2019-05-19 12:08:55 +00:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-only
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* linux/mm/page_alloc.c
|
|
|
|
*
|
|
|
|
* Manages the free list, the system allocates free pages here.
|
|
|
|
* Note that kmalloc() lives in slab.c
|
|
|
|
*
|
|
|
|
* Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
|
|
|
|
* Swap reorganised 29.12.95, Stephen Tweedie
|
|
|
|
* Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
|
|
|
|
* Reshaped it to be a zoned allocator, Ingo Molnar, Red Hat, 1999
|
|
|
|
* Discontiguous memory support, Kanoj Sarcar, SGI, Nov 1999
|
|
|
|
* Zone balancing, Kanoj Sarcar, SGI, Jan 2000
|
|
|
|
* Per cpu hot/cold page lists, bulk allocation, Martin J. Bligh, Sept 2002
|
|
|
|
* (lots of bits borrowed from Ingo Molnar & Andrew Morton)
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/stddef.h>
|
|
|
|
#include <linux/mm.h>
|
2018-12-28 08:34:29 +00:00
|
|
|
#include <linux/highmem.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/swap.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/pagemap.h>
|
2008-03-04 22:28:32 +00:00
|
|
|
#include <linux/jiffies.h>
|
2010-08-25 20:39:16 +00:00
|
|
|
#include <linux/memblock.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/compiler.h>
|
2005-09-13 08:25:16 +00:00
|
|
|
#include <linux/kernel.h>
|
2015-02-13 22:39:28 +00:00
|
|
|
#include <linux/kasan.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/suspend.h>
|
|
|
|
#include <linux/pagevec.h>
|
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/slab.h>
|
2011-05-25 00:12:16 +00:00
|
|
|
#include <linux/ratelimit.h>
|
2007-10-17 06:25:53 +00:00
|
|
|
#include <linux/oom.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/topology.h>
|
|
|
|
#include <linux/sysctl.h>
|
|
|
|
#include <linux/cpu.h>
|
|
|
|
#include <linux/cpuset.h>
|
2005-10-30 01:16:53 +00:00
|
|
|
#include <linux/memory_hotplug.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/nodemask.h>
|
|
|
|
#include <linux/vmalloc.h>
|
2011-05-25 00:11:33 +00:00
|
|
|
#include <linux/vmstat.h>
|
2006-01-06 08:11:17 +00:00
|
|
|
#include <linux/mempolicy.h>
|
2016-01-16 00:56:22 +00:00
|
|
|
#include <linux/memremap.h>
|
2006-06-23 09:03:11 +00:00
|
|
|
#include <linux/stop_machine.h>
|
2019-05-14 22:41:35 +00:00
|
|
|
#include <linux/random.h>
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
#include <linux/sort.h>
|
|
|
|
#include <linux/pfn.h>
|
2006-10-20 06:28:16 +00:00
|
|
|
#include <linux/backing-dev.h>
|
2006-12-08 10:39:45 +00:00
|
|
|
#include <linux/fault-inject.h>
|
2007-10-16 08:26:11 +00:00
|
|
|
#include <linux/page-isolation.h>
|
2008-04-30 07:55:01 +00:00
|
|
|
#include <linux/debugobjects.h>
|
2009-06-11 12:23:19 +00:00
|
|
|
#include <linux/kmemleak.h>
|
2010-05-24 21:32:30 +00:00
|
|
|
#include <linux/compaction.h>
|
2009-09-22 00:02:44 +00:00
|
|
|
#include <trace/events/kmem.h>
|
2017-02-22 23:42:00 +00:00
|
|
|
#include <trace/events/oom.h>
|
2011-05-20 19:50:29 +00:00
|
|
|
#include <linux/prefetch.h>
|
2013-09-11 21:22:36 +00:00
|
|
|
#include <linux/mm_inline.h>
|
2020-12-15 03:08:30 +00:00
|
|
|
#include <linux/mmu_notifier.h>
|
2011-12-29 12:09:50 +00:00
|
|
|
#include <linux/migrate.h>
|
2013-04-29 22:07:48 +00:00
|
|
|
#include <linux/hugetlb.h>
|
2013-02-07 15:47:07 +00:00
|
|
|
#include <linux/sched/rt.h>
|
2017-02-02 19:43:54 +00:00
|
|
|
#include <linux/sched/mm.h>
|
mm/page_owner: keep track of page owners
This is the page owner tracking code which is introduced so far ago. It
is resident on Andrew's tree, though, nobody tried to upstream so it
remain as is. Our company uses this feature actively to debug memory leak
or to find a memory hogger so I decide to upstream this feature.
This functionality help us to know who allocates the page. When
allocating a page, we store some information about allocation in extra
memory. Later, if we need to know status of all pages, we can get and
analyze it from this stored information.
In previous version of this feature, extra memory is statically defined in
struct page, but, in this version, extra memory is allocated outside of
struct page. It enables us to turn on/off this feature at boottime
without considerable memory waste.
Although we already have tracepoint for tracing page allocation/free,
using it to analyze page owner is rather complex. We need to enlarge the
trace buffer for preventing overlapping until userspace program launched.
And, launched program continually dump out the trace buffer for later
analysis and it would change system behaviour with more possibility rather
than just keeping it in memory, so bad for debug.
Moreover, we can use page_owner feature further for various purposes. For
example, we can use it for fragmentation statistics implemented in this
patch. And, I also plan to implement some CMA failure debugging feature
using this interface.
I'd like to give the credit for all developers contributed this feature,
but, it's not easy because I don't know exact history. Sorry about that.
Below is people who has "Signed-off-by" in the patches in Andrew's tree.
Contributor:
Alexander Nyberg <alexn@dsv.su.se>
Mel Gorman <mgorman@suse.de>
Dave Hansen <dave@linux.vnet.ibm.com>
Minchan Kim <minchan@kernel.org>
Michal Nazarewicz <mina86@mina86.com>
Andrew Morton <akpm@linux-foundation.org>
Jungsoo Son <jungsoo.son@lge.com>
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dave Hansen <dave@sr71.net>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Jungsoo Son <jungsoo.son@lge.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:01 +00:00
|
|
|
#include <linux/page_owner.h>
|
2015-06-30 21:57:27 +00:00
|
|
|
#include <linux/kthread.h>
|
mm: charge/uncharge kmemcg from generic page allocator paths
Currently, to charge a non-slab allocation to kmemcg one has to use
alloc_kmem_pages helper with __GFP_ACCOUNT flag. A page allocated with
this helper should finally be freed using free_kmem_pages, otherwise it
won't be uncharged.
This API suits its current users fine, but it turns out to be impossible
to use along with page reference counting, i.e. when an allocation is
supposed to be freed with put_page, as it is the case with pipe or unix
socket buffers.
To overcome this limitation, this patch moves charging/uncharging to
generic page allocator paths, i.e. to __alloc_pages_nodemask and
free_pages_prepare, and zaps alloc/free_kmem_pages helpers. This way,
one can use any of the available page allocation functions to get the
allocated page charged to kmemcg - it's enough to pass __GFP_ACCOUNT,
just like in case of kmalloc and friends. A charged page will be
automatically uncharged on free.
To make it possible, we need to mark pages charged to kmemcg somehow.
To avoid introducing a new page flag, we make use of page->_mapcount for
marking such pages. Since pages charged to kmemcg are not supposed to
be mapped to userspace, it should work just fine. There are other
(ab)users of page->_mapcount - buddy and balloon pages - but we don't
conflict with them.
In case kmemcg is compiled out or not used at runtime, this patch
introduces no overhead to generic page allocator paths. If kmemcg is
used, it will be plus one gfp flags check on alloc and plus one
page->_mapcount check on free, which shouldn't hurt performance, because
the data accessed are hot.
Link: http://lkml.kernel.org/r/a9736d856f895bcb465d9f257b54efe32eda6f99.1464079538.git.vdavydov@virtuozzo.com
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:24:24 +00:00
|
|
|
#include <linux/memcontrol.h>
|
2017-03-03 21:15:39 +00:00
|
|
|
#include <linux/ftrace.h>
|
2017-03-03 09:13:38 +00:00
|
|
|
#include <linux/lockdep.h>
|
2017-08-25 22:55:30 +00:00
|
|
|
#include <linux/nmi.h>
|
psi: pressure stall information for CPU, memory, and IO
When systems are overcommitted and resources become contended, it's hard
to tell exactly the impact this has on workload productivity, or how close
the system is to lockups and OOM kills. In particular, when machines work
multiple jobs concurrently, the impact of overcommit in terms of latency
and throughput on the individual job can be enormous.
In order to maximize hardware utilization without sacrificing individual
job health or risk complete machine lockups, this patch implements a way
to quantify resource pressure in the system.
A kernel built with CONFIG_PSI=y creates files in /proc/pressure/ that
expose the percentage of time the system is stalled on CPU, memory, or IO,
respectively. Stall states are aggregate versions of the per-task delay
accounting delays:
cpu: some tasks are runnable but not executing on a CPU
memory: tasks are reclaiming, or waiting for swapin or thrashing cache
io: tasks are waiting for io completions
These percentages of walltime can be thought of as pressure percentages,
and they give a general sense of system health and productivity loss
incurred by resource overcommit. They can also indicate when the system
is approaching lockup scenarios and OOMs.
To do this, psi keeps track of the task states associated with each CPU
and samples the time they spend in stall states. Every 2 seconds, the
samples are averaged across CPUs - weighted by the CPUs' non-idle time to
eliminate artifacts from unused CPUs - and translated into percentages of
walltime. A running average of those percentages is maintained over 10s,
1m, and 5m periods (similar to the loadaverage).
[hannes@cmpxchg.org: doc fixlet, per Randy]
Link: http://lkml.kernel.org/r/20180828205625.GA14030@cmpxchg.org
[hannes@cmpxchg.org: code optimization]
Link: http://lkml.kernel.org/r/20180907175015.GA8479@cmpxchg.org
[hannes@cmpxchg.org: rename psi_clock() to psi_update_work(), per Peter]
Link: http://lkml.kernel.org/r/20180907145404.GB11088@cmpxchg.org
[hannes@cmpxchg.org: fix build]
Link: http://lkml.kernel.org/r/20180913014222.GA2370@cmpxchg.org
Link: http://lkml.kernel.org/r/20180828172258.3185-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Daniel Drake <drake@endlessm.com>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:06:27 +00:00
|
|
|
#include <linux/psi.h>
|
mm: parallelize deferred_init_memmap()
Deferred struct page init is a significant bottleneck in kernel boot.
Optimizing it maximizes availability for large-memory systems and allows
spinning up short-lived VMs as needed without having to leave them
running. It also benefits bare metal machines hosting VMs that are
sensitive to downtime. In projects such as VMM Fast Restart[1], where
guest state is preserved across kexec reboot, it helps prevent application
and network timeouts in the guests.
Multithread to take full advantage of system memory bandwidth.
The maximum number of threads is capped at the number of CPUs on the node
because speedups always improve with additional threads on every system
tested, and at this phase of boot, the system is otherwise idle and
waiting on page init to finish.
Helper threads operate on section-aligned ranges to both avoid false
sharing when setting the pageblock's migrate type and to avoid accessing
uninitialized buddy pages, though max order alignment is enough for the
latter.
The minimum chunk size is also a section. There was benefit to using
multiple threads even on relatively small memory (1G) systems, and this is
the smallest size that the alignment allows.
The time (milliseconds) is the slowest node to initialize since boot
blocks until all nodes finish. intel_pstate is loaded in active mode
without hwp and with turbo enabled, and intel_idle is active as well.
Intel(R) Xeon(R) Platinum 8167M CPU @ 2.00GHz (Skylake, bare metal)
2 nodes * 26 cores * 2 threads = 104 CPUs
384G/node = 768G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 4089.7 ( 8.1) -- 1785.7 ( 7.6)
2% ( 1) 1.7% 4019.3 ( 1.5) 3.8% 1717.7 ( 11.8)
12% ( 6) 34.9% 2662.7 ( 2.9) 79.9% 359.3 ( 0.6)
25% ( 13) 39.9% 2459.0 ( 3.6) 91.2% 157.0 ( 0.0)
37% ( 19) 39.2% 2485.0 ( 29.7) 90.4% 172.0 ( 28.6)
50% ( 26) 39.3% 2482.7 ( 25.7) 90.3% 173.7 ( 30.0)
75% ( 39) 39.0% 2495.7 ( 5.5) 89.4% 190.0 ( 1.0)
100% ( 52) 40.2% 2443.7 ( 3.8) 92.3% 138.0 ( 1.0)
Intel(R) Xeon(R) CPU E5-2699C v4 @ 2.20GHz (Broadwell, kvm guest)
1 node * 16 cores * 2 threads = 32 CPUs
192G/node = 192G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1988.7 ( 9.6) -- 1096.0 ( 11.5)
3% ( 1) 1.1% 1967.0 ( 17.6) 0.3% 1092.7 ( 11.0)
12% ( 4) 41.1% 1170.3 ( 14.2) 73.8% 287.0 ( 3.6)
25% ( 8) 47.1% 1052.7 ( 21.9) 83.9% 177.0 ( 13.5)
38% ( 12) 48.9% 1016.3 ( 12.1) 86.8% 144.7 ( 1.5)
50% ( 16) 48.9% 1015.7 ( 8.1) 87.8% 134.0 ( 4.4)
75% ( 24) 49.1% 1012.3 ( 3.1) 88.1% 130.3 ( 2.3)
100% ( 32) 49.5% 1004.0 ( 5.3) 88.5% 125.7 ( 2.1)
Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz (Haswell, bare metal)
2 nodes * 18 cores * 2 threads = 72 CPUs
128G/node = 256G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1680.0 ( 4.6) -- 627.0 ( 4.0)
3% ( 1) 0.3% 1675.7 ( 4.5) -0.2% 628.0 ( 3.6)
11% ( 4) 25.6% 1250.7 ( 2.1) 67.9% 201.0 ( 0.0)
25% ( 9) 30.7% 1164.0 ( 17.3) 81.8% 114.3 ( 17.7)
36% ( 13) 31.4% 1152.7 ( 10.8) 84.0% 100.3 ( 17.9)
50% ( 18) 31.5% 1150.7 ( 9.3) 83.9% 101.0 ( 14.1)
75% ( 27) 31.7% 1148.0 ( 5.6) 84.5% 97.3 ( 6.4)
100% ( 36) 32.0% 1142.3 ( 4.0) 85.6% 90.0 ( 1.0)
AMD EPYC 7551 32-Core Processor (Zen, kvm guest)
1 node * 8 cores * 2 threads = 16 CPUs
64G/node = 64G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1029.3 ( 25.1) -- 240.7 ( 1.5)
6% ( 1) -0.6% 1036.0 ( 7.8) -2.2% 246.0 ( 0.0)
12% ( 2) 11.8% 907.7 ( 8.6) 44.7% 133.0 ( 1.0)
25% ( 4) 13.9% 886.0 ( 10.6) 62.6% 90.0 ( 6.0)
38% ( 6) 17.8% 845.7 ( 14.2) 69.1% 74.3 ( 3.8)
50% ( 8) 16.8% 856.0 ( 22.1) 72.9% 65.3 ( 5.7)
75% ( 12) 15.4% 871.0 ( 29.2) 79.8% 48.7 ( 7.4)
100% ( 16) 21.0% 813.7 ( 21.0) 80.5% 47.0 ( 5.2)
Server-oriented distros that enable deferred page init sometimes run in
small VMs, and they still benefit even though the fraction of boot time
saved is smaller:
AMD EPYC 7551 32-Core Processor (Zen, kvm guest)
1 node * 2 cores * 2 threads = 4 CPUs
16G/node = 16G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 716.0 ( 14.0) -- 49.7 ( 0.6)
25% ( 1) 1.8% 703.0 ( 5.3) -4.0% 51.7 ( 0.6)
50% ( 2) 1.6% 704.7 ( 1.2) 43.0% 28.3 ( 0.6)
75% ( 3) 2.7% 696.7 ( 13.1) 49.7% 25.0 ( 0.0)
100% ( 4) 4.1% 687.0 ( 10.4) 55.7% 22.0 ( 0.0)
Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz (Haswell, kvm guest)
1 node * 2 cores * 2 threads = 4 CPUs
14G/node = 14G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 787.7 ( 6.4) -- 122.3 ( 0.6)
25% ( 1) 0.2% 786.3 ( 10.8) -2.5% 125.3 ( 2.1)
50% ( 2) 5.9% 741.0 ( 13.9) 37.6% 76.3 ( 19.7)
75% ( 3) 8.3% 722.0 ( 19.0) 49.9% 61.3 ( 3.2)
100% ( 4) 9.3% 714.7 ( 9.5) 56.4% 53.3 ( 1.5)
On Josh's 96-CPU and 192G memory system:
Without this patch series:
[ 0.487132] node 0 initialised, 23398907 pages in 292ms
[ 0.499132] node 1 initialised, 24189223 pages in 304ms
...
[ 0.629376] Run /sbin/init as init process
With this patch series:
[ 0.231435] node 1 initialised, 24189223 pages in 32ms
[ 0.236718] node 0 initialised, 23398907 pages in 36ms
[1] https://static.sched.com/hosted_files/kvmforum2019/66/VMM-fast-restart_kvmforum2019.pdf
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Robert Elliott <elliott@hpe.com>
Cc: Shile Zhang <shile.zhang@linux.alibaba.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Zi Yan <ziy@nvidia.com>
Link: http://lkml.kernel.org/r/20200527173608.2885243-7-daniel.m.jordan@oracle.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-03 22:59:51 +00:00
|
|
|
#include <linux/padata.h>
|
2020-10-11 06:16:40 +00:00
|
|
|
#include <linux/khugepaged.h>
|
init/main: fix broken buffer_init when DEFERRED_STRUCT_PAGE_INIT set
In the booting phase if CONFIG_DEFERRED_STRUCT_PAGE_INIT is set,
we have following callchain:
start_kernel
...
mm_init
mem_init
memblock_free_all
reset_all_zones_managed_pages
free_low_memory_core_early
...
buffer_init
nr_free_buffer_pages
zone->managed_pages
...
rest_init
kernel_init
kernel_init_freeable
page_alloc_init_late
kthread_run(deferred_init_memmap, NODE_DATA(nid), "pgdatinit%d", nid);
wait_for_completion(&pgdat_init_all_done_comp);
...
files_maxfiles_init
It's clear that buffer_init depends on zone->managed_pages, but it's reset
in reset_all_zones_managed_pages after that pages are readded into
zone->managed_pages, but when buffer_init runs this process is half done
and most of them will finally be added till deferred_init_memmap done. In
large memory couting of nr_free_buffer_pages drifts too much, also
drifting from kernels to kernels on same hardware.
Fix is simple, it delays buffer_init run till deferred_init_memmap all
done.
But as corrected by this patch, max_buffer_heads becomes very large, the
value is roughly as many as 4 times of totalram_pages, formula:
max_buffer_heads = nrpages * (10%) * (PAGE_SIZE / sizeof(struct
buffer_head));
Say in a 64GB memory box we have 16777216 pages, then max_buffer_heads
turns out to be roughly 67,108,864. In common cases, should a buffer_head
be mapped to one page/block(4KB)? So max_buffer_heads never exceeds
totalram_pages. IMO it's likely to make buffer_heads_over_limit bool
value alwasy false, then make codes 'if (buffer_heads_over_limit)' test in
vmscan unnecessary.
So this patch will change the original behavior related to
buffer_heads_over_limit in vmscan since we used a half done value of
zone->managed_pages before, or should we use a smaller factor(<10%) in
previous formula.
akpm: I think this is OK - the max_buffer_heads code is only needed on
highmem machines, to prevent ZONE_NORMAL from being consumed by large
amounts of buffer_heads attached to highmem pagecache. This problem will
not occur on 64-bit machines, so this feature's non-functionality on such
machines is a feature, not a bug.
Link: https://lkml.kernel.org/r/20201123110500.103523-1-linf@wangsu.com
Signed-off-by: Lin Feng <linf@wangsu.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:11:19 +00:00
|
|
|
#include <linux/buffer_head.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-07-03 22:03:41 +00:00
|
|
|
#include <asm/sections.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <asm/tlbflush.h>
|
2006-05-15 16:43:59 +00:00
|
|
|
#include <asm/div64.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include "internal.h"
|
mm: shuffle initial free memory to improve memory-side-cache utilization
Patch series "mm: Randomize free memory", v10.
This patch (of 3):
Randomization of the page allocator improves the average utilization of
a direct-mapped memory-side-cache. Memory side caching is a platform
capability that Linux has been previously exposed to in HPC
(high-performance computing) environments on specialty platforms. In
that instance it was a smaller pool of high-bandwidth-memory relative to
higher-capacity / lower-bandwidth DRAM. Now, this capability is going
to be found on general purpose server platforms where DRAM is a cache in
front of higher latency persistent memory [1].
Robert offered an explanation of the state of the art of Linux
interactions with memory-side-caches [2], and I copy it here:
It's been a problem in the HPC space:
http://www.nersc.gov/research-and-development/knl-cache-mode-performance-coe/
A kernel module called zonesort is available to try to help:
https://software.intel.com/en-us/articles/xeon-phi-software
and this abandoned patch series proposed that for the kernel:
https://lkml.kernel.org/r/20170823100205.17311-1-lukasz.daniluk@intel.com
Dan's patch series doesn't attempt to ensure buffers won't conflict, but
also reduces the chance that the buffers will. This will make performance
more consistent, albeit slower than "optimal" (which is near impossible
to attain in a general-purpose kernel). That's better than forcing
users to deploy remedies like:
"To eliminate this gradual degradation, we have added a Stream
measurement to the Node Health Check that follows each job;
nodes are rebooted whenever their measured memory bandwidth
falls below 300 GB/s."
A replacement for zonesort was merged upstream in commit cc9aec03e58f
("x86/numa_emulation: Introduce uniform split capability"). With this
numa_emulation capability, memory can be split into cache sized
("near-memory" sized) numa nodes. A bind operation to such a node, and
disabling workloads on other nodes, enables full cache performance.
However, once the workload exceeds the cache size then cache conflicts
are unavoidable. While HPC environments might be able to tolerate
time-scheduling of cache sized workloads, for general purpose server
platforms, the oversubscribed cache case will be the common case.
The worst case scenario is that a server system owner benchmarks a
workload at boot with an un-contended cache only to see that performance
degrade over time, even below the average cache performance due to
excessive conflicts. Randomization clips the peaks and fills in the
valleys of cache utilization to yield steady average performance.
Here are some performance impact details of the patches:
1/ An Intel internal synthetic memory bandwidth measurement tool, saw a
3X speedup in a contrived case that tries to force cache conflicts.
The contrived cased used the numa_emulation capability to force an
instance of the benchmark to be run in two of the near-memory sized
numa nodes. If both instances were placed on the same emulated they
would fit and cause zero conflicts. While on separate emulated nodes
without randomization they underutilized the cache and conflicted
unnecessarily due to the in-order allocation per node.
2/ A well known Java server application benchmark was run with a heap
size that exceeded cache size by 3X. The cache conflict rate was 8%
for the first run and degraded to 21% after page allocator aging. With
randomization enabled the rate levelled out at 11%.
3/ A MongoDB workload did not observe measurable difference in
cache-conflict rates, but the overall throughput dropped by 7% with
randomization in one case.
4/ Mel Gorman ran his suite of performance workloads with randomization
enabled on platforms without a memory-side-cache and saw a mix of some
improvements and some losses [3].
While there is potentially significant improvement for applications that
depend on low latency access across a wide working-set, the performance
may be negligible to negative for other workloads. For this reason the
shuffle capability defaults to off unless a direct-mapped
memory-side-cache is detected. Even then, the page_alloc.shuffle=0
parameter can be specified to disable the randomization on those systems.
Outside of memory-side-cache utilization concerns there is potentially
security benefit from randomization. Some data exfiltration and
return-oriented-programming attacks rely on the ability to infer the
location of sensitive data objects. The kernel page allocator, especially
early in system boot, has predictable first-in-first out behavior for
physical pages. Pages are freed in physical address order when first
onlined.
Quoting Kees:
"While we already have a base-address randomization
(CONFIG_RANDOMIZE_MEMORY), attacks against the same hardware and
memory layouts would certainly be using the predictability of
allocation ordering (i.e. for attacks where the base address isn't
important: only the relative positions between allocated memory).
This is common in lots of heap-style attacks. They try to gain
control over ordering by spraying allocations, etc.
I'd really like to see this because it gives us something similar
to CONFIG_SLAB_FREELIST_RANDOM but for the page allocator."
While SLAB_FREELIST_RANDOM reduces the predictability of some local slab
caches it leaves vast bulk of memory to be predictably in order allocated.
However, it should be noted, the concrete security benefits are hard to
quantify, and no known CVE is mitigated by this randomization.
Introduce shuffle_free_memory(), and its helper shuffle_zone(), to perform
a Fisher-Yates shuffle of the page allocator 'free_area' lists when they
are initially populated with free memory at boot and at hotplug time. Do
this based on either the presence of a page_alloc.shuffle=Y command line
parameter, or autodetection of a memory-side-cache (to be added in a
follow-on patch).
The shuffling is done in terms of CONFIG_SHUFFLE_PAGE_ORDER sized free
pages where the default CONFIG_SHUFFLE_PAGE_ORDER is MAX_ORDER-1 i.e. 10,
4MB this trades off randomization granularity for time spent shuffling.
MAX_ORDER-1 was chosen to be minimally invasive to the page allocator
while still showing memory-side cache behavior improvements, and the
expectation that the security implications of finer granularity
randomization is mitigated by CONFIG_SLAB_FREELIST_RANDOM. The
performance impact of the shuffling appears to be in the noise compared to
other memory initialization work.
This initial randomization can be undone over time so a follow-on patch is
introduced to inject entropy on page free decisions. It is reasonable to
ask if the page free entropy is sufficient, but it is not enough due to
the in-order initial freeing of pages. At the start of that process
putting page1 in front or behind page0 still keeps them close together,
page2 is still near page1 and has a high chance of being adjacent. As
more pages are added ordering diversity improves, but there is still high
page locality for the low address pages and this leads to no significant
impact to the cache conflict rate.
[1]: https://itpeernetwork.intel.com/intel-optane-dc-persistent-memory-operating-modes/
[2]: https://lkml.kernel.org/r/AT5PR8401MB1169D656C8B5E121752FC0F8AB120@AT5PR8401MB1169.NAMPRD84.PROD.OUTLOOK.COM
[3]: https://lkml.org/lkml/2018/10/12/309
[dan.j.williams@intel.com: fix shuffle enable]
Link: http://lkml.kernel.org/r/154943713038.3858443.4125180191382062871.stgit@dwillia2-desk3.amr.corp.intel.com
[cai@lca.pw: fix SHUFFLE_PAGE_ALLOCATOR help texts]
Link: http://lkml.kernel.org/r/20190425201300.75650-1-cai@lca.pw
Link: http://lkml.kernel.org/r/154899811738.3165233.12325692939590944259.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Robert Elliott <elliott@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 22:41:28 +00:00
|
|
|
#include "shuffle.h"
|
2020-04-07 03:04:56 +00:00
|
|
|
#include "page_reporting.h"
|
2005-04-16 22:20:36 +00:00
|
|
|
|
mm/page_alloc: convert "report" flag of __free_one_page() to a proper flag
Patch series "mm: place pages to the freelist tail when onlining and undoing isolation", v2.
When adding separate memory blocks via add_memory*() and onlining them
immediately, the metadata (especially the memmap) of the next block will
be placed onto one of the just added+onlined block. This creates a chain
of unmovable allocations: If the last memory block cannot get
offlined+removed() so will all dependent ones. We directly have unmovable
allocations all over the place.
This can be observed quite easily using virtio-mem, however, it can also
be observed when using DIMMs. The freshly onlined pages will usually be
placed to the head of the freelists, meaning they will be allocated next,
turning the just-added memory usually immediately un-removable. The fresh
pages are cold, prefering to allocate others (that might be hot) also
feels to be the natural thing to do.
It also applies to the hyper-v balloon xen-balloon, and ppc64 dlpar: when
adding separate, successive memory blocks, each memory block will have
unmovable allocations on them - for example gigantic pages will fail to
allocate.
While the ZONE_NORMAL doesn't provide any guarantees that memory can get
offlined+removed again (any kind of fragmentation with unmovable
allocations is possible), there are many scenarios (hotplugging a lot of
memory, running workload, hotunplug some memory/as much as possible) where
we can offline+remove quite a lot with this patchset.
a) To visualize the problem, a very simple example:
Start a VM with 4GB and 8GB of virtio-mem memory:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000033fffffff 9G online yes 32-103
Memory block size: 128M
Total online memory: 12G
Total offline memory: 0B
Then try to unplug as much as possible using virtio-mem. Observe which
memory blocks are still around. Without this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
0x0000000148000000-0x000000014fffffff 128M online yes 41
0x0000000158000000-0x000000015fffffff 128M online yes 43
0x0000000168000000-0x000000016fffffff 128M online yes 45
0x0000000178000000-0x000000017fffffff 128M online yes 47
0x0000000188000000-0x0000000197ffffff 256M online yes 49-50
0x00000001a0000000-0x00000001a7ffffff 128M online yes 52
0x00000001b0000000-0x00000001b7ffffff 128M online yes 54
0x00000001c0000000-0x00000001c7ffffff 128M online yes 56
0x00000001d0000000-0x00000001d7ffffff 128M online yes 58
0x00000001e0000000-0x00000001e7ffffff 128M online yes 60
0x00000001f0000000-0x00000001f7ffffff 128M online yes 62
0x0000000200000000-0x0000000207ffffff 128M online yes 64
0x0000000210000000-0x0000000217ffffff 128M online yes 66
0x0000000220000000-0x0000000227ffffff 128M online yes 68
0x0000000230000000-0x0000000237ffffff 128M online yes 70
0x0000000240000000-0x0000000247ffffff 128M online yes 72
0x0000000250000000-0x0000000257ffffff 128M online yes 74
0x0000000260000000-0x0000000267ffffff 128M online yes 76
0x0000000270000000-0x0000000277ffffff 128M online yes 78
0x0000000280000000-0x0000000287ffffff 128M online yes 80
0x0000000290000000-0x0000000297ffffff 128M online yes 82
0x00000002a0000000-0x00000002a7ffffff 128M online yes 84
0x00000002b0000000-0x00000002b7ffffff 128M online yes 86
0x00000002c0000000-0x00000002c7ffffff 128M online yes 88
0x00000002d0000000-0x00000002d7ffffff 128M online yes 90
0x00000002e0000000-0x00000002e7ffffff 128M online yes 92
0x00000002f0000000-0x00000002f7ffffff 128M online yes 94
0x0000000300000000-0x0000000307ffffff 128M online yes 96
0x0000000310000000-0x0000000317ffffff 128M online yes 98
0x0000000320000000-0x0000000327ffffff 128M online yes 100
0x0000000330000000-0x000000033fffffff 256M online yes 102-103
Memory block size: 128M
Total online memory: 8.1G
Total offline memory: 0B
With this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
Memory block size: 128M
Total online memory: 4G
Total offline memory: 0B
All memory can get unplugged, all memory block can get removed. Of
course, no workload ran and the system was basically idle, but it
highlights the issue - the fairly deterministic chain of unmovable
allocations. When a huge page for the 2MB memmap is needed, a
just-onlined 4MB page will be split. The remaining 2MB page will be used
for the memmap of the next memory block. So one memory block will hold
the memmap of the two following memory blocks. Finally the pages of the
last-onlined memory block will get used for the next bigger allocations -
if any allocation is unmovable, all dependent memory blocks cannot get
unplugged and removed until that allocation is gone.
Note that with bigger memory blocks (e.g., 256MB), *all* memory
blocks are dependent and none can get unplugged again!
b) Experiment with memory intensive workload
I performed an experiment with an older version of this patch set (before
we used undo_isolate_page_range() in online_pages(): Hotplug 56GB to a VM
with an initial 4GB, onlining all memory to ZONE_NORMAL right from the
kernel when adding it. I then run various memory intensive workloads that
consume most system memory for a total of 45 minutes. Once finished, I
try to unplug as much memory as possible.
With this change, I am able to remove via virtio-mem (adding individual
128MB memory blocks) 413 out of 448 added memory blocks. Via individual
(256MB) DIMMs 380 out of 448 added memory blocks. (I don't have any
numbers without this patchset, but looking at the above example, it's at
most half of the 448 memory blocks for virtio-mem, and most probably none
for DIMMs).
Again, there are workloads that might behave very differently due to the
nature of ZONE_NORMAL.
This change also affects (besides memory onlining):
- Other users of undo_isolate_page_range(): Pages are always placed to the
tail.
-- When memory offlining fails
-- When memory isolation fails after having isolated some pageblocks
-- When alloc_contig_range() either succeeds or fails
- Other users of __putback_isolated_page(): Pages are always placed to the
tail.
-- Free page reporting
- Other users of __free_pages_core()
-- AFAIKs, any memory that is getting exposed to the buddy during boot.
IIUC we will now usually allocate memory from lower addresses within
a zone first (especially during boot).
- Other users of generic_online_page()
-- Hyper-V balloon
This patch (of 5):
Let's prepare for additional flags and avoid long parameter lists of
bools. Follow-up patches will also make use of the flags in
__free_pages_ok().
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-1-david@redhat.com
Link: https://lkml.kernel.org/r/20201005121534.15649-2-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:20 +00:00
|
|
|
/* Free Page Internal flags: for internal, non-pcp variants of free_pages(). */
|
|
|
|
typedef int __bitwise fpi_t;
|
|
|
|
|
|
|
|
/* No special request */
|
|
|
|
#define FPI_NONE ((__force fpi_t)0)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Skip free page reporting notification for the (possibly merged) page.
|
|
|
|
* This does not hinder free page reporting from grabbing the page,
|
|
|
|
* reporting it and marking it "reported" - it only skips notifying
|
|
|
|
* the free page reporting infrastructure about a newly freed page. For
|
|
|
|
* example, used when temporarily pulling a page from a freelist and
|
|
|
|
* putting it back unmodified.
|
|
|
|
*/
|
|
|
|
#define FPI_SKIP_REPORT_NOTIFY ((__force fpi_t)BIT(0))
|
|
|
|
|
mm/page_alloc: place pages to tail in __putback_isolated_page()
__putback_isolated_page() already documents that pages will be placed to
the tail of the freelist - this is, however, not the case for "order >=
MAX_ORDER - 2" (see buddy_merge_likely()) - which should be the case for
all existing users.
This change affects two users:
- free page reporting
- page isolation, when undoing the isolation (including memory onlining).
This behavior is desirable for pages that haven't really been touched
lately, so exactly the two users that don't actually read/write page
content, but rather move untouched pages.
The new behavior is especially desirable for memory onlining, where we
allow allocation of newly onlined pages via undo_isolate_page_range() in
online_pages(). Right now, we always place them to the head of the
freelist, resulting in undesireable behavior: Assume we add individual
memory chunks via add_memory() and online them right away to the NORMAL
zone. We create a dependency chain of unmovable allocations e.g., via the
memmap. The memmap of the next chunk will be placed onto previous chunks
- if the last block cannot get offlined+removed, all dependent ones cannot
get offlined+removed. While this can already be observed with individual
DIMMs, it's more of an issue for virtio-mem (and I suspect also ppc
DLPAR).
Document that this should only be used for optimizations, and no code
should rely on this behavior for correction (if the order of the freelists
ever changes).
We won't care about page shuffling: memory onlining already properly
shuffles after onlining. free page reporting doesn't care about
physically contiguous ranges, and there are already cases where page
isolation will simply move (physically close) free pages to (currently)
the head of the freelists via move_freepages_block() instead of shuffling.
If this becomes ever relevant, we should shuffle the whole zone when
undoing isolation of larger ranges, and after free_contig_range().
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-3-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:26 +00:00
|
|
|
/*
|
|
|
|
* Place the (possibly merged) page to the tail of the freelist. Will ignore
|
|
|
|
* page shuffling (relevant code - e.g., memory onlining - is expected to
|
|
|
|
* shuffle the whole zone).
|
|
|
|
*
|
|
|
|
* Note: No code should rely on this flag for correctness - it's purely
|
|
|
|
* to allow for optimizations when handing back either fresh pages
|
|
|
|
* (memory onlining) or untouched pages (page isolation, free page
|
|
|
|
* reporting).
|
|
|
|
*/
|
|
|
|
#define FPI_TO_TAIL ((__force fpi_t)BIT(1))
|
|
|
|
|
2013-07-03 22:01:29 +00:00
|
|
|
/* prevent >1 _updater_ of zone percpu pageset ->high and ->batch fields */
|
|
|
|
static DEFINE_MUTEX(pcp_batch_high_lock);
|
2014-06-23 20:22:04 +00:00
|
|
|
#define MIN_PERCPU_PAGELIST_FRACTION (8)
|
2013-07-03 22:01:29 +00:00
|
|
|
|
2010-05-26 21:44:56 +00:00
|
|
|
#ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
|
|
|
|
DEFINE_PER_CPU(int, numa_node);
|
|
|
|
EXPORT_PER_CPU_SYMBOL(numa_node);
|
|
|
|
#endif
|
|
|
|
|
2017-11-16 01:38:22 +00:00
|
|
|
DEFINE_STATIC_KEY_TRUE(vm_numa_stat_key);
|
|
|
|
|
2010-05-26 21:45:00 +00:00
|
|
|
#ifdef CONFIG_HAVE_MEMORYLESS_NODES
|
|
|
|
/*
|
|
|
|
* N.B., Do NOT reference the '_numa_mem_' per cpu variable directly.
|
|
|
|
* It will not be defined when CONFIG_HAVE_MEMORYLESS_NODES is not defined.
|
|
|
|
* Use the accessor functions set_numa_mem(), numa_mem_id() and cpu_to_mem()
|
|
|
|
* defined in <linux/topology.h>.
|
|
|
|
*/
|
|
|
|
DEFINE_PER_CPU(int, _numa_mem_); /* Kernel "local memory" node */
|
|
|
|
EXPORT_PER_CPU_SYMBOL(_numa_mem_);
|
|
|
|
#endif
|
|
|
|
|
2017-02-24 22:56:56 +00:00
|
|
|
/* work_structs for global per-cpu drains */
|
2018-12-28 08:38:58 +00:00
|
|
|
struct pcpu_drain {
|
|
|
|
struct zone *zone;
|
|
|
|
struct work_struct work;
|
|
|
|
};
|
2020-04-10 21:32:32 +00:00
|
|
|
static DEFINE_MUTEX(pcpu_drain_mutex);
|
|
|
|
static DEFINE_PER_CPU(struct pcpu_drain, pcpu_drain);
|
2017-02-24 22:56:56 +00:00
|
|
|
|
gcc-plugins: Add latent_entropy plugin
This adds a new gcc plugin named "latent_entropy". It is designed to
extract as much possible uncertainty from a running system at boot time as
possible, hoping to capitalize on any possible variation in CPU operation
(due to runtime data differences, hardware differences, SMP ordering,
thermal timing variation, cache behavior, etc).
At the very least, this plugin is a much more comprehensive example for
how to manipulate kernel code using the gcc plugin internals.
The need for very-early boot entropy tends to be very architecture or
system design specific, so this plugin is more suited for those sorts
of special cases. The existing kernel RNG already attempts to extract
entropy from reliable runtime variation, but this plugin takes the idea to
a logical extreme by permuting a global variable based on any variation
in code execution (e.g. a different value (and permutation function)
is used to permute the global based on loop count, case statement,
if/then/else branching, etc).
To do this, the plugin starts by inserting a local variable in every
marked function. The plugin then adds logic so that the value of this
variable is modified by randomly chosen operations (add, xor and rol) and
random values (gcc generates separate static values for each location at
compile time and also injects the stack pointer at runtime). The resulting
value depends on the control flow path (e.g., loops and branches taken).
Before the function returns, the plugin mixes this local variable into
the latent_entropy global variable. The value of this global variable
is added to the kernel entropy pool in do_one_initcall() and _do_fork(),
though it does not credit any bytes of entropy to the pool; the contents
of the global are just used to mix the pool.
Additionally, the plugin can pre-initialize arrays with build-time
random contents, so that two different kernel builds running on identical
hardware will not have the same starting values.
Signed-off-by: Emese Revfy <re.emese@gmail.com>
[kees: expanded commit message and code comments]
Signed-off-by: Kees Cook <keescook@chromium.org>
2016-06-20 18:41:19 +00:00
|
|
|
#ifdef CONFIG_GCC_PLUGIN_LATENT_ENTROPY
|
2016-10-18 22:08:04 +00:00
|
|
|
volatile unsigned long latent_entropy __latent_entropy;
|
gcc-plugins: Add latent_entropy plugin
This adds a new gcc plugin named "latent_entropy". It is designed to
extract as much possible uncertainty from a running system at boot time as
possible, hoping to capitalize on any possible variation in CPU operation
(due to runtime data differences, hardware differences, SMP ordering,
thermal timing variation, cache behavior, etc).
At the very least, this plugin is a much more comprehensive example for
how to manipulate kernel code using the gcc plugin internals.
The need for very-early boot entropy tends to be very architecture or
system design specific, so this plugin is more suited for those sorts
of special cases. The existing kernel RNG already attempts to extract
entropy from reliable runtime variation, but this plugin takes the idea to
a logical extreme by permuting a global variable based on any variation
in code execution (e.g. a different value (and permutation function)
is used to permute the global based on loop count, case statement,
if/then/else branching, etc).
To do this, the plugin starts by inserting a local variable in every
marked function. The plugin then adds logic so that the value of this
variable is modified by randomly chosen operations (add, xor and rol) and
random values (gcc generates separate static values for each location at
compile time and also injects the stack pointer at runtime). The resulting
value depends on the control flow path (e.g., loops and branches taken).
Before the function returns, the plugin mixes this local variable into
the latent_entropy global variable. The value of this global variable
is added to the kernel entropy pool in do_one_initcall() and _do_fork(),
though it does not credit any bytes of entropy to the pool; the contents
of the global are just used to mix the pool.
Additionally, the plugin can pre-initialize arrays with build-time
random contents, so that two different kernel builds running on identical
hardware will not have the same starting values.
Signed-off-by: Emese Revfy <re.emese@gmail.com>
[kees: expanded commit message and code comments]
Signed-off-by: Kees Cook <keescook@chromium.org>
2016-06-20 18:41:19 +00:00
|
|
|
EXPORT_SYMBOL(latent_entropy);
|
|
|
|
#endif
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
Memoryless nodes: Generic management of nodemasks for various purposes
Why do we need to support memoryless nodes?
KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> For fujitsu, problem is called "empty" node.
>
> When ACPI's SRAT table includes "possible nodes", ia64 bootstrap(acpi_numa_init)
> creates nodes, which includes no memory, no cpu.
>
> I tried to remove empty-node in past, but that was denied.
> It was because we can hot-add cpu to the empty node.
> (node-hotplug triggered by cpu is not implemented now. and it will be ugly.)
>
>
> For HP, (Lee can comment on this later), they have memory-less-node.
> As far as I hear, HP's machine can have following configration.
>
> (example)
> Node0: CPU0 memory AAA MB
> Node1: CPU1 memory AAA MB
> Node2: CPU2 memory AAA MB
> Node3: CPU3 memory AAA MB
> Node4: Memory XXX GB
>
> AAA is very small value (below 16MB) and will be omitted by ia64 bootstrap.
> After boot, only Node 4 has valid memory (but have no cpu.)
>
> Maybe this is memory-interleave by firmware config.
Christoph Lameter <clameter@sgi.com> wrote:
> Future SGI platforms (actually also current one can have but nothing like
> that is deployed to my knowledge) have nodes with only cpus. Current SGI
> platforms have nodes with just I/O that we so far cannot manage in the
> core. So the arch code maps them to the nearest memory node.
Lee Schermerhorn <Lee.Schermerhorn@hp.com> wrote:
> For the HP platforms, we can configure each cell with from 0% to 100%
> "cell local memory". When we configure with <100% CLM, the "missing
> percentages" are interleaved by hardware on a cache-line granularity to
> improve bandwidth at the expense of latency for numa-challenged
> applications [and OSes, but not our problem ;-)]. When we boot Linux on
> such a config, all of the real nodes have no memory--it all resides in a
> single interleaved pseudo-node.
>
> When we boot Linux on a 100% CLM configuration [== NUMA], we still have
> the interleaved pseudo-node. It contains a few hundred MB stolen from
> the real nodes to contain the DMA zone. [Interleaved memory resides at
> phys addr 0]. The memoryless-nodes patches, along with the zoneorder
> patches, support this config as well.
>
> Also, when we boot a NUMA config with the "mem=" command line,
> specifying less memory than actually exists, Linux takes the excluded
> memory "off the top" rather than distributing it across the nodes. This
> can result in memoryless nodes, as well.
>
This patch:
Preparation for memoryless node patches.
Provide a generic way to keep nodemasks describing various characteristics of
NUMA nodes.
Remove the node_online_map and the node_possible map and realize the same
functionality using two nodes stats: N_POSSIBLE and N_ONLINE.
[Lee.Schermerhorn@hp.com: Initialize N_*_MEMORY and N_CPU masks for non-NUMA config]
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Tested-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Bob Picco <bob.picco@hp.com>
Cc: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@skynet.ie>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:27 +00:00
|
|
|
* Array of node states.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
Memoryless nodes: Generic management of nodemasks for various purposes
Why do we need to support memoryless nodes?
KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> For fujitsu, problem is called "empty" node.
>
> When ACPI's SRAT table includes "possible nodes", ia64 bootstrap(acpi_numa_init)
> creates nodes, which includes no memory, no cpu.
>
> I tried to remove empty-node in past, but that was denied.
> It was because we can hot-add cpu to the empty node.
> (node-hotplug triggered by cpu is not implemented now. and it will be ugly.)
>
>
> For HP, (Lee can comment on this later), they have memory-less-node.
> As far as I hear, HP's machine can have following configration.
>
> (example)
> Node0: CPU0 memory AAA MB
> Node1: CPU1 memory AAA MB
> Node2: CPU2 memory AAA MB
> Node3: CPU3 memory AAA MB
> Node4: Memory XXX GB
>
> AAA is very small value (below 16MB) and will be omitted by ia64 bootstrap.
> After boot, only Node 4 has valid memory (but have no cpu.)
>
> Maybe this is memory-interleave by firmware config.
Christoph Lameter <clameter@sgi.com> wrote:
> Future SGI platforms (actually also current one can have but nothing like
> that is deployed to my knowledge) have nodes with only cpus. Current SGI
> platforms have nodes with just I/O that we so far cannot manage in the
> core. So the arch code maps them to the nearest memory node.
Lee Schermerhorn <Lee.Schermerhorn@hp.com> wrote:
> For the HP platforms, we can configure each cell with from 0% to 100%
> "cell local memory". When we configure with <100% CLM, the "missing
> percentages" are interleaved by hardware on a cache-line granularity to
> improve bandwidth at the expense of latency for numa-challenged
> applications [and OSes, but not our problem ;-)]. When we boot Linux on
> such a config, all of the real nodes have no memory--it all resides in a
> single interleaved pseudo-node.
>
> When we boot Linux on a 100% CLM configuration [== NUMA], we still have
> the interleaved pseudo-node. It contains a few hundred MB stolen from
> the real nodes to contain the DMA zone. [Interleaved memory resides at
> phys addr 0]. The memoryless-nodes patches, along with the zoneorder
> patches, support this config as well.
>
> Also, when we boot a NUMA config with the "mem=" command line,
> specifying less memory than actually exists, Linux takes the excluded
> memory "off the top" rather than distributing it across the nodes. This
> can result in memoryless nodes, as well.
>
This patch:
Preparation for memoryless node patches.
Provide a generic way to keep nodemasks describing various characteristics of
NUMA nodes.
Remove the node_online_map and the node_possible map and realize the same
functionality using two nodes stats: N_POSSIBLE and N_ONLINE.
[Lee.Schermerhorn@hp.com: Initialize N_*_MEMORY and N_CPU masks for non-NUMA config]
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Tested-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Bob Picco <bob.picco@hp.com>
Cc: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@skynet.ie>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:27 +00:00
|
|
|
nodemask_t node_states[NR_NODE_STATES] __read_mostly = {
|
|
|
|
[N_POSSIBLE] = NODE_MASK_ALL,
|
|
|
|
[N_ONLINE] = { { [0] = 1UL } },
|
|
|
|
#ifndef CONFIG_NUMA
|
|
|
|
[N_NORMAL_MEMORY] = { { [0] = 1UL } },
|
|
|
|
#ifdef CONFIG_HIGHMEM
|
|
|
|
[N_HIGH_MEMORY] = { { [0] = 1UL } },
|
2012-12-12 21:52:00 +00:00
|
|
|
#endif
|
|
|
|
[N_MEMORY] = { { [0] = 1UL } },
|
Memoryless nodes: Generic management of nodemasks for various purposes
Why do we need to support memoryless nodes?
KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> For fujitsu, problem is called "empty" node.
>
> When ACPI's SRAT table includes "possible nodes", ia64 bootstrap(acpi_numa_init)
> creates nodes, which includes no memory, no cpu.
>
> I tried to remove empty-node in past, but that was denied.
> It was because we can hot-add cpu to the empty node.
> (node-hotplug triggered by cpu is not implemented now. and it will be ugly.)
>
>
> For HP, (Lee can comment on this later), they have memory-less-node.
> As far as I hear, HP's machine can have following configration.
>
> (example)
> Node0: CPU0 memory AAA MB
> Node1: CPU1 memory AAA MB
> Node2: CPU2 memory AAA MB
> Node3: CPU3 memory AAA MB
> Node4: Memory XXX GB
>
> AAA is very small value (below 16MB) and will be omitted by ia64 bootstrap.
> After boot, only Node 4 has valid memory (but have no cpu.)
>
> Maybe this is memory-interleave by firmware config.
Christoph Lameter <clameter@sgi.com> wrote:
> Future SGI platforms (actually also current one can have but nothing like
> that is deployed to my knowledge) have nodes with only cpus. Current SGI
> platforms have nodes with just I/O that we so far cannot manage in the
> core. So the arch code maps them to the nearest memory node.
Lee Schermerhorn <Lee.Schermerhorn@hp.com> wrote:
> For the HP platforms, we can configure each cell with from 0% to 100%
> "cell local memory". When we configure with <100% CLM, the "missing
> percentages" are interleaved by hardware on a cache-line granularity to
> improve bandwidth at the expense of latency for numa-challenged
> applications [and OSes, but not our problem ;-)]. When we boot Linux on
> such a config, all of the real nodes have no memory--it all resides in a
> single interleaved pseudo-node.
>
> When we boot Linux on a 100% CLM configuration [== NUMA], we still have
> the interleaved pseudo-node. It contains a few hundred MB stolen from
> the real nodes to contain the DMA zone. [Interleaved memory resides at
> phys addr 0]. The memoryless-nodes patches, along with the zoneorder
> patches, support this config as well.
>
> Also, when we boot a NUMA config with the "mem=" command line,
> specifying less memory than actually exists, Linux takes the excluded
> memory "off the top" rather than distributing it across the nodes. This
> can result in memoryless nodes, as well.
>
This patch:
Preparation for memoryless node patches.
Provide a generic way to keep nodemasks describing various characteristics of
NUMA nodes.
Remove the node_online_map and the node_possible map and realize the same
functionality using two nodes stats: N_POSSIBLE and N_ONLINE.
[Lee.Schermerhorn@hp.com: Initialize N_*_MEMORY and N_CPU masks for non-NUMA config]
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Tested-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Bob Picco <bob.picco@hp.com>
Cc: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@skynet.ie>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:27 +00:00
|
|
|
[N_CPU] = { { [0] = 1UL } },
|
|
|
|
#endif /* NUMA */
|
|
|
|
};
|
|
|
|
EXPORT_SYMBOL(node_states);
|
|
|
|
|
2018-12-28 08:34:29 +00:00
|
|
|
atomic_long_t _totalram_pages __read_mostly;
|
|
|
|
EXPORT_SYMBOL(_totalram_pages);
|
2006-04-11 05:52:59 +00:00
|
|
|
unsigned long totalreserve_pages __read_mostly;
|
mm: cma: split cma-reserved in dmesg log
When the system boots up, in the dmesg logs we can see the memory
statistics along with total reserved as below. Memory: 458840k/458840k
available, 65448k reserved, 0K highmem
When CMA is enabled, still the total reserved memory remains the same.
However, the CMA memory is not considered as reserved. But, when we see
/proc/meminfo, the CMA memory is part of free memory. This creates
confusion. This patch corrects the problem by properly subtracting the
CMA reserved memory from the total reserved memory in dmesg logs.
Below is the dmesg snapshot from an arm based device with 512MB RAM and
12MB single CMA region.
Before this change:
Memory: 458840k/458840k available, 65448k reserved, 0K highmem
After this change:
Memory: 458840k/458840k available, 53160k reserved, 12288k cma-reserved, 0K highmem
Signed-off-by: Pintu Kumar <pintu.k@samsung.com>
Signed-off-by: Vishnu Pratap Singh <vishnu.ps@samsung.com>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: Rafael Aquini <aquini@redhat.com>
Cc: Jerome Marchand <jmarchan@redhat.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-19 00:17:15 +00:00
|
|
|
unsigned long totalcma_pages __read_mostly;
|
2012-01-10 23:07:42 +00:00
|
|
|
|
2012-05-11 08:00:07 +00:00
|
|
|
int percpu_pagelist_fraction;
|
2009-06-18 03:24:12 +00:00
|
|
|
gfp_t gfp_allowed_mask __read_mostly = GFP_BOOT_MASK;
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
DEFINE_STATIC_KEY_FALSE(init_on_alloc);
|
|
|
|
EXPORT_SYMBOL(init_on_alloc);
|
|
|
|
|
|
|
|
DEFINE_STATIC_KEY_FALSE(init_on_free);
|
|
|
|
EXPORT_SYMBOL(init_on_free);
|
|
|
|
|
mm, page_alloc: do not rely on the order of page_poison and init_on_alloc/free parameters
Patch series "cleanup page poisoning", v3.
I have identified a number of issues and opportunities for cleanup with
CONFIG_PAGE_POISON and friends:
- interaction with init_on_alloc and init_on_free parameters depends on
the order of parameters (Patch 1)
- the boot time enabling uses static key, but inefficienty (Patch 2)
- sanity checking is incompatible with hibernation (Patch 3)
- CONFIG_PAGE_POISONING_NO_SANITY can be removed now that we have
init_on_free (Patch 4)
- CONFIG_PAGE_POISONING_ZERO can be most likely removed now that we
have init_on_free (Patch 5)
This patch (of 5):
Enabling page_poison=1 together with init_on_alloc=1 or init_on_free=1
produces a warning in dmesg that page_poison takes precedence. However,
as these warnings are printed in early_param handlers for
init_on_alloc/free, they are not printed if page_poison is enabled later
on the command line (handlers are called in the order of their
parameters), or when init_on_alloc/free is always enabled by the
respective config option - before the page_poison early param handler is
called, it is not considered to be enabled. This is inconsistent.
We can remove the dependency on order by making the init_on_* parameters
only set a boolean variable, and postponing the evaluation after all early
params have been processed. Introduce a new
init_mem_debugging_and_hardening() function for that, and move the related
debug_pagealloc processing there as well.
As a result init_mem_debugging_and_hardening() knows always accurately if
init_on_* and/or page_poison options were enabled. Thus we can also
optimize want_init_on_alloc() and want_init_on_free(). We don't need to
check page_poisoning_enabled() there, we can instead not enable the
init_on_* static keys at all, if page poisoning is enabled. This results
in a simpler and more effective code.
Link: https://lkml.kernel.org/r/20201113104033.22907-1-vbabka@suse.cz
Link: https://lkml.kernel.org/r/20201113104033.22907-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mateusz Nosek <mateusznosek0@gmail.com>
Cc: Laura Abbott <labbott@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:13:30 +00:00
|
|
|
static bool _init_on_alloc_enabled_early __read_mostly
|
|
|
|
= IS_ENABLED(CONFIG_INIT_ON_ALLOC_DEFAULT_ON);
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
static int __init early_init_on_alloc(char *buf)
|
|
|
|
{
|
|
|
|
|
mm, page_alloc: do not rely on the order of page_poison and init_on_alloc/free parameters
Patch series "cleanup page poisoning", v3.
I have identified a number of issues and opportunities for cleanup with
CONFIG_PAGE_POISON and friends:
- interaction with init_on_alloc and init_on_free parameters depends on
the order of parameters (Patch 1)
- the boot time enabling uses static key, but inefficienty (Patch 2)
- sanity checking is incompatible with hibernation (Patch 3)
- CONFIG_PAGE_POISONING_NO_SANITY can be removed now that we have
init_on_free (Patch 4)
- CONFIG_PAGE_POISONING_ZERO can be most likely removed now that we
have init_on_free (Patch 5)
This patch (of 5):
Enabling page_poison=1 together with init_on_alloc=1 or init_on_free=1
produces a warning in dmesg that page_poison takes precedence. However,
as these warnings are printed in early_param handlers for
init_on_alloc/free, they are not printed if page_poison is enabled later
on the command line (handlers are called in the order of their
parameters), or when init_on_alloc/free is always enabled by the
respective config option - before the page_poison early param handler is
called, it is not considered to be enabled. This is inconsistent.
We can remove the dependency on order by making the init_on_* parameters
only set a boolean variable, and postponing the evaluation after all early
params have been processed. Introduce a new
init_mem_debugging_and_hardening() function for that, and move the related
debug_pagealloc processing there as well.
As a result init_mem_debugging_and_hardening() knows always accurately if
init_on_* and/or page_poison options were enabled. Thus we can also
optimize want_init_on_alloc() and want_init_on_free(). We don't need to
check page_poisoning_enabled() there, we can instead not enable the
init_on_* static keys at all, if page poisoning is enabled. This results
in a simpler and more effective code.
Link: https://lkml.kernel.org/r/20201113104033.22907-1-vbabka@suse.cz
Link: https://lkml.kernel.org/r/20201113104033.22907-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mateusz Nosek <mateusznosek0@gmail.com>
Cc: Laura Abbott <labbott@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:13:30 +00:00
|
|
|
return kstrtobool(buf, &_init_on_alloc_enabled_early);
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
}
|
|
|
|
early_param("init_on_alloc", early_init_on_alloc);
|
|
|
|
|
mm, page_alloc: do not rely on the order of page_poison and init_on_alloc/free parameters
Patch series "cleanup page poisoning", v3.
I have identified a number of issues and opportunities for cleanup with
CONFIG_PAGE_POISON and friends:
- interaction with init_on_alloc and init_on_free parameters depends on
the order of parameters (Patch 1)
- the boot time enabling uses static key, but inefficienty (Patch 2)
- sanity checking is incompatible with hibernation (Patch 3)
- CONFIG_PAGE_POISONING_NO_SANITY can be removed now that we have
init_on_free (Patch 4)
- CONFIG_PAGE_POISONING_ZERO can be most likely removed now that we
have init_on_free (Patch 5)
This patch (of 5):
Enabling page_poison=1 together with init_on_alloc=1 or init_on_free=1
produces a warning in dmesg that page_poison takes precedence. However,
as these warnings are printed in early_param handlers for
init_on_alloc/free, they are not printed if page_poison is enabled later
on the command line (handlers are called in the order of their
parameters), or when init_on_alloc/free is always enabled by the
respective config option - before the page_poison early param handler is
called, it is not considered to be enabled. This is inconsistent.
We can remove the dependency on order by making the init_on_* parameters
only set a boolean variable, and postponing the evaluation after all early
params have been processed. Introduce a new
init_mem_debugging_and_hardening() function for that, and move the related
debug_pagealloc processing there as well.
As a result init_mem_debugging_and_hardening() knows always accurately if
init_on_* and/or page_poison options were enabled. Thus we can also
optimize want_init_on_alloc() and want_init_on_free(). We don't need to
check page_poisoning_enabled() there, we can instead not enable the
init_on_* static keys at all, if page poisoning is enabled. This results
in a simpler and more effective code.
Link: https://lkml.kernel.org/r/20201113104033.22907-1-vbabka@suse.cz
Link: https://lkml.kernel.org/r/20201113104033.22907-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mateusz Nosek <mateusznosek0@gmail.com>
Cc: Laura Abbott <labbott@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:13:30 +00:00
|
|
|
static bool _init_on_free_enabled_early __read_mostly
|
|
|
|
= IS_ENABLED(CONFIG_INIT_ON_FREE_DEFAULT_ON);
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
static int __init early_init_on_free(char *buf)
|
|
|
|
{
|
mm, page_alloc: do not rely on the order of page_poison and init_on_alloc/free parameters
Patch series "cleanup page poisoning", v3.
I have identified a number of issues and opportunities for cleanup with
CONFIG_PAGE_POISON and friends:
- interaction with init_on_alloc and init_on_free parameters depends on
the order of parameters (Patch 1)
- the boot time enabling uses static key, but inefficienty (Patch 2)
- sanity checking is incompatible with hibernation (Patch 3)
- CONFIG_PAGE_POISONING_NO_SANITY can be removed now that we have
init_on_free (Patch 4)
- CONFIG_PAGE_POISONING_ZERO can be most likely removed now that we
have init_on_free (Patch 5)
This patch (of 5):
Enabling page_poison=1 together with init_on_alloc=1 or init_on_free=1
produces a warning in dmesg that page_poison takes precedence. However,
as these warnings are printed in early_param handlers for
init_on_alloc/free, they are not printed if page_poison is enabled later
on the command line (handlers are called in the order of their
parameters), or when init_on_alloc/free is always enabled by the
respective config option - before the page_poison early param handler is
called, it is not considered to be enabled. This is inconsistent.
We can remove the dependency on order by making the init_on_* parameters
only set a boolean variable, and postponing the evaluation after all early
params have been processed. Introduce a new
init_mem_debugging_and_hardening() function for that, and move the related
debug_pagealloc processing there as well.
As a result init_mem_debugging_and_hardening() knows always accurately if
init_on_* and/or page_poison options were enabled. Thus we can also
optimize want_init_on_alloc() and want_init_on_free(). We don't need to
check page_poisoning_enabled() there, we can instead not enable the
init_on_* static keys at all, if page poisoning is enabled. This results
in a simpler and more effective code.
Link: https://lkml.kernel.org/r/20201113104033.22907-1-vbabka@suse.cz
Link: https://lkml.kernel.org/r/20201113104033.22907-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mateusz Nosek <mateusznosek0@gmail.com>
Cc: Laura Abbott <labbott@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:13:30 +00:00
|
|
|
return kstrtobool(buf, &_init_on_free_enabled_early);
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
}
|
|
|
|
early_param("init_on_free", early_init_on_free);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2015-09-08 22:01:25 +00:00
|
|
|
/*
|
|
|
|
* A cached value of the page's pageblock's migratetype, used when the page is
|
|
|
|
* put on a pcplist. Used to avoid the pageblock migratetype lookup when
|
|
|
|
* freeing from pcplists in most cases, at the cost of possibly becoming stale.
|
|
|
|
* Also the migratetype set in the page does not necessarily match the pcplist
|
|
|
|
* index, e.g. page might have MIGRATE_CMA set but be on a pcplist with any
|
|
|
|
* other index - this ensures that it will be put on the correct CMA freelist.
|
|
|
|
*/
|
|
|
|
static inline int get_pcppage_migratetype(struct page *page)
|
|
|
|
{
|
|
|
|
return page->index;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void set_pcppage_migratetype(struct page *page, int migratetype)
|
|
|
|
{
|
|
|
|
page->index = migratetype;
|
|
|
|
}
|
|
|
|
|
2010-03-05 21:42:13 +00:00
|
|
|
#ifdef CONFIG_PM_SLEEP
|
|
|
|
/*
|
|
|
|
* The following functions are used by the suspend/hibernate code to temporarily
|
|
|
|
* change gfp_allowed_mask in order to avoid using I/O during memory allocations
|
|
|
|
* while devices are suspended. To avoid races with the suspend/hibernate code,
|
2018-07-31 08:51:32 +00:00
|
|
|
* they should always be called with system_transition_mutex held
|
|
|
|
* (gfp_allowed_mask also should only be modified with system_transition_mutex
|
|
|
|
* held, unless the suspend/hibernate code is guaranteed not to run in parallel
|
|
|
|
* with that modification).
|
2010-03-05 21:42:13 +00:00
|
|
|
*/
|
2010-12-03 21:57:45 +00:00
|
|
|
|
|
|
|
static gfp_t saved_gfp_mask;
|
|
|
|
|
|
|
|
void pm_restore_gfp_mask(void)
|
2010-03-05 21:42:13 +00:00
|
|
|
{
|
2018-07-31 08:51:32 +00:00
|
|
|
WARN_ON(!mutex_is_locked(&system_transition_mutex));
|
2010-12-03 21:57:45 +00:00
|
|
|
if (saved_gfp_mask) {
|
|
|
|
gfp_allowed_mask = saved_gfp_mask;
|
|
|
|
saved_gfp_mask = 0;
|
|
|
|
}
|
2010-03-05 21:42:13 +00:00
|
|
|
}
|
|
|
|
|
2010-12-03 21:57:45 +00:00
|
|
|
void pm_restrict_gfp_mask(void)
|
2010-03-05 21:42:13 +00:00
|
|
|
{
|
2018-07-31 08:51:32 +00:00
|
|
|
WARN_ON(!mutex_is_locked(&system_transition_mutex));
|
2010-12-03 21:57:45 +00:00
|
|
|
WARN_ON(saved_gfp_mask);
|
|
|
|
saved_gfp_mask = gfp_allowed_mask;
|
2015-11-07 00:28:21 +00:00
|
|
|
gfp_allowed_mask &= ~(__GFP_IO | __GFP_FS);
|
2010-03-05 21:42:13 +00:00
|
|
|
}
|
2012-01-10 23:07:15 +00:00
|
|
|
|
|
|
|
bool pm_suspended_storage(void)
|
|
|
|
{
|
2015-11-07 00:28:21 +00:00
|
|
|
if ((gfp_allowed_mask & (__GFP_IO | __GFP_FS)) == (__GFP_IO | __GFP_FS))
|
2012-01-10 23:07:15 +00:00
|
|
|
return false;
|
|
|
|
return true;
|
|
|
|
}
|
2010-03-05 21:42:13 +00:00
|
|
|
#endif /* CONFIG_PM_SLEEP */
|
|
|
|
|
2007-10-16 08:26:01 +00:00
|
|
|
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
|
2015-11-07 00:29:57 +00:00
|
|
|
unsigned int pageblock_order __read_mostly;
|
2007-10-16 08:26:01 +00:00
|
|
|
#endif
|
|
|
|
|
2020-10-16 03:09:35 +00:00
|
|
|
static void __free_pages_ok(struct page *page, unsigned int order,
|
|
|
|
fpi_t fpi_flags);
|
2006-01-06 08:11:08 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* results with 256, 32 in the lowmem_reserve sysctl:
|
|
|
|
* 1G machine -> (16M dma, 800M-16M normal, 1G-800M high)
|
|
|
|
* 1G machine -> (16M dma, 784M normal, 224M high)
|
|
|
|
* NORMAL allocation will leave 784M/256 of ram reserved in the ZONE_DMA
|
|
|
|
* HIGHMEM allocation will leave 224M/32 of ram reserved in ZONE_NORMAL
|
2015-02-12 23:00:22 +00:00
|
|
|
* HIGHMEM allocation will leave (224M+784M)/256 of ram reserved in ZONE_DMA
|
2005-11-05 16:25:53 +00:00
|
|
|
*
|
|
|
|
* TBD: should special case ZONE_DMA32 machines here - in those we normally
|
|
|
|
* don't need any ZONE_NORMAL reservation
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2018-04-10 23:30:11 +00:00
|
|
|
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES] = {
|
2007-02-10 09:43:10 +00:00
|
|
|
#ifdef CONFIG_ZONE_DMA
|
2018-04-10 23:30:11 +00:00
|
|
|
[ZONE_DMA] = 256,
|
2007-02-10 09:43:10 +00:00
|
|
|
#endif
|
2006-09-26 06:31:13 +00:00
|
|
|
#ifdef CONFIG_ZONE_DMA32
|
2018-04-10 23:30:11 +00:00
|
|
|
[ZONE_DMA32] = 256,
|
2006-09-26 06:31:13 +00:00
|
|
|
#endif
|
2018-04-10 23:30:11 +00:00
|
|
|
[ZONE_NORMAL] = 32,
|
2006-09-26 06:31:14 +00:00
|
|
|
#ifdef CONFIG_HIGHMEM
|
2018-04-10 23:30:11 +00:00
|
|
|
[ZONE_HIGHMEM] = 0,
|
2006-09-26 06:31:14 +00:00
|
|
|
#endif
|
2018-04-10 23:30:11 +00:00
|
|
|
[ZONE_MOVABLE] = 0,
|
2006-09-26 06:31:13 +00:00
|
|
|
};
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-12-07 04:40:36 +00:00
|
|
|
static char * const zone_names[MAX_NR_ZONES] = {
|
2007-02-10 09:43:10 +00:00
|
|
|
#ifdef CONFIG_ZONE_DMA
|
2006-09-26 06:31:13 +00:00
|
|
|
"DMA",
|
2007-02-10 09:43:10 +00:00
|
|
|
#endif
|
2006-09-26 06:31:13 +00:00
|
|
|
#ifdef CONFIG_ZONE_DMA32
|
2006-09-26 06:31:13 +00:00
|
|
|
"DMA32",
|
2006-09-26 06:31:13 +00:00
|
|
|
#endif
|
2006-09-26 06:31:13 +00:00
|
|
|
"Normal",
|
2006-09-26 06:31:14 +00:00
|
|
|
#ifdef CONFIG_HIGHMEM
|
2007-07-17 11:03:12 +00:00
|
|
|
"HighMem",
|
2006-09-26 06:31:14 +00:00
|
|
|
#endif
|
2007-07-17 11:03:12 +00:00
|
|
|
"Movable",
|
2015-08-09 19:29:06 +00:00
|
|
|
#ifdef CONFIG_ZONE_DEVICE
|
|
|
|
"Device",
|
|
|
|
#endif
|
2006-09-26 06:31:13 +00:00
|
|
|
};
|
|
|
|
|
2018-12-28 08:35:55 +00:00
|
|
|
const char * const migratetype_names[MIGRATE_TYPES] = {
|
mm, page_owner: print migratetype of page and pageblock, symbolic flags
The information in /sys/kernel/debug/page_owner includes the migratetype
of the pageblock the page belongs to. This is also checked against the
page's migratetype (as declared by gfp_flags during its allocation), and
the page is reported as Fallback if its migratetype differs from the
pageblock's one. t This is somewhat misleading because in fact fallback
allocation is not the only reason why these two can differ. It also
doesn't direcly provide the page's migratetype, although it's possible
to derive that from the gfp_flags.
It's arguably better to print both page and pageblock's migratetype and
leave the interpretation to the consumer than to suggest fallback
allocation as the only possible reason. While at it, we can print the
migratetypes as string the same way as /proc/pagetypeinfo does, as some
of the numeric values depend on kernel configuration. For that, this
patch moves the migratetype_names array from #ifdef CONFIG_PROC_FS part
of mm/vmstat.c to mm/page_alloc.c and exports it.
With the new format strings for flags, we can now also provide symbolic
page and gfp flags in the /sys/kernel/debug/page_owner file. This
replaces the positional printing of page flags as single letters, which
might have looked nicer, but was limited to a subset of flags, and
required the user to remember the letters.
Example page_owner entry after the patch:
Page allocated via order 0, mask 0x24213ca(GFP_HIGHUSER_MOVABLE|__GFP_COLD|__GFP_NOWARN|__GFP_NORETRY)
PFN 520 type Movable Block 1 type Movable Flags 0xfffff8001006c(referenced|uptodate|lru|active|mappedtodisk)
[<ffffffff811682c4>] __alloc_pages_nodemask+0x134/0x230
[<ffffffff811b4058>] alloc_pages_current+0x88/0x120
[<ffffffff8115e386>] __page_cache_alloc+0xe6/0x120
[<ffffffff8116ba6c>] __do_page_cache_readahead+0xdc/0x240
[<ffffffff8116bd05>] ondemand_readahead+0x135/0x260
[<ffffffff8116bfb1>] page_cache_sync_readahead+0x31/0x50
[<ffffffff81160523>] generic_file_read_iter+0x453/0x760
[<ffffffff811e0d57>] __vfs_read+0xa7/0xd0
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Sasha Levin <sasha.levin@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-15 21:56:08 +00:00
|
|
|
"Unmovable",
|
|
|
|
"Movable",
|
|
|
|
"Reclaimable",
|
|
|
|
"HighAtomic",
|
|
|
|
#ifdef CONFIG_CMA
|
|
|
|
"CMA",
|
|
|
|
#endif
|
|
|
|
#ifdef CONFIG_MEMORY_ISOLATION
|
|
|
|
"Isolate",
|
|
|
|
#endif
|
|
|
|
};
|
|
|
|
|
2020-06-03 22:59:17 +00:00
|
|
|
compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS] = {
|
|
|
|
[NULL_COMPOUND_DTOR] = NULL,
|
|
|
|
[COMPOUND_PAGE_DTOR] = free_compound_page,
|
2015-11-07 00:29:50 +00:00
|
|
|
#ifdef CONFIG_HUGETLB_PAGE
|
2020-06-03 22:59:17 +00:00
|
|
|
[HUGETLB_PAGE_DTOR] = free_huge_page,
|
2015-11-07 00:29:50 +00:00
|
|
|
#endif
|
2016-01-16 00:54:17 +00:00
|
|
|
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
|
2020-06-03 22:59:17 +00:00
|
|
|
[TRANSHUGE_PAGE_DTOR] = free_transhuge_page,
|
2016-01-16 00:54:17 +00:00
|
|
|
#endif
|
2015-11-07 00:29:50 +00:00
|
|
|
};
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
int min_free_kbytes = 1024;
|
2014-01-23 23:53:28 +00:00
|
|
|
int user_min_free_kbytes = -1;
|
2019-04-26 05:23:51 +00:00
|
|
|
#ifdef CONFIG_DISCONTIGMEM
|
|
|
|
/*
|
|
|
|
* DiscontigMem defines memory ranges as separate pg_data_t even if the ranges
|
|
|
|
* are not on separate NUMA nodes. Functionally this works but with
|
|
|
|
* watermark_boost_factor, it can reclaim prematurely as the ranges can be
|
|
|
|
* quite small. By default, do not boost watermarks on discontigmem as in
|
|
|
|
* many cases very high-order allocations like THP are likely to be
|
|
|
|
* unsupported and the premature reclaim offsets the advantage of long-term
|
|
|
|
* fragmentation avoidance.
|
|
|
|
*/
|
|
|
|
int watermark_boost_factor __read_mostly;
|
|
|
|
#else
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
int watermark_boost_factor __read_mostly = 15000;
|
2019-04-26 05:23:51 +00:00
|
|
|
#endif
|
2016-03-17 21:19:14 +00:00
|
|
|
int watermark_scale_factor = 10;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2018-12-28 08:37:24 +00:00
|
|
|
static unsigned long nr_kernel_pages __initdata;
|
|
|
|
static unsigned long nr_all_pages __initdata;
|
|
|
|
static unsigned long dma_reserve __initdata;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2018-12-28 08:37:24 +00:00
|
|
|
static unsigned long arch_zone_lowest_possible_pfn[MAX_NR_ZONES] __initdata;
|
|
|
|
static unsigned long arch_zone_highest_possible_pfn[MAX_NR_ZONES] __initdata;
|
2018-04-05 23:23:12 +00:00
|
|
|
static unsigned long required_kernelcore __initdata;
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
static unsigned long required_kernelcore_percent __initdata;
|
2018-04-05 23:23:12 +00:00
|
|
|
static unsigned long required_movablecore __initdata;
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
static unsigned long required_movablecore_percent __initdata;
|
2018-12-28 08:37:24 +00:00
|
|
|
static unsigned long zone_movable_pfn[MAX_NUMNODES] __initdata;
|
2018-04-05 23:23:12 +00:00
|
|
|
static bool mirrored_kernelcore __meminitdata;
|
2011-12-08 18:22:09 +00:00
|
|
|
|
|
|
|
/* movable_zone is the "real" zone pages in ZONE_MOVABLE are taken from */
|
|
|
|
int movable_zone;
|
|
|
|
EXPORT_SYMBOL(movable_zone);
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
|
2007-05-23 20:57:55 +00:00
|
|
|
#if MAX_NUMNODES > 1
|
2019-03-05 23:48:26 +00:00
|
|
|
unsigned int nr_node_ids __read_mostly = MAX_NUMNODES;
|
2019-03-05 23:48:29 +00:00
|
|
|
unsigned int nr_online_nodes __read_mostly = 1;
|
2007-05-23 20:57:55 +00:00
|
|
|
EXPORT_SYMBOL(nr_node_ids);
|
2009-06-16 22:32:15 +00:00
|
|
|
EXPORT_SYMBOL(nr_online_nodes);
|
2007-05-23 20:57:55 +00:00
|
|
|
#endif
|
|
|
|
|
2007-10-16 08:25:54 +00:00
|
|
|
int page_group_by_mobility_disabled __read_mostly;
|
|
|
|
|
2015-06-30 21:57:02 +00:00
|
|
|
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
|
mm/page_alloc.c: don't call kasan_free_pages() at deferred mem init
When CONFIG_KASAN is enabled on large memory SMP systems, the deferrred
pages initialization can take a long time. Below were the reported init
times on a 8-socket 96-core 4TB IvyBridge system.
1) Non-debug kernel without CONFIG_KASAN
[ 8.764222] node 1 initialised, 132086516 pages in 7027ms
2) Debug kernel with CONFIG_KASAN
[ 146.288115] node 1 initialised, 132075466 pages in 143052ms
So the page init time in a debug kernel was 20X of the non-debug kernel.
The long init time can be problematic as the page initialization is done
with interrupt disabled. In this particular case, it caused the
appearance of following warning messages as well as NMI backtraces of all
the cores that were doing the initialization.
[ 68.240049] rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
[ 68.241000] rcu: 25-...0: (100 ticks this GP) idle=b72/1/0x4000000000000000 softirq=915/915 fqs=16252
[ 68.241000] rcu: 44-...0: (95 ticks this GP) idle=49a/1/0x4000000000000000 softirq=788/788 fqs=16253
[ 68.241000] rcu: 54-...0: (104 ticks this GP) idle=03a/1/0x4000000000000000 softirq=721/825 fqs=16253
[ 68.241000] rcu: 60-...0: (103 ticks this GP) idle=cbe/1/0x4000000000000000 softirq=637/740 fqs=16253
[ 68.241000] rcu: 72-...0: (105 ticks this GP) idle=786/1/0x4000000000000000 softirq=536/641 fqs=16253
[ 68.241000] rcu: 84-...0: (99 ticks this GP) idle=292/1/0x4000000000000000 softirq=537/537 fqs=16253
[ 68.241000] rcu: 111-...0: (104 ticks this GP) idle=bde/1/0x4000000000000000 softirq=474/476 fqs=16253
[ 68.241000] rcu: (detected by 13, t=65018 jiffies, g=249, q=2)
The long init time was mainly caused by the call to kasan_free_pages() to
poison the newly initialized pages. On a 4TB system, we are talking about
almost 500GB of memory probably on the same node.
In reality, we may not need to poison the newly initialized pages before
they are ever allocated. So KASAN poisoning of freed pages before the
completion of deferred memory initialization is now disabled. Those pages
will be properly poisoned when they are allocated or freed after deferred
pages initialization is done.
With this change, the new page initialization time became:
[ 21.948010] node 1 initialised, 132075466 pages in 18702ms
This was still about double the non-debug kernel time, but was much
better than before.
Link: http://lkml.kernel.org/r/1544459388-8736-1-git-send-email-longman@redhat.com
Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:38:51 +00:00
|
|
|
/*
|
|
|
|
* During boot we initialize deferred pages on-demand, as needed, but once
|
|
|
|
* page_alloc_init_late() has finished, the deferred pages are all initialized,
|
|
|
|
* and we can permanently disable that path.
|
|
|
|
*/
|
|
|
|
static DEFINE_STATIC_KEY_TRUE(deferred_pages);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Calling kasan_free_pages() only after deferred memory initialization
|
|
|
|
* has completed. Poisoning pages during deferred memory init will greatly
|
|
|
|
* lengthen the process and cause problem in large memory systems as the
|
|
|
|
* deferred pages initialization is done with interrupt disabled.
|
|
|
|
*
|
|
|
|
* Assuming that there will be no reference to those newly initialized
|
|
|
|
* pages before they are ever allocated, this should have no effect on
|
|
|
|
* KASAN memory tracking as the poison will be properly inserted at page
|
|
|
|
* allocation time. The only corner case is when pages are allocated by
|
|
|
|
* on-demand allocation and then freed again before the deferred pages
|
|
|
|
* initialization is done, but this is not likely to happen.
|
|
|
|
*/
|
|
|
|
static inline void kasan_free_nondeferred_pages(struct page *page, int order)
|
|
|
|
{
|
|
|
|
if (!static_branch_unlikely(&deferred_pages))
|
|
|
|
kasan_free_pages(page, order);
|
|
|
|
}
|
|
|
|
|
2015-06-30 21:57:02 +00:00
|
|
|
/* Returns true if the struct page for the pfn is uninitialised */
|
2015-06-30 21:57:27 +00:00
|
|
|
static inline bool __meminit early_page_uninitialised(unsigned long pfn)
|
2015-06-30 21:57:02 +00:00
|
|
|
{
|
2016-07-14 19:07:23 +00:00
|
|
|
int nid = early_pfn_to_nid(pfn);
|
|
|
|
|
|
|
|
if (node_online(nid) && pfn >= NODE_DATA(nid)->first_deferred_pfn)
|
2015-06-30 21:57:02 +00:00
|
|
|
return true;
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2018-10-26 22:09:37 +00:00
|
|
|
* Returns true when the remaining initialisation should be deferred until
|
2015-06-30 21:57:02 +00:00
|
|
|
* later in the boot cycle when it can be parallelised.
|
|
|
|
*/
|
2018-10-26 22:09:37 +00:00
|
|
|
static bool __meminit
|
|
|
|
defer_init(int nid, unsigned long pfn, unsigned long end_pfn)
|
2015-06-30 21:57:02 +00:00
|
|
|
{
|
2018-10-26 22:09:37 +00:00
|
|
|
static unsigned long prev_end_pfn, nr_initialised;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* prev_end_pfn static that contains the end of previous zone
|
|
|
|
* No need to protect because called very early in boot before smp_init.
|
|
|
|
*/
|
|
|
|
if (prev_end_pfn != end_pfn) {
|
|
|
|
prev_end_pfn = end_pfn;
|
|
|
|
nr_initialised = 0;
|
|
|
|
}
|
|
|
|
|
2018-02-01 00:20:07 +00:00
|
|
|
/* Always populate low zones for address-constrained allocations */
|
2018-10-26 22:09:37 +00:00
|
|
|
if (end_pfn < pgdat_end_pfn(NODE_DATA(nid)))
|
2015-06-30 21:57:02 +00:00
|
|
|
return false;
|
mm: check nr_initialised with PAGES_PER_SECTION directly in defer_init()
When DEFERRED_STRUCT_PAGE_INIT is configured, only the first section of
each node's highest zone is initialized before defer stage.
static_init_pgcnt is used to store the number of pages like this:
pgdat->static_init_pgcnt = min_t(unsigned long, PAGES_PER_SECTION,
pgdat->node_spanned_pages);
because we don't want to overflow zone's range.
But this is not necessary, since defer_init() is called like this:
memmap_init_zone()
for pfn in [start_pfn, end_pfn)
defer_init(pfn, end_pfn)
In case (pgdat->node_spanned_pages < PAGES_PER_SECTION), the loop would
stop before calling defer_init().
BTW, comparing PAGES_PER_SECTION with node_spanned_pages is not correct,
since nr_initialised is zone based instead of node based. Even
node_spanned_pages is bigger than PAGES_PER_SECTION, its highest zone
would have pages less than PAGES_PER_SECTION.
Link: http://lkml.kernel.org/r/20181122094807.6985-1-richard.weiyang@gmail.com
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:36:18 +00:00
|
|
|
|
mm: memmap defer init doesn't work as expected
VMware observed a performance regression during memmap init on their
platform, and bisected to commit 73a6e474cb376 ("mm: memmap_init:
iterate over memblock regions rather that check each PFN") causing it.
Before the commit:
[0.033176] Normal zone: 1445888 pages used for memmap
[0.033176] Normal zone: 89391104 pages, LIFO batch:63
[0.035851] ACPI: PM-Timer IO Port: 0x448
With commit
[0.026874] Normal zone: 1445888 pages used for memmap
[0.026875] Normal zone: 89391104 pages, LIFO batch:63
[2.028450] ACPI: PM-Timer IO Port: 0x448
The root cause is the current memmap defer init doesn't work as expected.
Before, memmap_init_zone() was used to do memmap init of one whole zone,
to initialize all low zones of one numa node, but defer memmap init of
the last zone in that numa node. However, since commit 73a6e474cb376,
function memmap_init() is adapted to iterater over memblock regions
inside one zone, then call memmap_init_zone() to do memmap init for each
region.
E.g, on VMware's system, the memory layout is as below, there are two
memory regions in node 2. The current code will mistakenly initialize the
whole 1st region [mem 0xab00000000-0xfcffffffff], then do memmap defer to
iniatialize only one memmory section on the 2nd region [mem
0x10000000000-0x1033fffffff]. In fact, we only expect to see that there's
only one memory section's memmap initialized. That's why more time is
costed at the time.
[ 0.008842] ACPI: SRAT: Node 0 PXM 0 [mem 0x00000000-0x0009ffff]
[ 0.008842] ACPI: SRAT: Node 0 PXM 0 [mem 0x00100000-0xbfffffff]
[ 0.008843] ACPI: SRAT: Node 0 PXM 0 [mem 0x100000000-0x55ffffffff]
[ 0.008844] ACPI: SRAT: Node 1 PXM 1 [mem 0x5600000000-0xaaffffffff]
[ 0.008844] ACPI: SRAT: Node 2 PXM 2 [mem 0xab00000000-0xfcffffffff]
[ 0.008845] ACPI: SRAT: Node 2 PXM 2 [mem 0x10000000000-0x1033fffffff]
Now, let's add a parameter 'zone_end_pfn' to memmap_init_zone() to pass
down the real zone end pfn so that defer_init() can use it to judge
whether defer need be taken in zone wide.
Link: https://lkml.kernel.org/r/20201223080811.16211-1-bhe@redhat.com
Link: https://lkml.kernel.org/r/20201223080811.16211-2-bhe@redhat.com
Fixes: commit 73a6e474cb376 ("mm: memmap_init: iterate over memblock regions rather that check each PFN")
Signed-off-by: Baoquan He <bhe@redhat.com>
Reported-by: Rahul Gopakumar <gopakumarr@vmware.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-29 23:14:37 +00:00
|
|
|
if (NODE_DATA(nid)->first_deferred_pfn != ULONG_MAX)
|
|
|
|
return true;
|
mm: check nr_initialised with PAGES_PER_SECTION directly in defer_init()
When DEFERRED_STRUCT_PAGE_INIT is configured, only the first section of
each node's highest zone is initialized before defer stage.
static_init_pgcnt is used to store the number of pages like this:
pgdat->static_init_pgcnt = min_t(unsigned long, PAGES_PER_SECTION,
pgdat->node_spanned_pages);
because we don't want to overflow zone's range.
But this is not necessary, since defer_init() is called like this:
memmap_init_zone()
for pfn in [start_pfn, end_pfn)
defer_init(pfn, end_pfn)
In case (pgdat->node_spanned_pages < PAGES_PER_SECTION), the loop would
stop before calling defer_init().
BTW, comparing PAGES_PER_SECTION with node_spanned_pages is not correct,
since nr_initialised is zone based instead of node based. Even
node_spanned_pages is bigger than PAGES_PER_SECTION, its highest zone
would have pages less than PAGES_PER_SECTION.
Link: http://lkml.kernel.org/r/20181122094807.6985-1-richard.weiyang@gmail.com
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:36:18 +00:00
|
|
|
/*
|
|
|
|
* We start only with one section of pages, more pages are added as
|
|
|
|
* needed until the rest of deferred pages are initialized.
|
|
|
|
*/
|
2018-10-26 22:09:37 +00:00
|
|
|
nr_initialised++;
|
mm: check nr_initialised with PAGES_PER_SECTION directly in defer_init()
When DEFERRED_STRUCT_PAGE_INIT is configured, only the first section of
each node's highest zone is initialized before defer stage.
static_init_pgcnt is used to store the number of pages like this:
pgdat->static_init_pgcnt = min_t(unsigned long, PAGES_PER_SECTION,
pgdat->node_spanned_pages);
because we don't want to overflow zone's range.
But this is not necessary, since defer_init() is called like this:
memmap_init_zone()
for pfn in [start_pfn, end_pfn)
defer_init(pfn, end_pfn)
In case (pgdat->node_spanned_pages < PAGES_PER_SECTION), the loop would
stop before calling defer_init().
BTW, comparing PAGES_PER_SECTION with node_spanned_pages is not correct,
since nr_initialised is zone based instead of node based. Even
node_spanned_pages is bigger than PAGES_PER_SECTION, its highest zone
would have pages less than PAGES_PER_SECTION.
Link: http://lkml.kernel.org/r/20181122094807.6985-1-richard.weiyang@gmail.com
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:36:18 +00:00
|
|
|
if ((nr_initialised > PAGES_PER_SECTION) &&
|
2018-10-26 22:09:37 +00:00
|
|
|
(pfn & (PAGES_PER_SECTION - 1)) == 0) {
|
|
|
|
NODE_DATA(nid)->first_deferred_pfn = pfn;
|
|
|
|
return true;
|
2015-06-30 21:57:02 +00:00
|
|
|
}
|
2018-10-26 22:09:37 +00:00
|
|
|
return false;
|
2015-06-30 21:57:02 +00:00
|
|
|
}
|
|
|
|
#else
|
mm/page_alloc.c: don't call kasan_free_pages() at deferred mem init
When CONFIG_KASAN is enabled on large memory SMP systems, the deferrred
pages initialization can take a long time. Below were the reported init
times on a 8-socket 96-core 4TB IvyBridge system.
1) Non-debug kernel without CONFIG_KASAN
[ 8.764222] node 1 initialised, 132086516 pages in 7027ms
2) Debug kernel with CONFIG_KASAN
[ 146.288115] node 1 initialised, 132075466 pages in 143052ms
So the page init time in a debug kernel was 20X of the non-debug kernel.
The long init time can be problematic as the page initialization is done
with interrupt disabled. In this particular case, it caused the
appearance of following warning messages as well as NMI backtraces of all
the cores that were doing the initialization.
[ 68.240049] rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
[ 68.241000] rcu: 25-...0: (100 ticks this GP) idle=b72/1/0x4000000000000000 softirq=915/915 fqs=16252
[ 68.241000] rcu: 44-...0: (95 ticks this GP) idle=49a/1/0x4000000000000000 softirq=788/788 fqs=16253
[ 68.241000] rcu: 54-...0: (104 ticks this GP) idle=03a/1/0x4000000000000000 softirq=721/825 fqs=16253
[ 68.241000] rcu: 60-...0: (103 ticks this GP) idle=cbe/1/0x4000000000000000 softirq=637/740 fqs=16253
[ 68.241000] rcu: 72-...0: (105 ticks this GP) idle=786/1/0x4000000000000000 softirq=536/641 fqs=16253
[ 68.241000] rcu: 84-...0: (99 ticks this GP) idle=292/1/0x4000000000000000 softirq=537/537 fqs=16253
[ 68.241000] rcu: 111-...0: (104 ticks this GP) idle=bde/1/0x4000000000000000 softirq=474/476 fqs=16253
[ 68.241000] rcu: (detected by 13, t=65018 jiffies, g=249, q=2)
The long init time was mainly caused by the call to kasan_free_pages() to
poison the newly initialized pages. On a 4TB system, we are talking about
almost 500GB of memory probably on the same node.
In reality, we may not need to poison the newly initialized pages before
they are ever allocated. So KASAN poisoning of freed pages before the
completion of deferred memory initialization is now disabled. Those pages
will be properly poisoned when they are allocated or freed after deferred
pages initialization is done.
With this change, the new page initialization time became:
[ 21.948010] node 1 initialised, 132075466 pages in 18702ms
This was still about double the non-debug kernel time, but was much
better than before.
Link: http://lkml.kernel.org/r/1544459388-8736-1-git-send-email-longman@redhat.com
Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:38:51 +00:00
|
|
|
#define kasan_free_nondeferred_pages(p, o) kasan_free_pages(p, o)
|
|
|
|
|
2015-06-30 21:57:02 +00:00
|
|
|
static inline bool early_page_uninitialised(unsigned long pfn)
|
|
|
|
{
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2018-10-26 22:09:37 +00:00
|
|
|
static inline bool defer_init(int nid, unsigned long pfn, unsigned long end_pfn)
|
2015-06-30 21:57:02 +00:00
|
|
|
{
|
2018-10-26 22:09:37 +00:00
|
|
|
return false;
|
2015-06-30 21:57:02 +00:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2016-05-20 00:14:27 +00:00
|
|
|
/* Return a pointer to the bitmap storing bits affecting a block of pages */
|
|
|
|
static inline unsigned long *get_pageblock_bitmap(struct page *page,
|
|
|
|
unsigned long pfn)
|
|
|
|
{
|
|
|
|
#ifdef CONFIG_SPARSEMEM
|
mm/sparsemem: introduce struct mem_section_usage
Patch series "mm: Sub-section memory hotplug support", v10.
The memory hotplug section is an arbitrary / convenient unit for memory
hotplug. 'Section-size' units have bled into the user interface
('memblock' sysfs) and can not be changed without breaking existing
userspace. The section-size constraint, while mostly benign for typical
memory hotplug, has and continues to wreak havoc with 'device-memory'
use cases, persistent memory (pmem) in particular. Recall that pmem
uses devm_memremap_pages(), and subsequently arch_add_memory(), to
allocate a 'struct page' memmap for pmem. However, it does not use the
'bottom half' of memory hotplug, i.e. never marks pmem pages online and
never exposes the userspace memblock interface for pmem. This leaves an
opening to redress the section-size constraint.
To date, the libnvdimm subsystem has attempted to inject padding to
satisfy the internal constraints of arch_add_memory(). Beyond
complicating the code, leading to bugs [2], wasting memory, and limiting
configuration flexibility, the padding hack is broken when the platform
changes this physical memory alignment of pmem from one boot to the
next. Device failure (intermittent or permanent) and physical
reconfiguration are events that can cause the platform firmware to
change the physical placement of pmem on a subsequent boot, and device
failure is an everyday event in a data-center.
It turns out that sections are only a hard requirement of the
user-facing interface for memory hotplug and with a bit more
infrastructure sub-section arch_add_memory() support can be added for
kernel internal usages like devm_memremap_pages(). Here is an analysis
of the current design assumptions in the current code and how they are
addressed in the new implementation:
Current design assumptions:
- Sections that describe boot memory (early sections) are never
unplugged / removed.
- pfn_valid(), in the CONFIG_SPARSEMEM_VMEMMAP=y, case devolves to a
valid_section() check
- __add_pages() and helper routines assume all operations occur in
PAGES_PER_SECTION units.
- The memblock sysfs interface only comprehends full sections
New design assumptions:
- Sections are instrumented with a sub-section bitmask to track (on
x86) individual 2MB sub-divisions of a 128MB section.
- Partially populated early sections can be extended with additional
sub-sections, and those sub-sections can be removed with
arch_remove_memory(). With this in place we no longer lose usable
memory capacity to padding.
- pfn_valid() is updated to look deeper than valid_section() to also
check the active-sub-section mask. This indication is in the same
cacheline as the valid_section() so the performance impact is
expected to be negligible. So far the lkp robot has not reported any
regressions.
- Outside of the core vmemmap population routines which are replaced,
other helper routines like shrink_{zone,pgdat}_span() are updated to
handle the smaller granularity. Core memory hotplug routines that
deal with online memory are not touched.
- The existing memblock sysfs user api guarantees / assumptions are not
touched since this capability is limited to !online
!memblock-sysfs-accessible sections.
Meanwhile the issue reports continue to roll in from users that do not
understand when and how the 128MB constraint will bite them. The current
implementation relied on being able to support at least one misaligned
namespace, but that immediately falls over on any moderately complex
namespace creation attempt. Beyond the initial problem of 'System RAM'
colliding with pmem, and the unsolvable problem of physical alignment
changes, Linux is now being exposed to platforms that collide pmem ranges
with other pmem ranges by default [3]. In short, devm_memremap_pages()
has pushed the venerable section-size constraint past the breaking point,
and the simplicity of section-aligned arch_add_memory() is no longer
tenable.
These patches are exposed to the kbuild robot on a subsection-v10 branch
[4], and a preview of the unit test for this functionality is available
on the 'subsection-pending' branch of ndctl [5].
[2]: https://lore.kernel.org/r/155000671719.348031.2347363160141119237.stgit@dwillia2-desk3.amr.corp.intel.com
[3]: https://github.com/pmem/ndctl/issues/76
[4]: https://git.kernel.org/pub/scm/linux/kernel/git/djbw/nvdimm.git/log/?h=subsection-v10
[5]: https://github.com/pmem/ndctl/commit/7c59b4867e1c
This patch (of 13):
Towards enabling memory hotplug to track partial population of a section,
introduce 'struct mem_section_usage'.
A pointer to a 'struct mem_section_usage' instance replaces the existing
pointer to a 'pageblock_flags' bitmap. Effectively it adds one more
'unsigned long' beyond the 'pageblock_flags' (usemap) allocation to house
a new 'subsection_map' bitmap. The new bitmap enables the memory
hot{plug,remove} implementation to act on incremental sub-divisions of a
section.
SUBSECTION_SHIFT is defined as global constant instead of per-architecture
value like SECTION_SIZE_BITS in order to allow cross-arch compatibility of
subsection users. Specifically a common subsection size allows for the
possibility that persistent memory namespace configurations be made
compatible across architectures.
The primary motivation for this functionality is to support platforms that
mix "System RAM" and "Persistent Memory" within a single section, or
multiple PMEM ranges with different mapping lifetimes within a single
section. The section restriction for hotplug has caused an ongoing saga
of hacks and bugs for devm_memremap_pages() users.
Beyond the fixups to teach existing paths how to retrieve the 'usemap'
from a section, and updates to usemap allocation path, there are no
expected behavior changes.
Link: http://lkml.kernel.org/r/156092349845.979959.73333291612799019.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richardw.yang@linux.intel.com>
Tested-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> [ppc64]
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jérôme Glisse <jglisse@redhat.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Qian Cai <cai@lca.pw>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-18 22:57:57 +00:00
|
|
|
return section_to_usemap(__pfn_to_section(pfn));
|
2016-05-20 00:14:27 +00:00
|
|
|
#else
|
|
|
|
return page_zone(page)->pageblock_flags;
|
|
|
|
#endif /* CONFIG_SPARSEMEM */
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int pfn_to_bitidx(struct page *page, unsigned long pfn)
|
|
|
|
{
|
|
|
|
#ifdef CONFIG_SPARSEMEM
|
|
|
|
pfn &= (PAGES_PER_SECTION-1);
|
|
|
|
#else
|
|
|
|
pfn = pfn - round_down(page_zone(page)->zone_start_pfn, pageblock_nr_pages);
|
|
|
|
#endif /* CONFIG_SPARSEMEM */
|
2020-08-07 06:25:44 +00:00
|
|
|
return (pfn >> pageblock_order) * NR_PAGEBLOCK_BITS;
|
2016-05-20 00:14:27 +00:00
|
|
|
}
|
|
|
|
|
2020-08-07 06:25:51 +00:00
|
|
|
static __always_inline
|
|
|
|
unsigned long __get_pfnblock_flags_mask(struct page *page,
|
2016-05-20 00:14:27 +00:00
|
|
|
unsigned long pfn,
|
|
|
|
unsigned long mask)
|
|
|
|
{
|
|
|
|
unsigned long *bitmap;
|
|
|
|
unsigned long bitidx, word_bitidx;
|
|
|
|
unsigned long word;
|
|
|
|
|
|
|
|
bitmap = get_pageblock_bitmap(page, pfn);
|
|
|
|
bitidx = pfn_to_bitidx(page, pfn);
|
|
|
|
word_bitidx = bitidx / BITS_PER_LONG;
|
|
|
|
bitidx &= (BITS_PER_LONG-1);
|
|
|
|
|
|
|
|
word = bitmap[word_bitidx];
|
2020-08-07 06:25:48 +00:00
|
|
|
return (word >> bitidx) & mask;
|
2016-05-20 00:14:27 +00:00
|
|
|
}
|
|
|
|
|
2020-12-15 03:14:39 +00:00
|
|
|
/**
|
|
|
|
* get_pfnblock_flags_mask - Return the requested group of flags for the pageblock_nr_pages block of pages
|
|
|
|
* @page: The page within the block of interest
|
|
|
|
* @pfn: The target page frame number
|
|
|
|
* @mask: mask of bits that the caller is interested in
|
|
|
|
*
|
|
|
|
* Return: pageblock_bits flags
|
|
|
|
*/
|
2016-05-20 00:14:27 +00:00
|
|
|
unsigned long get_pfnblock_flags_mask(struct page *page, unsigned long pfn,
|
|
|
|
unsigned long mask)
|
|
|
|
{
|
2020-08-07 06:25:51 +00:00
|
|
|
return __get_pfnblock_flags_mask(page, pfn, mask);
|
2016-05-20 00:14:27 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static __always_inline int get_pfnblock_migratetype(struct page *page, unsigned long pfn)
|
|
|
|
{
|
2020-08-07 06:25:51 +00:00
|
|
|
return __get_pfnblock_flags_mask(page, pfn, MIGRATETYPE_MASK);
|
2016-05-20 00:14:27 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* set_pfnblock_flags_mask - Set the requested group of flags for a pageblock_nr_pages block of pages
|
|
|
|
* @page: The page within the block of interest
|
|
|
|
* @flags: The flags to set
|
|
|
|
* @pfn: The target page frame number
|
|
|
|
* @mask: mask of bits that the caller is interested in
|
|
|
|
*/
|
|
|
|
void set_pfnblock_flags_mask(struct page *page, unsigned long flags,
|
|
|
|
unsigned long pfn,
|
|
|
|
unsigned long mask)
|
|
|
|
{
|
|
|
|
unsigned long *bitmap;
|
|
|
|
unsigned long bitidx, word_bitidx;
|
|
|
|
unsigned long old_word, word;
|
|
|
|
|
|
|
|
BUILD_BUG_ON(NR_PAGEBLOCK_BITS != 4);
|
2018-12-28 08:38:43 +00:00
|
|
|
BUILD_BUG_ON(MIGRATE_TYPES > (1 << PB_migratetype_bits));
|
2016-05-20 00:14:27 +00:00
|
|
|
|
|
|
|
bitmap = get_pageblock_bitmap(page, pfn);
|
|
|
|
bitidx = pfn_to_bitidx(page, pfn);
|
|
|
|
word_bitidx = bitidx / BITS_PER_LONG;
|
|
|
|
bitidx &= (BITS_PER_LONG-1);
|
|
|
|
|
|
|
|
VM_BUG_ON_PAGE(!zone_spans_pfn(page_zone(page), pfn), page);
|
|
|
|
|
2020-08-07 06:25:48 +00:00
|
|
|
mask <<= bitidx;
|
|
|
|
flags <<= bitidx;
|
2016-05-20 00:14:27 +00:00
|
|
|
|
|
|
|
word = READ_ONCE(bitmap[word_bitidx]);
|
|
|
|
for (;;) {
|
|
|
|
old_word = cmpxchg(&bitmap[word_bitidx], word, (word & ~mask) | flags);
|
|
|
|
if (word == old_word)
|
|
|
|
break;
|
|
|
|
word = old_word;
|
|
|
|
}
|
|
|
|
}
|
2015-06-30 21:57:02 +00:00
|
|
|
|
2012-07-31 23:43:50 +00:00
|
|
|
void set_pageblock_migratetype(struct page *page, int migratetype)
|
2007-10-16 08:25:48 +00:00
|
|
|
{
|
2013-11-12 23:08:18 +00:00
|
|
|
if (unlikely(page_group_by_mobility_disabled &&
|
|
|
|
migratetype < MIGRATE_PCPTYPES))
|
2009-06-16 22:31:58 +00:00
|
|
|
migratetype = MIGRATE_UNMOVABLE;
|
|
|
|
|
2020-08-07 06:25:48 +00:00
|
|
|
set_pfnblock_flags_mask(page, (unsigned long)migratetype,
|
2020-08-07 06:25:51 +00:00
|
|
|
page_to_pfn(page), MIGRATETYPE_MASK);
|
2007-10-16 08:25:48 +00:00
|
|
|
}
|
|
|
|
|
2006-01-06 08:10:58 +00:00
|
|
|
#ifdef CONFIG_DEBUG_VM
|
2005-10-30 01:16:52 +00:00
|
|
|
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2005-10-30 01:16:53 +00:00
|
|
|
int ret = 0;
|
|
|
|
unsigned seq;
|
|
|
|
unsigned long pfn = page_to_pfn(page);
|
2013-02-23 00:35:28 +00:00
|
|
|
unsigned long sp, start_pfn;
|
2005-10-30 01:16:52 +00:00
|
|
|
|
2005-10-30 01:16:53 +00:00
|
|
|
do {
|
|
|
|
seq = zone_span_seqbegin(zone);
|
2013-02-23 00:35:28 +00:00
|
|
|
start_pfn = zone->zone_start_pfn;
|
|
|
|
sp = zone->spanned_pages;
|
2013-02-23 00:35:23 +00:00
|
|
|
if (!zone_spans_pfn(zone, pfn))
|
2005-10-30 01:16:53 +00:00
|
|
|
ret = 1;
|
|
|
|
} while (zone_span_seqretry(zone, seq));
|
|
|
|
|
2013-02-23 00:35:28 +00:00
|
|
|
if (ret)
|
2014-06-04 23:07:27 +00:00
|
|
|
pr_err("page 0x%lx outside node %d zone %s [ 0x%lx - 0x%lx ]\n",
|
|
|
|
pfn, zone_to_nid(zone), zone->name,
|
|
|
|
start_pfn, start_pfn + sp);
|
2013-02-23 00:35:28 +00:00
|
|
|
|
2005-10-30 01:16:53 +00:00
|
|
|
return ret;
|
2005-10-30 01:16:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int page_is_consistent(struct zone *zone, struct page *page)
|
|
|
|
{
|
2007-05-06 21:49:14 +00:00
|
|
|
if (!pfn_valid_within(page_to_pfn(page)))
|
2005-10-30 01:16:52 +00:00
|
|
|
return 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
if (zone != page_zone(page))
|
2005-10-30 01:16:52 +00:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Temporary debugging check for pages not lying within a given zone.
|
|
|
|
*/
|
2017-07-06 22:39:23 +00:00
|
|
|
static int __maybe_unused bad_range(struct zone *zone, struct page *page)
|
2005-10-30 01:16:52 +00:00
|
|
|
{
|
|
|
|
if (page_outside_zone_boundaries(zone, page))
|
2005-04-16 22:20:36 +00:00
|
|
|
return 1;
|
2005-10-30 01:16:52 +00:00
|
|
|
if (!page_is_consistent(zone, page))
|
|
|
|
return 1;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2006-01-06 08:10:58 +00:00
|
|
|
#else
|
2017-07-06 22:39:23 +00:00
|
|
|
static inline int __maybe_unused bad_range(struct zone *zone, struct page *page)
|
2006-01-06 08:10:58 +00:00
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2020-06-03 22:58:29 +00:00
|
|
|
static void bad_page(struct page *page, const char *reason)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2009-01-06 22:40:12 +00:00
|
|
|
static unsigned long resume;
|
|
|
|
static unsigned long nr_shown;
|
|
|
|
static unsigned long nr_unshown;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allow a burst of 60 reports, then keep quiet for that minute;
|
|
|
|
* or allow a steady drip of one report per second.
|
|
|
|
*/
|
|
|
|
if (nr_shown == 60) {
|
|
|
|
if (time_before(jiffies, resume)) {
|
|
|
|
nr_unshown++;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (nr_unshown) {
|
2016-03-15 21:56:24 +00:00
|
|
|
pr_alert(
|
2009-01-06 22:40:13 +00:00
|
|
|
"BUG: Bad page state: %lu messages suppressed\n",
|
2009-01-06 22:40:12 +00:00
|
|
|
nr_unshown);
|
|
|
|
nr_unshown = 0;
|
|
|
|
}
|
|
|
|
nr_shown = 0;
|
|
|
|
}
|
|
|
|
if (nr_shown++ == 0)
|
|
|
|
resume = jiffies + 60 * HZ;
|
|
|
|
|
2016-03-15 21:56:24 +00:00
|
|
|
pr_alert("BUG: Bad page state in process %s pfn:%05lx\n",
|
badpage: replace page_remove_rmap Eeek and BUG
Now that bad pages are kept out of circulation, there is no need for the
infamous page_remove_rmap() BUG() - once that page is freed, its negative
mapcount will issue a "Bad page state" message and the page won't be
freed. Removing the BUG() allows more info, on subsequent pages, to be
gathered.
We do have more info about the page at this point than bad_page() can know
- notably, what the pmd is, which might pinpoint something like low 64kB
corruption - but page_remove_rmap() isn't given the address to find that.
In practice, there is only one call to page_remove_rmap() which has ever
reported anything, that from zap_pte_range() (usually on exit, sometimes
on munmap). It has all the info, so remove page_remove_rmap()'s "Eeek"
message and leave it all to zap_pte_range().
mm/memory.c already has a hardly used print_bad_pte() function, showing
some of the appropriate info: extend it to show what we want for the rmap
case: pte info, page info (when there is a page) and vma info to compare.
zap_pte_range() already knows the pmd, but print_bad_pte() is easier to
use if it works that out for itself.
Some of this info is also shown in bad_page()'s "Bad page state" message.
Keep them separate, but adjust them to match each other as far as
possible. Say "Bad page map" in print_bad_pte(), and add a TAINT_BAD_PAGE
there too.
print_bad_pte() show current->comm unconditionally (though it should get
repeated in the usually irrelevant stack trace): sorry, I misled Nick
Piggin to make it conditional on vm_mm == current->mm, but current->mm is
already NULL in the exit case. Usually current->comm is good, though
exceptionally it may not be that of the mm (when "swapoff" for example).
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-06 22:40:08 +00:00
|
|
|
current->comm, page_to_pfn(page));
|
2016-03-15 21:56:24 +00:00
|
|
|
__dump_page(page, reason);
|
2016-03-15 21:56:21 +00:00
|
|
|
dump_page_owner(page);
|
badpage: replace page_remove_rmap Eeek and BUG
Now that bad pages are kept out of circulation, there is no need for the
infamous page_remove_rmap() BUG() - once that page is freed, its negative
mapcount will issue a "Bad page state" message and the page won't be
freed. Removing the BUG() allows more info, on subsequent pages, to be
gathered.
We do have more info about the page at this point than bad_page() can know
- notably, what the pmd is, which might pinpoint something like low 64kB
corruption - but page_remove_rmap() isn't given the address to find that.
In practice, there is only one call to page_remove_rmap() which has ever
reported anything, that from zap_pte_range() (usually on exit, sometimes
on munmap). It has all the info, so remove page_remove_rmap()'s "Eeek"
message and leave it all to zap_pte_range().
mm/memory.c already has a hardly used print_bad_pte() function, showing
some of the appropriate info: extend it to show what we want for the rmap
case: pte info, page info (when there is a page) and vma info to compare.
zap_pte_range() already knows the pmd, but print_bad_pte() is easier to
use if it works that out for itself.
Some of this info is also shown in bad_page()'s "Bad page state" message.
Keep them separate, but adjust them to match each other as far as
possible. Say "Bad page map" in print_bad_pte(), and add a TAINT_BAD_PAGE
there too.
print_bad_pte() show current->comm unconditionally (though it should get
repeated in the usually irrelevant stack trace): sorry, I misled Nick
Piggin to make it conditional on vm_mm == current->mm, but current->mm is
already NULL in the exit case. Usually current->comm is good, though
exceptionally it may not be that of the mm (when "swapoff" for example).
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-06 22:40:08 +00:00
|
|
|
|
2011-11-01 00:07:24 +00:00
|
|
|
print_modules();
|
2005-04-16 22:20:36 +00:00
|
|
|
dump_stack();
|
2009-01-06 22:40:12 +00:00
|
|
|
out:
|
2009-01-06 22:40:06 +00:00
|
|
|
/* Leave bad fields for debug, except PageBuddy could make trouble */
|
2013-02-23 00:34:59 +00:00
|
|
|
page_mapcount_reset(page); /* remove PageBuddy */
|
2013-01-21 06:47:39 +00:00
|
|
|
add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Higher-order pages are called "compound pages". They are structured thusly:
|
|
|
|
*
|
2015-11-07 00:29:54 +00:00
|
|
|
* The first PAGE_SIZE page is called the "head page" and have PG_head set.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
2015-11-07 00:29:54 +00:00
|
|
|
* The remaining PAGE_SIZE pages are called "tail pages". PageTail() is encoded
|
|
|
|
* in bit 0 of page->compound_head. The rest of bits is pointer to head page.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
2015-11-07 00:29:54 +00:00
|
|
|
* The first tail page's ->compound_dtor holds the offset in array of compound
|
|
|
|
* page destructors. See compound_page_dtors.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
2015-11-07 00:29:54 +00:00
|
|
|
* The first tail page's ->compound_order holds the order of allocation.
|
[PATCH] compound page: use page[1].lru
If a compound page has its own put_page_testzero destructor (the only current
example is free_huge_page), that is noted in page[1].mapping of the compound
page. But that's rather a poor place to keep it: functions which call
set_page_dirty_lock after get_user_pages (e.g. Infiniband's
__ib_umem_release) ought to be checking first, otherwise set_page_dirty is
liable to crash on what's not the address of a struct address_space.
And now I'm about to make that worse: it turns out that every compound page
needs a destructor, so we can no longer rely on hugetlb pages going their own
special way, to avoid further problems of page->mapping reuse. For example,
not many people know that: on 50% of i386 -Os builds, the first tail page of a
compound page purports to be PageAnon (when its destructor has an odd
address), which surprises page_add_file_rmap.
Keep the compound page destructor in page[1].lru.next instead. And to free up
the common pairing of mapping and index, also move compound page order from
index to lru.prev. Slab reuses page->lru too: but if we ever need slab to use
compound pages, it can easily stack its use above this.
(akpm: decoded version of the above: the tail pages of a compound page now
have ->mapping==NULL, so there's no need for the set_page_dirty[_lock]()
caller to check that they're not compund pages before doing the dirty).
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-14 21:52:58 +00:00
|
|
|
* This usage means that zero-order pages may not be compound.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2006-02-14 21:52:59 +00:00
|
|
|
|
2016-01-16 00:54:17 +00:00
|
|
|
void free_compound_page(struct page *page)
|
2006-02-14 21:52:59 +00:00
|
|
|
{
|
2019-09-23 22:38:09 +00:00
|
|
|
mem_cgroup_uncharge(page);
|
2020-10-16 03:09:35 +00:00
|
|
|
__free_pages_ok(page, compound_order(page), FPI_NONE);
|
2006-02-14 21:52:59 +00:00
|
|
|
}
|
|
|
|
|
2015-11-07 00:29:57 +00:00
|
|
|
void prep_compound_page(struct page *page, unsigned int order)
|
2008-11-06 20:53:27 +00:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
int nr_pages = 1 << order;
|
|
|
|
|
|
|
|
__SetPageHead(page);
|
|
|
|
for (i = 1; i < nr_pages; i++) {
|
|
|
|
struct page *p = page + i;
|
2011-12-08 22:34:18 +00:00
|
|
|
set_page_count(p, 0);
|
2016-01-16 00:52:07 +00:00
|
|
|
p->mapping = TAIL_MAPPING;
|
2015-11-07 00:29:54 +00:00
|
|
|
set_compound_head(p, page);
|
2008-11-06 20:53:27 +00:00
|
|
|
}
|
2020-08-15 00:30:23 +00:00
|
|
|
|
|
|
|
set_compound_page_dtor(page, COMPOUND_PAGE_DTOR);
|
|
|
|
set_compound_order(page, order);
|
2016-01-16 00:53:42 +00:00
|
|
|
atomic_set(compound_mapcount_ptr(page), -1);
|
mm/gup: page->hpage_pinned_refcount: exact pin counts for huge pages
For huge pages (and in fact, any compound page), the GUP_PIN_COUNTING_BIAS
scheme tends to overflow too easily, each tail page increments the head
page->_refcount by GUP_PIN_COUNTING_BIAS (1024). That limits the number
of huge pages that can be pinned.
This patch removes that limitation, by using an exact form of pin counting
for compound pages of order > 1. The "order > 1" is required because this
approach uses the 3rd struct page in the compound page, and order 1
compound pages only have two pages, so that won't work there.
A new struct page field, hpage_pinned_refcount, has been added, replacing
a padding field in the union (so no new space is used).
This enhancement also has a useful side effect: huge pages and compound
pages (of order > 1) do not suffer from the "potential false positives"
problem that is discussed in the page_dma_pinned() comment block. That is
because these compound pages have extra space for tracking things, so they
get exact pin counts instead of overloading page->_refcount.
Documentation/core-api/pin_user_pages.rst is updated accordingly.
Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jérôme Glisse <jglisse@redhat.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Link: http://lkml.kernel.org/r/20200211001536.1027652-8-jhubbard@nvidia.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-04-02 04:05:33 +00:00
|
|
|
if (hpage_pincount_available(page))
|
|
|
|
atomic_set(compound_pincount_ptr(page), 0);
|
2008-11-06 20:53:27 +00:00
|
|
|
}
|
|
|
|
|
2012-01-10 23:07:28 +00:00
|
|
|
#ifdef CONFIG_DEBUG_PAGEALLOC
|
|
|
|
unsigned int _debug_guardpage_minorder;
|
2019-07-12 03:55:06 +00:00
|
|
|
|
mm, debug_pagealloc: don't rely on static keys too early
Commit 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable
debugging") has introduced a static key to reduce overhead when
debug_pagealloc is compiled in but not enabled. It relied on the
assumption that jump_label_init() is called before parse_early_param()
as in start_kernel(), so when the "debug_pagealloc=on" option is parsed,
it is safe to enable the static key.
However, it turns out multiple architectures call parse_early_param()
earlier from their setup_arch(). x86 also calls jump_label_init() even
earlier, so no issue was found while testing the commit, but same is not
true for e.g. ppc64 and s390 where the kernel would not boot with
debug_pagealloc=on as found by our QA.
To fix this without tricky changes to init code of multiple
architectures, this patch partially reverts the static key conversion
from 96a2b03f281d. Init-time and non-fastpath calls (such as in arch
code) of debug_pagealloc_enabled() will again test a simple bool
variable. Fastpath mm code is converted to a new
debug_pagealloc_enabled_static() variant that relies on the static key,
which is enabled in a well-defined point in mm_init() where it's
guaranteed that jump_label_init() has been called, regardless of
architecture.
[sfr@canb.auug.org.au: export _debug_pagealloc_enabled_early]
Link: http://lkml.kernel.org/r/20200106164944.063ac07b@canb.auug.org.au
Link: http://lkml.kernel.org/r/20191219130612.23171-1-vbabka@suse.cz
Fixes: 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable debugging")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Qian Cai <cai@lca.pw>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-14 00:29:20 +00:00
|
|
|
bool _debug_pagealloc_enabled_early __read_mostly
|
|
|
|
= IS_ENABLED(CONFIG_DEBUG_PAGEALLOC_ENABLE_DEFAULT);
|
|
|
|
EXPORT_SYMBOL(_debug_pagealloc_enabled_early);
|
2019-07-12 03:55:06 +00:00
|
|
|
DEFINE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
|
2016-03-17 21:17:56 +00:00
|
|
|
EXPORT_SYMBOL(_debug_pagealloc_enabled);
|
2019-07-12 03:55:06 +00:00
|
|
|
|
|
|
|
DEFINE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
|
mm/debug-pagealloc: prepare boottime configurable on/off
Until now, debug-pagealloc needs extra flags in struct page, so we need to
recompile whole source code when we decide to use it. This is really
painful, because it takes some time to recompile and sometimes rebuild is
not possible due to third party module depending on struct page. So, we
can't use this good feature in many cases.
Now, we have the page extension feature that allows us to insert extra
flags to outside of struct page. This gets rid of third party module
issue mentioned above. And, this allows us to determine if we need extra
memory for this page extension in boottime. With these property, we can
avoid using debug-pagealloc in boottime with low computational overhead in
the kernel built with CONFIG_DEBUG_PAGEALLOC. This will help our
development process greatly.
This patch is the preparation step to achive above goal. debug-pagealloc
originally uses extra field of struct page, but, after this patch, it will
use field of struct page_ext. Because memory for page_ext is allocated
later than initialization of page allocator in CONFIG_SPARSEMEM, we should
disable debug-pagealloc feature temporarily until initialization of
page_ext. This patch implements this.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dave Hansen <dave@sr71.net>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Jungsoo Son <jungsoo.son@lge.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:55:49 +00:00
|
|
|
|
2014-12-13 00:55:52 +00:00
|
|
|
static int __init early_debug_pagealloc(char *buf)
|
|
|
|
{
|
mm, debug_pagealloc: don't rely on static keys too early
Commit 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable
debugging") has introduced a static key to reduce overhead when
debug_pagealloc is compiled in but not enabled. It relied on the
assumption that jump_label_init() is called before parse_early_param()
as in start_kernel(), so when the "debug_pagealloc=on" option is parsed,
it is safe to enable the static key.
However, it turns out multiple architectures call parse_early_param()
earlier from their setup_arch(). x86 also calls jump_label_init() even
earlier, so no issue was found while testing the commit, but same is not
true for e.g. ppc64 and s390 where the kernel would not boot with
debug_pagealloc=on as found by our QA.
To fix this without tricky changes to init code of multiple
architectures, this patch partially reverts the static key conversion
from 96a2b03f281d. Init-time and non-fastpath calls (such as in arch
code) of debug_pagealloc_enabled() will again test a simple bool
variable. Fastpath mm code is converted to a new
debug_pagealloc_enabled_static() variant that relies on the static key,
which is enabled in a well-defined point in mm_init() where it's
guaranteed that jump_label_init() has been called, regardless of
architecture.
[sfr@canb.auug.org.au: export _debug_pagealloc_enabled_early]
Link: http://lkml.kernel.org/r/20200106164944.063ac07b@canb.auug.org.au
Link: http://lkml.kernel.org/r/20191219130612.23171-1-vbabka@suse.cz
Fixes: 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable debugging")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Qian Cai <cai@lca.pw>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-14 00:29:20 +00:00
|
|
|
return kstrtobool(buf, &_debug_pagealloc_enabled_early);
|
2014-12-13 00:55:52 +00:00
|
|
|
}
|
|
|
|
early_param("debug_pagealloc", early_debug_pagealloc);
|
|
|
|
|
2012-01-10 23:07:28 +00:00
|
|
|
static int __init debug_guardpage_minorder_setup(char *buf)
|
|
|
|
{
|
|
|
|
unsigned long res;
|
|
|
|
|
|
|
|
if (kstrtoul(buf, 10, &res) < 0 || res > MAX_ORDER / 2) {
|
2016-03-17 21:19:50 +00:00
|
|
|
pr_err("Bad debug_guardpage_minorder value\n");
|
2012-01-10 23:07:28 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
_debug_guardpage_minorder = res;
|
2016-03-17 21:19:50 +00:00
|
|
|
pr_info("Setting debug_guardpage_minorder to %lu\n", res);
|
2012-01-10 23:07:28 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2016-10-07 23:58:18 +00:00
|
|
|
early_param("debug_guardpage_minorder", debug_guardpage_minorder_setup);
|
2012-01-10 23:07:28 +00:00
|
|
|
|
2016-10-07 23:58:15 +00:00
|
|
|
static inline bool set_page_guard(struct zone *zone, struct page *page,
|
2014-12-13 00:55:01 +00:00
|
|
|
unsigned int order, int migratetype)
|
2012-01-10 23:07:28 +00:00
|
|
|
{
|
mm/debug-pagealloc: prepare boottime configurable on/off
Until now, debug-pagealloc needs extra flags in struct page, so we need to
recompile whole source code when we decide to use it. This is really
painful, because it takes some time to recompile and sometimes rebuild is
not possible due to third party module depending on struct page. So, we
can't use this good feature in many cases.
Now, we have the page extension feature that allows us to insert extra
flags to outside of struct page. This gets rid of third party module
issue mentioned above. And, this allows us to determine if we need extra
memory for this page extension in boottime. With these property, we can
avoid using debug-pagealloc in boottime with low computational overhead in
the kernel built with CONFIG_DEBUG_PAGEALLOC. This will help our
development process greatly.
This patch is the preparation step to achive above goal. debug-pagealloc
originally uses extra field of struct page, but, after this patch, it will
use field of struct page_ext. Because memory for page_ext is allocated
later than initialization of page allocator in CONFIG_SPARSEMEM, we should
disable debug-pagealloc feature temporarily until initialization of
page_ext. This patch implements this.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dave Hansen <dave@sr71.net>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Jungsoo Son <jungsoo.son@lge.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:55:49 +00:00
|
|
|
if (!debug_guardpage_enabled())
|
2016-10-07 23:58:15 +00:00
|
|
|
return false;
|
|
|
|
|
|
|
|
if (order >= debug_guardpage_minorder())
|
|
|
|
return false;
|
mm/debug-pagealloc: prepare boottime configurable on/off
Until now, debug-pagealloc needs extra flags in struct page, so we need to
recompile whole source code when we decide to use it. This is really
painful, because it takes some time to recompile and sometimes rebuild is
not possible due to third party module depending on struct page. So, we
can't use this good feature in many cases.
Now, we have the page extension feature that allows us to insert extra
flags to outside of struct page. This gets rid of third party module
issue mentioned above. And, this allows us to determine if we need extra
memory for this page extension in boottime. With these property, we can
avoid using debug-pagealloc in boottime with low computational overhead in
the kernel built with CONFIG_DEBUG_PAGEALLOC. This will help our
development process greatly.
This patch is the preparation step to achive above goal. debug-pagealloc
originally uses extra field of struct page, but, after this patch, it will
use field of struct page_ext. Because memory for page_ext is allocated
later than initialization of page allocator in CONFIG_SPARSEMEM, we should
disable debug-pagealloc feature temporarily until initialization of
page_ext. This patch implements this.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dave Hansen <dave@sr71.net>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Jungsoo Son <jungsoo.son@lge.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:55:49 +00:00
|
|
|
|
2019-07-12 03:55:13 +00:00
|
|
|
__SetPageGuard(page);
|
2014-12-13 00:55:01 +00:00
|
|
|
INIT_LIST_HEAD(&page->lru);
|
|
|
|
set_page_private(page, order);
|
|
|
|
/* Guard pages are not available for any usage */
|
|
|
|
__mod_zone_freepage_state(zone, -(1 << order), migratetype);
|
2016-10-07 23:58:15 +00:00
|
|
|
|
|
|
|
return true;
|
2012-01-10 23:07:28 +00:00
|
|
|
}
|
|
|
|
|
2014-12-13 00:55:01 +00:00
|
|
|
static inline void clear_page_guard(struct zone *zone, struct page *page,
|
|
|
|
unsigned int order, int migratetype)
|
2012-01-10 23:07:28 +00:00
|
|
|
{
|
mm/debug-pagealloc: prepare boottime configurable on/off
Until now, debug-pagealloc needs extra flags in struct page, so we need to
recompile whole source code when we decide to use it. This is really
painful, because it takes some time to recompile and sometimes rebuild is
not possible due to third party module depending on struct page. So, we
can't use this good feature in many cases.
Now, we have the page extension feature that allows us to insert extra
flags to outside of struct page. This gets rid of third party module
issue mentioned above. And, this allows us to determine if we need extra
memory for this page extension in boottime. With these property, we can
avoid using debug-pagealloc in boottime with low computational overhead in
the kernel built with CONFIG_DEBUG_PAGEALLOC. This will help our
development process greatly.
This patch is the preparation step to achive above goal. debug-pagealloc
originally uses extra field of struct page, but, after this patch, it will
use field of struct page_ext. Because memory for page_ext is allocated
later than initialization of page allocator in CONFIG_SPARSEMEM, we should
disable debug-pagealloc feature temporarily until initialization of
page_ext. This patch implements this.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dave Hansen <dave@sr71.net>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Jungsoo Son <jungsoo.son@lge.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:55:49 +00:00
|
|
|
if (!debug_guardpage_enabled())
|
|
|
|
return;
|
|
|
|
|
2019-07-12 03:55:13 +00:00
|
|
|
__ClearPageGuard(page);
|
mm/debug-pagealloc: prepare boottime configurable on/off
Until now, debug-pagealloc needs extra flags in struct page, so we need to
recompile whole source code when we decide to use it. This is really
painful, because it takes some time to recompile and sometimes rebuild is
not possible due to third party module depending on struct page. So, we
can't use this good feature in many cases.
Now, we have the page extension feature that allows us to insert extra
flags to outside of struct page. This gets rid of third party module
issue mentioned above. And, this allows us to determine if we need extra
memory for this page extension in boottime. With these property, we can
avoid using debug-pagealloc in boottime with low computational overhead in
the kernel built with CONFIG_DEBUG_PAGEALLOC. This will help our
development process greatly.
This patch is the preparation step to achive above goal. debug-pagealloc
originally uses extra field of struct page, but, after this patch, it will
use field of struct page_ext. Because memory for page_ext is allocated
later than initialization of page allocator in CONFIG_SPARSEMEM, we should
disable debug-pagealloc feature temporarily until initialization of
page_ext. This patch implements this.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dave Hansen <dave@sr71.net>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Jungsoo Son <jungsoo.son@lge.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:55:49 +00:00
|
|
|
|
2014-12-13 00:55:01 +00:00
|
|
|
set_page_private(page, 0);
|
|
|
|
if (!is_migrate_isolate(migratetype))
|
|
|
|
__mod_zone_freepage_state(zone, (1 << order), migratetype);
|
2012-01-10 23:07:28 +00:00
|
|
|
}
|
|
|
|
#else
|
2016-10-07 23:58:15 +00:00
|
|
|
static inline bool set_page_guard(struct zone *zone, struct page *page,
|
|
|
|
unsigned int order, int migratetype) { return false; }
|
2014-12-13 00:55:01 +00:00
|
|
|
static inline void clear_page_guard(struct zone *zone, struct page *page,
|
|
|
|
unsigned int order, int migratetype) {}
|
2012-01-10 23:07:28 +00:00
|
|
|
#endif
|
|
|
|
|
mm, page_alloc: do not rely on the order of page_poison and init_on_alloc/free parameters
Patch series "cleanup page poisoning", v3.
I have identified a number of issues and opportunities for cleanup with
CONFIG_PAGE_POISON and friends:
- interaction with init_on_alloc and init_on_free parameters depends on
the order of parameters (Patch 1)
- the boot time enabling uses static key, but inefficienty (Patch 2)
- sanity checking is incompatible with hibernation (Patch 3)
- CONFIG_PAGE_POISONING_NO_SANITY can be removed now that we have
init_on_free (Patch 4)
- CONFIG_PAGE_POISONING_ZERO can be most likely removed now that we
have init_on_free (Patch 5)
This patch (of 5):
Enabling page_poison=1 together with init_on_alloc=1 or init_on_free=1
produces a warning in dmesg that page_poison takes precedence. However,
as these warnings are printed in early_param handlers for
init_on_alloc/free, they are not printed if page_poison is enabled later
on the command line (handlers are called in the order of their
parameters), or when init_on_alloc/free is always enabled by the
respective config option - before the page_poison early param handler is
called, it is not considered to be enabled. This is inconsistent.
We can remove the dependency on order by making the init_on_* parameters
only set a boolean variable, and postponing the evaluation after all early
params have been processed. Introduce a new
init_mem_debugging_and_hardening() function for that, and move the related
debug_pagealloc processing there as well.
As a result init_mem_debugging_and_hardening() knows always accurately if
init_on_* and/or page_poison options were enabled. Thus we can also
optimize want_init_on_alloc() and want_init_on_free(). We don't need to
check page_poisoning_enabled() there, we can instead not enable the
init_on_* static keys at all, if page poisoning is enabled. This results
in a simpler and more effective code.
Link: https://lkml.kernel.org/r/20201113104033.22907-1-vbabka@suse.cz
Link: https://lkml.kernel.org/r/20201113104033.22907-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mateusz Nosek <mateusznosek0@gmail.com>
Cc: Laura Abbott <labbott@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:13:30 +00:00
|
|
|
/*
|
|
|
|
* Enable static keys related to various memory debugging and hardening options.
|
|
|
|
* Some override others, and depend on early params that are evaluated in the
|
|
|
|
* order of appearance. So we need to first gather the full picture of what was
|
|
|
|
* enabled, and then make decisions.
|
|
|
|
*/
|
|
|
|
void init_mem_debugging_and_hardening(void)
|
|
|
|
{
|
|
|
|
if (_init_on_alloc_enabled_early) {
|
|
|
|
if (page_poisoning_enabled())
|
|
|
|
pr_info("mem auto-init: CONFIG_PAGE_POISONING is on, "
|
|
|
|
"will take precedence over init_on_alloc\n");
|
|
|
|
else
|
|
|
|
static_branch_enable(&init_on_alloc);
|
|
|
|
}
|
|
|
|
if (_init_on_free_enabled_early) {
|
|
|
|
if (page_poisoning_enabled())
|
|
|
|
pr_info("mem auto-init: CONFIG_PAGE_POISONING is on, "
|
|
|
|
"will take precedence over init_on_free\n");
|
|
|
|
else
|
|
|
|
static_branch_enable(&init_on_free);
|
|
|
|
}
|
|
|
|
|
2020-12-15 03:13:34 +00:00
|
|
|
#ifdef CONFIG_PAGE_POISONING
|
|
|
|
/*
|
|
|
|
* Page poisoning is debug page alloc for some arches. If
|
|
|
|
* either of those options are enabled, enable poisoning.
|
|
|
|
*/
|
|
|
|
if (page_poisoning_enabled() ||
|
|
|
|
(!IS_ENABLED(CONFIG_ARCH_SUPPORTS_DEBUG_PAGEALLOC) &&
|
|
|
|
debug_pagealloc_enabled()))
|
|
|
|
static_branch_enable(&_page_poisoning_enabled);
|
|
|
|
#endif
|
|
|
|
|
mm, page_alloc: do not rely on the order of page_poison and init_on_alloc/free parameters
Patch series "cleanup page poisoning", v3.
I have identified a number of issues and opportunities for cleanup with
CONFIG_PAGE_POISON and friends:
- interaction with init_on_alloc and init_on_free parameters depends on
the order of parameters (Patch 1)
- the boot time enabling uses static key, but inefficienty (Patch 2)
- sanity checking is incompatible with hibernation (Patch 3)
- CONFIG_PAGE_POISONING_NO_SANITY can be removed now that we have
init_on_free (Patch 4)
- CONFIG_PAGE_POISONING_ZERO can be most likely removed now that we
have init_on_free (Patch 5)
This patch (of 5):
Enabling page_poison=1 together with init_on_alloc=1 or init_on_free=1
produces a warning in dmesg that page_poison takes precedence. However,
as these warnings are printed in early_param handlers for
init_on_alloc/free, they are not printed if page_poison is enabled later
on the command line (handlers are called in the order of their
parameters), or when init_on_alloc/free is always enabled by the
respective config option - before the page_poison early param handler is
called, it is not considered to be enabled. This is inconsistent.
We can remove the dependency on order by making the init_on_* parameters
only set a boolean variable, and postponing the evaluation after all early
params have been processed. Introduce a new
init_mem_debugging_and_hardening() function for that, and move the related
debug_pagealloc processing there as well.
As a result init_mem_debugging_and_hardening() knows always accurately if
init_on_* and/or page_poison options were enabled. Thus we can also
optimize want_init_on_alloc() and want_init_on_free(). We don't need to
check page_poisoning_enabled() there, we can instead not enable the
init_on_* static keys at all, if page poisoning is enabled. This results
in a simpler and more effective code.
Link: https://lkml.kernel.org/r/20201113104033.22907-1-vbabka@suse.cz
Link: https://lkml.kernel.org/r/20201113104033.22907-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mateusz Nosek <mateusznosek0@gmail.com>
Cc: Laura Abbott <labbott@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:13:30 +00:00
|
|
|
#ifdef CONFIG_DEBUG_PAGEALLOC
|
|
|
|
if (!debug_pagealloc_enabled())
|
|
|
|
return;
|
|
|
|
|
|
|
|
static_branch_enable(&_debug_pagealloc_enabled);
|
|
|
|
|
|
|
|
if (!debug_guardpage_minorder())
|
|
|
|
return;
|
|
|
|
|
|
|
|
static_branch_enable(&_debug_guardpage_enabled);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2020-10-16 03:10:15 +00:00
|
|
|
static inline void set_buddy_order(struct page *page, unsigned int order)
|
2006-04-19 05:20:52 +00:00
|
|
|
{
|
[PATCH] mm: split page table lock
Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with
a many-threaded application which concurrently initializes different parts of
a large anonymous area.
This patch corrects that, by using a separate spinlock per page table page, to
guard the page table entries in that page, instead of using the mm's single
page_table_lock. (But even then, page_table_lock is still used to guard page
table allocation, and anon_vma allocation.)
In this implementation, the spinlock is tucked inside the struct page of the
page table page: with a BUILD_BUG_ON in case it overflows - which it would in
the case of 32-bit PA-RISC with spinlock debugging enabled.
Splitting the lock is not quite for free: another cacheline access. Ideally,
I suppose we would use split ptlock only for multi-threaded processes on
multi-cpu machines; but deciding that dynamically would have its own costs.
So for now enable it by config, at some number of cpus - since the Kconfig
language doesn't support inequalities, let preprocessor compare that with
NR_CPUS. But I don't think it's worth being user-configurable: for good
testing of both split and unsplit configs, split now at 4 cpus, and perhaps
change that to 8 later.
There is a benefit even for singly threaded processes: kswapd can be attacking
one part of the mm while another part is busy faulting.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
|
|
|
set_page_private(page, order);
|
2006-04-10 01:21:48 +00:00
|
|
|
__SetPageBuddy(page);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This function checks whether a page is free && is the buddy
|
2018-06-08 00:08:18 +00:00
|
|
|
* we can coalesce a page and its buddy if
|
2017-02-22 23:41:51 +00:00
|
|
|
* (a) the buddy is not in a hole (check before calling!) &&
|
2006-04-10 01:21:48 +00:00
|
|
|
* (b) the buddy is in the buddy system &&
|
2006-06-23 09:03:01 +00:00
|
|
|
* (c) a page and its buddy have the same order &&
|
|
|
|
* (d) a page and its buddy are in the same zone.
|
2006-04-10 01:21:48 +00:00
|
|
|
*
|
2018-06-08 00:08:18 +00:00
|
|
|
* For recording whether a page is in the buddy system, we set PageBuddy.
|
|
|
|
* Setting, clearing, and testing PageBuddy is serialized by zone->lock.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
2006-04-10 01:21:48 +00:00
|
|
|
* For recording page's order, we use page_private(page).
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2020-04-02 04:09:56 +00:00
|
|
|
static inline bool page_is_buddy(struct page *page, struct page *buddy,
|
2014-06-04 23:10:21 +00:00
|
|
|
unsigned int order)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2020-04-02 04:09:56 +00:00
|
|
|
if (!page_is_guard(buddy) && !PageBuddy(buddy))
|
|
|
|
return false;
|
2015-02-10 22:11:39 +00:00
|
|
|
|
2020-10-16 03:10:15 +00:00
|
|
|
if (buddy_order(buddy) != order)
|
2020-04-02 04:09:56 +00:00
|
|
|
return false;
|
2012-01-10 23:07:28 +00:00
|
|
|
|
2020-04-02 04:09:56 +00:00
|
|
|
/*
|
|
|
|
* zone check is done late to avoid uselessly calculating
|
|
|
|
* zone/node ids for pages that could never merge.
|
|
|
|
*/
|
|
|
|
if (page_zone_id(page) != page_zone_id(buddy))
|
|
|
|
return false;
|
2014-06-04 23:10:10 +00:00
|
|
|
|
2020-04-02 04:09:56 +00:00
|
|
|
VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
|
2015-02-10 22:11:39 +00:00
|
|
|
|
2020-04-02 04:09:56 +00:00
|
|
|
return true;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2019-03-05 23:45:41 +00:00
|
|
|
#ifdef CONFIG_COMPACTION
|
|
|
|
static inline struct capture_control *task_capc(struct zone *zone)
|
|
|
|
{
|
|
|
|
struct capture_control *capc = current->capture_control;
|
|
|
|
|
2020-08-07 06:25:16 +00:00
|
|
|
return unlikely(capc) &&
|
2019-03-05 23:45:41 +00:00
|
|
|
!(current->flags & PF_KTHREAD) &&
|
|
|
|
!capc->page &&
|
2020-08-07 06:25:16 +00:00
|
|
|
capc->cc->zone == zone ? capc : NULL;
|
2019-03-05 23:45:41 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline bool
|
|
|
|
compaction_capture(struct capture_control *capc, struct page *page,
|
|
|
|
int order, int migratetype)
|
|
|
|
{
|
|
|
|
if (!capc || order != capc->cc->order)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
/* Do not accidentally pollute CMA or isolated regions*/
|
|
|
|
if (is_migrate_cma(migratetype) ||
|
|
|
|
is_migrate_isolate(migratetype))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do not let lower order allocations polluate a movable pageblock.
|
|
|
|
* This might let an unmovable request use a reclaimable pageblock
|
|
|
|
* and vice-versa but no more than normal fallback logic which can
|
|
|
|
* have trouble finding a high-order free page.
|
|
|
|
*/
|
|
|
|
if (order < pageblock_order && migratetype == MIGRATE_MOVABLE)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
capc->page = page;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
#else
|
|
|
|
static inline struct capture_control *task_capc(struct zone *zone)
|
|
|
|
{
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline bool
|
|
|
|
compaction_capture(struct capture_control *capc, struct page *page,
|
|
|
|
int order, int migratetype)
|
|
|
|
{
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_COMPACTION */
|
|
|
|
|
2020-04-07 03:04:49 +00:00
|
|
|
/* Used for pages not on another list */
|
|
|
|
static inline void add_to_free_list(struct page *page, struct zone *zone,
|
|
|
|
unsigned int order, int migratetype)
|
|
|
|
{
|
|
|
|
struct free_area *area = &zone->free_area[order];
|
|
|
|
|
|
|
|
list_add(&page->lru, &area->free_list[migratetype]);
|
|
|
|
area->nr_free++;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Used for pages not on another list */
|
|
|
|
static inline void add_to_free_list_tail(struct page *page, struct zone *zone,
|
|
|
|
unsigned int order, int migratetype)
|
|
|
|
{
|
|
|
|
struct free_area *area = &zone->free_area[order];
|
|
|
|
|
|
|
|
list_add_tail(&page->lru, &area->free_list[migratetype]);
|
|
|
|
area->nr_free++;
|
|
|
|
}
|
|
|
|
|
mm/page_alloc: move pages to tail in move_to_free_list()
Whenever we move pages between freelists via move_to_free_list()/
move_freepages_block(), we don't actually touch the pages:
1. Page isolation doesn't actually touch the pages, it simply isolates
pageblocks and moves all free pages to the MIGRATE_ISOLATE freelist.
When undoing isolation, we move the pages back to the target list.
2. Page stealing (steal_suitable_fallback()) moves free pages directly
between lists without touching them.
3. reserve_highatomic_pageblock()/unreserve_highatomic_pageblock() moves
free pages directly between freelists without touching them.
We already place pages to the tail of the freelists when undoing isolation
via __putback_isolated_page(), let's do it in any case (e.g., if order <=
pageblock_order) and document the behavior. To simplify, let's move the
pages to the tail for all move_to_free_list()/move_freepages_block() users.
In 2., the target list is empty, so there should be no change. In 3., we
might observe a change, however, highatomic is more concerned about
allocations succeeding than cache hotness - if we ever realize this change
degrades a workload, we can special-case this instance and add a proper
comment.
This change results in all pages getting onlined via online_pages() to be
placed to the tail of the freelist.
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Acked-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-4-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:30 +00:00
|
|
|
/*
|
|
|
|
* Used for pages which are on another list. Move the pages to the tail
|
|
|
|
* of the list - so the moved pages won't immediately be considered for
|
|
|
|
* allocation again (e.g., optimization for memory onlining).
|
|
|
|
*/
|
2020-04-07 03:04:49 +00:00
|
|
|
static inline void move_to_free_list(struct page *page, struct zone *zone,
|
|
|
|
unsigned int order, int migratetype)
|
|
|
|
{
|
|
|
|
struct free_area *area = &zone->free_area[order];
|
|
|
|
|
mm/page_alloc: move pages to tail in move_to_free_list()
Whenever we move pages between freelists via move_to_free_list()/
move_freepages_block(), we don't actually touch the pages:
1. Page isolation doesn't actually touch the pages, it simply isolates
pageblocks and moves all free pages to the MIGRATE_ISOLATE freelist.
When undoing isolation, we move the pages back to the target list.
2. Page stealing (steal_suitable_fallback()) moves free pages directly
between lists without touching them.
3. reserve_highatomic_pageblock()/unreserve_highatomic_pageblock() moves
free pages directly between freelists without touching them.
We already place pages to the tail of the freelists when undoing isolation
via __putback_isolated_page(), let's do it in any case (e.g., if order <=
pageblock_order) and document the behavior. To simplify, let's move the
pages to the tail for all move_to_free_list()/move_freepages_block() users.
In 2., the target list is empty, so there should be no change. In 3., we
might observe a change, however, highatomic is more concerned about
allocations succeeding than cache hotness - if we ever realize this change
degrades a workload, we can special-case this instance and add a proper
comment.
This change results in all pages getting onlined via online_pages() to be
placed to the tail of the freelist.
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Acked-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-4-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:30 +00:00
|
|
|
list_move_tail(&page->lru, &area->free_list[migratetype]);
|
2020-04-07 03:04:49 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void del_page_from_free_list(struct page *page, struct zone *zone,
|
|
|
|
unsigned int order)
|
|
|
|
{
|
2020-04-07 03:04:56 +00:00
|
|
|
/* clear reported state and update reported page count */
|
|
|
|
if (page_reported(page))
|
|
|
|
__ClearPageReported(page);
|
|
|
|
|
2020-04-07 03:04:49 +00:00
|
|
|
list_del(&page->lru);
|
|
|
|
__ClearPageBuddy(page);
|
|
|
|
set_page_private(page, 0);
|
|
|
|
zone->free_area[order].nr_free--;
|
|
|
|
}
|
|
|
|
|
2020-04-07 03:04:45 +00:00
|
|
|
/*
|
|
|
|
* If this is not the largest possible page, check if the buddy
|
|
|
|
* of the next-highest order is free. If it is, it's possible
|
|
|
|
* that pages are being freed that will coalesce soon. In case,
|
|
|
|
* that is happening, add the free page to the tail of the list
|
|
|
|
* so it's less likely to be used soon and more likely to be merged
|
|
|
|
* as a higher order page
|
|
|
|
*/
|
|
|
|
static inline bool
|
|
|
|
buddy_merge_likely(unsigned long pfn, unsigned long buddy_pfn,
|
|
|
|
struct page *page, unsigned int order)
|
|
|
|
{
|
|
|
|
struct page *higher_page, *higher_buddy;
|
|
|
|
unsigned long combined_pfn;
|
|
|
|
|
|
|
|
if (order >= MAX_ORDER - 2)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (!pfn_valid_within(buddy_pfn))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
combined_pfn = buddy_pfn & pfn;
|
|
|
|
higher_page = page + (combined_pfn - pfn);
|
|
|
|
buddy_pfn = __find_buddy_pfn(combined_pfn, order + 1);
|
|
|
|
higher_buddy = higher_page + (buddy_pfn - combined_pfn);
|
|
|
|
|
|
|
|
return pfn_valid_within(buddy_pfn) &&
|
|
|
|
page_is_buddy(higher_page, higher_buddy, order + 1);
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Freeing function for a buddy system allocator.
|
|
|
|
*
|
|
|
|
* The concept of a buddy system is to maintain direct-mapped table
|
|
|
|
* (containing bit values) for memory blocks of various "orders".
|
|
|
|
* The bottom level table contains the map for the smallest allocatable
|
|
|
|
* units of memory (here, pages), and each level above it describes
|
|
|
|
* pairs of units from the levels below, hence, "buddies".
|
|
|
|
* At a high level, all that happens here is marking the table entry
|
|
|
|
* at the bottom level available, and propagating the changes upward
|
|
|
|
* as necessary, plus some accounting needed to play nicely with other
|
|
|
|
* parts of the VM system.
|
|
|
|
* At each level, we keep a list of pages, which are heads of continuous
|
2018-06-08 00:08:18 +00:00
|
|
|
* free pages of length of (1 << order) and marked with PageBuddy.
|
|
|
|
* Page's order is recorded in page_private(page) field.
|
2005-04-16 22:20:36 +00:00
|
|
|
* So when we are allocating or freeing one, we can derive the state of the
|
2012-01-11 14:16:11 +00:00
|
|
|
* other. That is, if we allocate a small block, and both were
|
|
|
|
* free, the remainder of the region must be split into blocks.
|
2005-04-16 22:20:36 +00:00
|
|
|
* If a block is freed, and its buddy is also free, then this
|
2012-01-11 14:16:11 +00:00
|
|
|
* triggers coalescing into a block of larger size.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
2012-12-06 09:39:54 +00:00
|
|
|
* -- nyc
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
|
|
|
|
2006-01-08 09:00:42 +00:00
|
|
|
static inline void __free_one_page(struct page *page,
|
2014-06-04 23:10:17 +00:00
|
|
|
unsigned long pfn,
|
2009-06-16 22:32:07 +00:00
|
|
|
struct zone *zone, unsigned int order,
|
mm/page_alloc: convert "report" flag of __free_one_page() to a proper flag
Patch series "mm: place pages to the freelist tail when onlining and undoing isolation", v2.
When adding separate memory blocks via add_memory*() and onlining them
immediately, the metadata (especially the memmap) of the next block will
be placed onto one of the just added+onlined block. This creates a chain
of unmovable allocations: If the last memory block cannot get
offlined+removed() so will all dependent ones. We directly have unmovable
allocations all over the place.
This can be observed quite easily using virtio-mem, however, it can also
be observed when using DIMMs. The freshly onlined pages will usually be
placed to the head of the freelists, meaning they will be allocated next,
turning the just-added memory usually immediately un-removable. The fresh
pages are cold, prefering to allocate others (that might be hot) also
feels to be the natural thing to do.
It also applies to the hyper-v balloon xen-balloon, and ppc64 dlpar: when
adding separate, successive memory blocks, each memory block will have
unmovable allocations on them - for example gigantic pages will fail to
allocate.
While the ZONE_NORMAL doesn't provide any guarantees that memory can get
offlined+removed again (any kind of fragmentation with unmovable
allocations is possible), there are many scenarios (hotplugging a lot of
memory, running workload, hotunplug some memory/as much as possible) where
we can offline+remove quite a lot with this patchset.
a) To visualize the problem, a very simple example:
Start a VM with 4GB and 8GB of virtio-mem memory:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000033fffffff 9G online yes 32-103
Memory block size: 128M
Total online memory: 12G
Total offline memory: 0B
Then try to unplug as much as possible using virtio-mem. Observe which
memory blocks are still around. Without this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
0x0000000148000000-0x000000014fffffff 128M online yes 41
0x0000000158000000-0x000000015fffffff 128M online yes 43
0x0000000168000000-0x000000016fffffff 128M online yes 45
0x0000000178000000-0x000000017fffffff 128M online yes 47
0x0000000188000000-0x0000000197ffffff 256M online yes 49-50
0x00000001a0000000-0x00000001a7ffffff 128M online yes 52
0x00000001b0000000-0x00000001b7ffffff 128M online yes 54
0x00000001c0000000-0x00000001c7ffffff 128M online yes 56
0x00000001d0000000-0x00000001d7ffffff 128M online yes 58
0x00000001e0000000-0x00000001e7ffffff 128M online yes 60
0x00000001f0000000-0x00000001f7ffffff 128M online yes 62
0x0000000200000000-0x0000000207ffffff 128M online yes 64
0x0000000210000000-0x0000000217ffffff 128M online yes 66
0x0000000220000000-0x0000000227ffffff 128M online yes 68
0x0000000230000000-0x0000000237ffffff 128M online yes 70
0x0000000240000000-0x0000000247ffffff 128M online yes 72
0x0000000250000000-0x0000000257ffffff 128M online yes 74
0x0000000260000000-0x0000000267ffffff 128M online yes 76
0x0000000270000000-0x0000000277ffffff 128M online yes 78
0x0000000280000000-0x0000000287ffffff 128M online yes 80
0x0000000290000000-0x0000000297ffffff 128M online yes 82
0x00000002a0000000-0x00000002a7ffffff 128M online yes 84
0x00000002b0000000-0x00000002b7ffffff 128M online yes 86
0x00000002c0000000-0x00000002c7ffffff 128M online yes 88
0x00000002d0000000-0x00000002d7ffffff 128M online yes 90
0x00000002e0000000-0x00000002e7ffffff 128M online yes 92
0x00000002f0000000-0x00000002f7ffffff 128M online yes 94
0x0000000300000000-0x0000000307ffffff 128M online yes 96
0x0000000310000000-0x0000000317ffffff 128M online yes 98
0x0000000320000000-0x0000000327ffffff 128M online yes 100
0x0000000330000000-0x000000033fffffff 256M online yes 102-103
Memory block size: 128M
Total online memory: 8.1G
Total offline memory: 0B
With this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
Memory block size: 128M
Total online memory: 4G
Total offline memory: 0B
All memory can get unplugged, all memory block can get removed. Of
course, no workload ran and the system was basically idle, but it
highlights the issue - the fairly deterministic chain of unmovable
allocations. When a huge page for the 2MB memmap is needed, a
just-onlined 4MB page will be split. The remaining 2MB page will be used
for the memmap of the next memory block. So one memory block will hold
the memmap of the two following memory blocks. Finally the pages of the
last-onlined memory block will get used for the next bigger allocations -
if any allocation is unmovable, all dependent memory blocks cannot get
unplugged and removed until that allocation is gone.
Note that with bigger memory blocks (e.g., 256MB), *all* memory
blocks are dependent and none can get unplugged again!
b) Experiment with memory intensive workload
I performed an experiment with an older version of this patch set (before
we used undo_isolate_page_range() in online_pages(): Hotplug 56GB to a VM
with an initial 4GB, onlining all memory to ZONE_NORMAL right from the
kernel when adding it. I then run various memory intensive workloads that
consume most system memory for a total of 45 minutes. Once finished, I
try to unplug as much memory as possible.
With this change, I am able to remove via virtio-mem (adding individual
128MB memory blocks) 413 out of 448 added memory blocks. Via individual
(256MB) DIMMs 380 out of 448 added memory blocks. (I don't have any
numbers without this patchset, but looking at the above example, it's at
most half of the 448 memory blocks for virtio-mem, and most probably none
for DIMMs).
Again, there are workloads that might behave very differently due to the
nature of ZONE_NORMAL.
This change also affects (besides memory onlining):
- Other users of undo_isolate_page_range(): Pages are always placed to the
tail.
-- When memory offlining fails
-- When memory isolation fails after having isolated some pageblocks
-- When alloc_contig_range() either succeeds or fails
- Other users of __putback_isolated_page(): Pages are always placed to the
tail.
-- Free page reporting
- Other users of __free_pages_core()
-- AFAIKs, any memory that is getting exposed to the buddy during boot.
IIUC we will now usually allocate memory from lower addresses within
a zone first (especially during boot).
- Other users of generic_online_page()
-- Hyper-V balloon
This patch (of 5):
Let's prepare for additional flags and avoid long parameter lists of
bools. Follow-up patches will also make use of the flags in
__free_pages_ok().
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-1-david@redhat.com
Link: https://lkml.kernel.org/r/20201005121534.15649-2-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:20 +00:00
|
|
|
int migratetype, fpi_t fpi_flags)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2020-04-07 03:04:45 +00:00
|
|
|
struct capture_control *capc = task_capc(zone);
|
treewide: Remove uninitialized_var() usage
Using uninitialized_var() is dangerous as it papers over real bugs[1]
(or can in the future), and suppresses unrelated compiler warnings
(e.g. "unused variable"). If the compiler thinks it is uninitialized,
either simply initialize the variable or make compiler changes.
In preparation for removing[2] the[3] macro[4], remove all remaining
needless uses with the following script:
git grep '\buninitialized_var\b' | cut -d: -f1 | sort -u | \
xargs perl -pi -e \
's/\buninitialized_var\(([^\)]+)\)/\1/g;
s:\s*/\* (GCC be quiet|to make compiler happy) \*/$::g;'
drivers/video/fbdev/riva/riva_hw.c was manually tweaked to avoid
pathological white-space.
No outstanding warnings were found building allmodconfig with GCC 9.3.0
for x86_64, i386, arm64, arm, powerpc, powerpc64le, s390x, mips, sparc64,
alpha, and m68k.
[1] https://lore.kernel.org/lkml/20200603174714.192027-1-glider@google.com/
[2] https://lore.kernel.org/lkml/CA+55aFw+Vbj0i=1TGqCR5vQkCzWJ0QxK6CernOU6eedsudAixw@mail.gmail.com/
[3] https://lore.kernel.org/lkml/CA+55aFwgbgqhbp1fkxvRKEpzyR5J8n1vKT1VZdz9knmPuXhOeg@mail.gmail.com/
[4] https://lore.kernel.org/lkml/CA+55aFz2500WfbKXAx8s67wrm9=yVJu65TpLgN_ybYNv0VEOKA@mail.gmail.com/
Reviewed-by: Leon Romanovsky <leonro@mellanox.com> # drivers/infiniband and mlx4/mlx5
Acked-by: Jason Gunthorpe <jgg@mellanox.com> # IB
Acked-by: Kalle Valo <kvalo@codeaurora.org> # wireless drivers
Reviewed-by: Chao Yu <yuchao0@huawei.com> # erofs
Signed-off-by: Kees Cook <keescook@chromium.org>
2020-06-03 20:09:38 +00:00
|
|
|
unsigned long buddy_pfn;
|
2020-04-07 03:04:45 +00:00
|
|
|
unsigned long combined_pfn;
|
mm/page_alloc: prevent merging between isolated and other pageblocks
Hanjun Guo has reported that a CMA stress test causes broken accounting of
CMA and free pages:
> Before the test, I got:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 195044 kB
>
>
> After running the test:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 6602584 kB
>
> So the freed CMA memory is more than total..
>
> Also the the MemFree is more than mem total:
>
> -bash-4.3# cat /proc/meminfo
> MemTotal: 16342016 kB
> MemFree: 22367268 kB
> MemAvailable: 22370528 kB
Laura Abbott has confirmed the issue and suspected the freepage accounting
rewrite around 3.18/4.0 by Joonsoo Kim. Joonsoo had a theory that this is
caused by unexpected merging between MIGRATE_ISOLATE and MIGRATE_CMA
pageblocks:
> CMA isolates MAX_ORDER aligned blocks, but, during the process,
> partialy isolated block exists. If MAX_ORDER is 11 and
> pageblock_order is 9, two pageblocks make up MAX_ORDER
> aligned block and I can think following scenario because pageblock
> (un)isolation would be done one by one.
>
> (each character means one pageblock. 'C', 'I' means MIGRATE_CMA,
> MIGRATE_ISOLATE, respectively.
>
> CC -> IC -> II (Isolation)
> II -> CI -> CC (Un-isolation)
>
> If some pages are freed at this intermediate state such as IC or CI,
> that page could be merged to the other page that is resident on
> different type of pageblock and it will cause wrong freepage count.
This was supposed to be prevented by CMA operating on MAX_ORDER blocks,
but since it doesn't hold the zone->lock between pageblocks, a race
window does exist.
It's also likely that unexpected merging can occur between
MIGRATE_ISOLATE and non-CMA pageblocks. This should be prevented in
__free_one_page() since commit 3c605096d315 ("mm/page_alloc: restrict
max order of merging on isolated pageblock"). However, we only check
the migratetype of the pageblock where buddy merging has been initiated,
not the migratetype of the buddy pageblock (or group of pageblocks)
which can be MIGRATE_ISOLATE.
Joonsoo has suggested checking for buddy migratetype as part of
page_is_buddy(), but that would add extra checks in allocator hotpath
and bloat-o-meter has shown significant code bloat (the function is
inline).
This patch reduces the bloat at some expense of more complicated code.
The buddy-merging while-loop in __free_one_page() is initially bounded
to pageblock_border and without any migratetype checks. The checks are
placed outside, bumping the max_order if merging is allowed, and
returning to the while-loop with a statement which can't be possibly
considered harmful.
This fixes the accounting bug and also removes the arguably weird state
in the original commit 3c605096d315 where buddies could be left
unmerged.
Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
Link: https://lkml.org/lkml/2016/3/2/280
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Debugged-by: Laura Abbott <labbott@redhat.com>
Debugged-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org> [3.18+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-25 21:21:50 +00:00
|
|
|
unsigned int max_order;
|
2020-04-07 03:04:45 +00:00
|
|
|
struct page *buddy;
|
|
|
|
bool to_tail;
|
mm/page_alloc: prevent merging between isolated and other pageblocks
Hanjun Guo has reported that a CMA stress test causes broken accounting of
CMA and free pages:
> Before the test, I got:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 195044 kB
>
>
> After running the test:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 6602584 kB
>
> So the freed CMA memory is more than total..
>
> Also the the MemFree is more than mem total:
>
> -bash-4.3# cat /proc/meminfo
> MemTotal: 16342016 kB
> MemFree: 22367268 kB
> MemAvailable: 22370528 kB
Laura Abbott has confirmed the issue and suspected the freepage accounting
rewrite around 3.18/4.0 by Joonsoo Kim. Joonsoo had a theory that this is
caused by unexpected merging between MIGRATE_ISOLATE and MIGRATE_CMA
pageblocks:
> CMA isolates MAX_ORDER aligned blocks, but, during the process,
> partialy isolated block exists. If MAX_ORDER is 11 and
> pageblock_order is 9, two pageblocks make up MAX_ORDER
> aligned block and I can think following scenario because pageblock
> (un)isolation would be done one by one.
>
> (each character means one pageblock. 'C', 'I' means MIGRATE_CMA,
> MIGRATE_ISOLATE, respectively.
>
> CC -> IC -> II (Isolation)
> II -> CI -> CC (Un-isolation)
>
> If some pages are freed at this intermediate state such as IC or CI,
> that page could be merged to the other page that is resident on
> different type of pageblock and it will cause wrong freepage count.
This was supposed to be prevented by CMA operating on MAX_ORDER blocks,
but since it doesn't hold the zone->lock between pageblocks, a race
window does exist.
It's also likely that unexpected merging can occur between
MIGRATE_ISOLATE and non-CMA pageblocks. This should be prevented in
__free_one_page() since commit 3c605096d315 ("mm/page_alloc: restrict
max order of merging on isolated pageblock"). However, we only check
the migratetype of the pageblock where buddy merging has been initiated,
not the migratetype of the buddy pageblock (or group of pageblocks)
which can be MIGRATE_ISOLATE.
Joonsoo has suggested checking for buddy migratetype as part of
page_is_buddy(), but that would add extra checks in allocator hotpath
and bloat-o-meter has shown significant code bloat (the function is
inline).
This patch reduces the bloat at some expense of more complicated code.
The buddy-merging while-loop in __free_one_page() is initially bounded
to pageblock_border and without any migratetype checks. The checks are
placed outside, bumping the max_order if merging is allowed, and
returning to the while-loop with a statement which can't be possibly
considered harmful.
This fixes the accounting bug and also removes the arguably weird state
in the original commit 3c605096d315 where buddies could be left
unmerged.
Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
Link: https://lkml.org/lkml/2016/3/2/280
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Debugged-by: Laura Abbott <labbott@redhat.com>
Debugged-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org> [3.18+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-25 21:21:50 +00:00
|
|
|
|
2020-12-15 03:11:25 +00:00
|
|
|
max_order = min_t(unsigned int, MAX_ORDER - 1, pageblock_order);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-02-23 00:35:25 +00:00
|
|
|
VM_BUG_ON(!zone_is_initialized(zone));
|
2015-02-11 23:25:50 +00:00
|
|
|
VM_BUG_ON_PAGE(page->flags & PAGE_FLAGS_CHECK_AT_PREP, page);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2009-06-16 22:32:07 +00:00
|
|
|
VM_BUG_ON(migratetype == -1);
|
mm/page_alloc: prevent merging between isolated and other pageblocks
Hanjun Guo has reported that a CMA stress test causes broken accounting of
CMA and free pages:
> Before the test, I got:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 195044 kB
>
>
> After running the test:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 6602584 kB
>
> So the freed CMA memory is more than total..
>
> Also the the MemFree is more than mem total:
>
> -bash-4.3# cat /proc/meminfo
> MemTotal: 16342016 kB
> MemFree: 22367268 kB
> MemAvailable: 22370528 kB
Laura Abbott has confirmed the issue and suspected the freepage accounting
rewrite around 3.18/4.0 by Joonsoo Kim. Joonsoo had a theory that this is
caused by unexpected merging between MIGRATE_ISOLATE and MIGRATE_CMA
pageblocks:
> CMA isolates MAX_ORDER aligned blocks, but, during the process,
> partialy isolated block exists. If MAX_ORDER is 11 and
> pageblock_order is 9, two pageblocks make up MAX_ORDER
> aligned block and I can think following scenario because pageblock
> (un)isolation would be done one by one.
>
> (each character means one pageblock. 'C', 'I' means MIGRATE_CMA,
> MIGRATE_ISOLATE, respectively.
>
> CC -> IC -> II (Isolation)
> II -> CI -> CC (Un-isolation)
>
> If some pages are freed at this intermediate state such as IC or CI,
> that page could be merged to the other page that is resident on
> different type of pageblock and it will cause wrong freepage count.
This was supposed to be prevented by CMA operating on MAX_ORDER blocks,
but since it doesn't hold the zone->lock between pageblocks, a race
window does exist.
It's also likely that unexpected merging can occur between
MIGRATE_ISOLATE and non-CMA pageblocks. This should be prevented in
__free_one_page() since commit 3c605096d315 ("mm/page_alloc: restrict
max order of merging on isolated pageblock"). However, we only check
the migratetype of the pageblock where buddy merging has been initiated,
not the migratetype of the buddy pageblock (or group of pageblocks)
which can be MIGRATE_ISOLATE.
Joonsoo has suggested checking for buddy migratetype as part of
page_is_buddy(), but that would add extra checks in allocator hotpath
and bloat-o-meter has shown significant code bloat (the function is
inline).
This patch reduces the bloat at some expense of more complicated code.
The buddy-merging while-loop in __free_one_page() is initially bounded
to pageblock_border and without any migratetype checks. The checks are
placed outside, bumping the max_order if merging is allowed, and
returning to the while-loop with a statement which can't be possibly
considered harmful.
This fixes the accounting bug and also removes the arguably weird state
in the original commit 3c605096d315 where buddies could be left
unmerged.
Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
Link: https://lkml.org/lkml/2016/3/2/280
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Debugged-by: Laura Abbott <labbott@redhat.com>
Debugged-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org> [3.18+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-25 21:21:50 +00:00
|
|
|
if (likely(!is_migrate_isolate(migratetype)))
|
2014-11-13 23:19:18 +00:00
|
|
|
__mod_zone_freepage_state(zone, 1 << order, migratetype);
|
2009-06-16 22:32:07 +00:00
|
|
|
|
2017-02-22 23:41:48 +00:00
|
|
|
VM_BUG_ON_PAGE(pfn & ((1 << order) - 1), page);
|
2014-01-23 23:52:54 +00:00
|
|
|
VM_BUG_ON_PAGE(bad_range(zone, page), page);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
mm/page_alloc: prevent merging between isolated and other pageblocks
Hanjun Guo has reported that a CMA stress test causes broken accounting of
CMA and free pages:
> Before the test, I got:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 195044 kB
>
>
> After running the test:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 6602584 kB
>
> So the freed CMA memory is more than total..
>
> Also the the MemFree is more than mem total:
>
> -bash-4.3# cat /proc/meminfo
> MemTotal: 16342016 kB
> MemFree: 22367268 kB
> MemAvailable: 22370528 kB
Laura Abbott has confirmed the issue and suspected the freepage accounting
rewrite around 3.18/4.0 by Joonsoo Kim. Joonsoo had a theory that this is
caused by unexpected merging between MIGRATE_ISOLATE and MIGRATE_CMA
pageblocks:
> CMA isolates MAX_ORDER aligned blocks, but, during the process,
> partialy isolated block exists. If MAX_ORDER is 11 and
> pageblock_order is 9, two pageblocks make up MAX_ORDER
> aligned block and I can think following scenario because pageblock
> (un)isolation would be done one by one.
>
> (each character means one pageblock. 'C', 'I' means MIGRATE_CMA,
> MIGRATE_ISOLATE, respectively.
>
> CC -> IC -> II (Isolation)
> II -> CI -> CC (Un-isolation)
>
> If some pages are freed at this intermediate state such as IC or CI,
> that page could be merged to the other page that is resident on
> different type of pageblock and it will cause wrong freepage count.
This was supposed to be prevented by CMA operating on MAX_ORDER blocks,
but since it doesn't hold the zone->lock between pageblocks, a race
window does exist.
It's also likely that unexpected merging can occur between
MIGRATE_ISOLATE and non-CMA pageblocks. This should be prevented in
__free_one_page() since commit 3c605096d315 ("mm/page_alloc: restrict
max order of merging on isolated pageblock"). However, we only check
the migratetype of the pageblock where buddy merging has been initiated,
not the migratetype of the buddy pageblock (or group of pageblocks)
which can be MIGRATE_ISOLATE.
Joonsoo has suggested checking for buddy migratetype as part of
page_is_buddy(), but that would add extra checks in allocator hotpath
and bloat-o-meter has shown significant code bloat (the function is
inline).
This patch reduces the bloat at some expense of more complicated code.
The buddy-merging while-loop in __free_one_page() is initially bounded
to pageblock_border and without any migratetype checks. The checks are
placed outside, bumping the max_order if merging is allowed, and
returning to the while-loop with a statement which can't be possibly
considered harmful.
This fixes the accounting bug and also removes the arguably weird state
in the original commit 3c605096d315 where buddies could be left
unmerged.
Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
Link: https://lkml.org/lkml/2016/3/2/280
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Debugged-by: Laura Abbott <labbott@redhat.com>
Debugged-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org> [3.18+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-25 21:21:50 +00:00
|
|
|
continue_merging:
|
2020-12-15 03:11:25 +00:00
|
|
|
while (order < max_order) {
|
2019-03-05 23:45:41 +00:00
|
|
|
if (compaction_capture(capc, page, order, migratetype)) {
|
|
|
|
__mod_zone_freepage_state(zone, -(1 << order),
|
|
|
|
migratetype);
|
|
|
|
return;
|
|
|
|
}
|
2017-02-22 23:41:48 +00:00
|
|
|
buddy_pfn = __find_buddy_pfn(pfn, order);
|
|
|
|
buddy = page + (buddy_pfn - pfn);
|
2017-02-22 23:41:51 +00:00
|
|
|
|
|
|
|
if (!pfn_valid_within(buddy_pfn))
|
|
|
|
goto done_merging;
|
2006-06-23 09:03:01 +00:00
|
|
|
if (!page_is_buddy(page, buddy, order))
|
mm/page_alloc: prevent merging between isolated and other pageblocks
Hanjun Guo has reported that a CMA stress test causes broken accounting of
CMA and free pages:
> Before the test, I got:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 195044 kB
>
>
> After running the test:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 6602584 kB
>
> So the freed CMA memory is more than total..
>
> Also the the MemFree is more than mem total:
>
> -bash-4.3# cat /proc/meminfo
> MemTotal: 16342016 kB
> MemFree: 22367268 kB
> MemAvailable: 22370528 kB
Laura Abbott has confirmed the issue and suspected the freepage accounting
rewrite around 3.18/4.0 by Joonsoo Kim. Joonsoo had a theory that this is
caused by unexpected merging between MIGRATE_ISOLATE and MIGRATE_CMA
pageblocks:
> CMA isolates MAX_ORDER aligned blocks, but, during the process,
> partialy isolated block exists. If MAX_ORDER is 11 and
> pageblock_order is 9, two pageblocks make up MAX_ORDER
> aligned block and I can think following scenario because pageblock
> (un)isolation would be done one by one.
>
> (each character means one pageblock. 'C', 'I' means MIGRATE_CMA,
> MIGRATE_ISOLATE, respectively.
>
> CC -> IC -> II (Isolation)
> II -> CI -> CC (Un-isolation)
>
> If some pages are freed at this intermediate state such as IC or CI,
> that page could be merged to the other page that is resident on
> different type of pageblock and it will cause wrong freepage count.
This was supposed to be prevented by CMA operating on MAX_ORDER blocks,
but since it doesn't hold the zone->lock between pageblocks, a race
window does exist.
It's also likely that unexpected merging can occur between
MIGRATE_ISOLATE and non-CMA pageblocks. This should be prevented in
__free_one_page() since commit 3c605096d315 ("mm/page_alloc: restrict
max order of merging on isolated pageblock"). However, we only check
the migratetype of the pageblock where buddy merging has been initiated,
not the migratetype of the buddy pageblock (or group of pageblocks)
which can be MIGRATE_ISOLATE.
Joonsoo has suggested checking for buddy migratetype as part of
page_is_buddy(), but that would add extra checks in allocator hotpath
and bloat-o-meter has shown significant code bloat (the function is
inline).
This patch reduces the bloat at some expense of more complicated code.
The buddy-merging while-loop in __free_one_page() is initially bounded
to pageblock_border and without any migratetype checks. The checks are
placed outside, bumping the max_order if merging is allowed, and
returning to the while-loop with a statement which can't be possibly
considered harmful.
This fixes the accounting bug and also removes the arguably weird state
in the original commit 3c605096d315 where buddies could be left
unmerged.
Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
Link: https://lkml.org/lkml/2016/3/2/280
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Debugged-by: Laura Abbott <labbott@redhat.com>
Debugged-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org> [3.18+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-25 21:21:50 +00:00
|
|
|
goto done_merging;
|
2012-01-10 23:07:28 +00:00
|
|
|
/*
|
|
|
|
* Our buddy is free or it is CONFIG_DEBUG_PAGEALLOC guard page,
|
|
|
|
* merge with it and move up one order.
|
|
|
|
*/
|
2019-05-14 22:41:32 +00:00
|
|
|
if (page_is_guard(buddy))
|
2014-12-13 00:55:01 +00:00
|
|
|
clear_page_guard(zone, buddy, order, migratetype);
|
2019-05-14 22:41:32 +00:00
|
|
|
else
|
2020-04-07 03:04:49 +00:00
|
|
|
del_page_from_free_list(buddy, zone, order);
|
2017-02-22 23:41:48 +00:00
|
|
|
combined_pfn = buddy_pfn & pfn;
|
|
|
|
page = page + (combined_pfn - pfn);
|
|
|
|
pfn = combined_pfn;
|
2005-04-16 22:20:36 +00:00
|
|
|
order++;
|
|
|
|
}
|
2020-12-15 03:11:25 +00:00
|
|
|
if (order < MAX_ORDER - 1) {
|
mm/page_alloc: prevent merging between isolated and other pageblocks
Hanjun Guo has reported that a CMA stress test causes broken accounting of
CMA and free pages:
> Before the test, I got:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 195044 kB
>
>
> After running the test:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 6602584 kB
>
> So the freed CMA memory is more than total..
>
> Also the the MemFree is more than mem total:
>
> -bash-4.3# cat /proc/meminfo
> MemTotal: 16342016 kB
> MemFree: 22367268 kB
> MemAvailable: 22370528 kB
Laura Abbott has confirmed the issue and suspected the freepage accounting
rewrite around 3.18/4.0 by Joonsoo Kim. Joonsoo had a theory that this is
caused by unexpected merging between MIGRATE_ISOLATE and MIGRATE_CMA
pageblocks:
> CMA isolates MAX_ORDER aligned blocks, but, during the process,
> partialy isolated block exists. If MAX_ORDER is 11 and
> pageblock_order is 9, two pageblocks make up MAX_ORDER
> aligned block and I can think following scenario because pageblock
> (un)isolation would be done one by one.
>
> (each character means one pageblock. 'C', 'I' means MIGRATE_CMA,
> MIGRATE_ISOLATE, respectively.
>
> CC -> IC -> II (Isolation)
> II -> CI -> CC (Un-isolation)
>
> If some pages are freed at this intermediate state such as IC or CI,
> that page could be merged to the other page that is resident on
> different type of pageblock and it will cause wrong freepage count.
This was supposed to be prevented by CMA operating on MAX_ORDER blocks,
but since it doesn't hold the zone->lock between pageblocks, a race
window does exist.
It's also likely that unexpected merging can occur between
MIGRATE_ISOLATE and non-CMA pageblocks. This should be prevented in
__free_one_page() since commit 3c605096d315 ("mm/page_alloc: restrict
max order of merging on isolated pageblock"). However, we only check
the migratetype of the pageblock where buddy merging has been initiated,
not the migratetype of the buddy pageblock (or group of pageblocks)
which can be MIGRATE_ISOLATE.
Joonsoo has suggested checking for buddy migratetype as part of
page_is_buddy(), but that would add extra checks in allocator hotpath
and bloat-o-meter has shown significant code bloat (the function is
inline).
This patch reduces the bloat at some expense of more complicated code.
The buddy-merging while-loop in __free_one_page() is initially bounded
to pageblock_border and without any migratetype checks. The checks are
placed outside, bumping the max_order if merging is allowed, and
returning to the while-loop with a statement which can't be possibly
considered harmful.
This fixes the accounting bug and also removes the arguably weird state
in the original commit 3c605096d315 where buddies could be left
unmerged.
Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
Link: https://lkml.org/lkml/2016/3/2/280
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Debugged-by: Laura Abbott <labbott@redhat.com>
Debugged-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org> [3.18+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-25 21:21:50 +00:00
|
|
|
/* If we are here, it means order is >= pageblock_order.
|
|
|
|
* We want to prevent merge between freepages on isolate
|
|
|
|
* pageblock and normal pageblock. Without this, pageblock
|
|
|
|
* isolation could cause incorrect freepage or CMA accounting.
|
|
|
|
*
|
|
|
|
* We don't want to hit this code for the more frequent
|
|
|
|
* low-order merging.
|
|
|
|
*/
|
|
|
|
if (unlikely(has_isolate_pageblock(zone))) {
|
|
|
|
int buddy_mt;
|
|
|
|
|
2017-02-22 23:41:48 +00:00
|
|
|
buddy_pfn = __find_buddy_pfn(pfn, order);
|
|
|
|
buddy = page + (buddy_pfn - pfn);
|
mm/page_alloc: prevent merging between isolated and other pageblocks
Hanjun Guo has reported that a CMA stress test causes broken accounting of
CMA and free pages:
> Before the test, I got:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 195044 kB
>
>
> After running the test:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 6602584 kB
>
> So the freed CMA memory is more than total..
>
> Also the the MemFree is more than mem total:
>
> -bash-4.3# cat /proc/meminfo
> MemTotal: 16342016 kB
> MemFree: 22367268 kB
> MemAvailable: 22370528 kB
Laura Abbott has confirmed the issue and suspected the freepage accounting
rewrite around 3.18/4.0 by Joonsoo Kim. Joonsoo had a theory that this is
caused by unexpected merging between MIGRATE_ISOLATE and MIGRATE_CMA
pageblocks:
> CMA isolates MAX_ORDER aligned blocks, but, during the process,
> partialy isolated block exists. If MAX_ORDER is 11 and
> pageblock_order is 9, two pageblocks make up MAX_ORDER
> aligned block and I can think following scenario because pageblock
> (un)isolation would be done one by one.
>
> (each character means one pageblock. 'C', 'I' means MIGRATE_CMA,
> MIGRATE_ISOLATE, respectively.
>
> CC -> IC -> II (Isolation)
> II -> CI -> CC (Un-isolation)
>
> If some pages are freed at this intermediate state such as IC or CI,
> that page could be merged to the other page that is resident on
> different type of pageblock and it will cause wrong freepage count.
This was supposed to be prevented by CMA operating on MAX_ORDER blocks,
but since it doesn't hold the zone->lock between pageblocks, a race
window does exist.
It's also likely that unexpected merging can occur between
MIGRATE_ISOLATE and non-CMA pageblocks. This should be prevented in
__free_one_page() since commit 3c605096d315 ("mm/page_alloc: restrict
max order of merging on isolated pageblock"). However, we only check
the migratetype of the pageblock where buddy merging has been initiated,
not the migratetype of the buddy pageblock (or group of pageblocks)
which can be MIGRATE_ISOLATE.
Joonsoo has suggested checking for buddy migratetype as part of
page_is_buddy(), but that would add extra checks in allocator hotpath
and bloat-o-meter has shown significant code bloat (the function is
inline).
This patch reduces the bloat at some expense of more complicated code.
The buddy-merging while-loop in __free_one_page() is initially bounded
to pageblock_border and without any migratetype checks. The checks are
placed outside, bumping the max_order if merging is allowed, and
returning to the while-loop with a statement which can't be possibly
considered harmful.
This fixes the accounting bug and also removes the arguably weird state
in the original commit 3c605096d315 where buddies could be left
unmerged.
Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
Link: https://lkml.org/lkml/2016/3/2/280
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Debugged-by: Laura Abbott <labbott@redhat.com>
Debugged-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org> [3.18+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-25 21:21:50 +00:00
|
|
|
buddy_mt = get_pageblock_migratetype(buddy);
|
|
|
|
|
|
|
|
if (migratetype != buddy_mt
|
|
|
|
&& (is_migrate_isolate(migratetype) ||
|
|
|
|
is_migrate_isolate(buddy_mt)))
|
|
|
|
goto done_merging;
|
|
|
|
}
|
2020-12-15 03:11:25 +00:00
|
|
|
max_order = order + 1;
|
mm/page_alloc: prevent merging between isolated and other pageblocks
Hanjun Guo has reported that a CMA stress test causes broken accounting of
CMA and free pages:
> Before the test, I got:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 195044 kB
>
>
> After running the test:
> -bash-4.3# cat /proc/meminfo | grep Cma
> CmaTotal: 204800 kB
> CmaFree: 6602584 kB
>
> So the freed CMA memory is more than total..
>
> Also the the MemFree is more than mem total:
>
> -bash-4.3# cat /proc/meminfo
> MemTotal: 16342016 kB
> MemFree: 22367268 kB
> MemAvailable: 22370528 kB
Laura Abbott has confirmed the issue and suspected the freepage accounting
rewrite around 3.18/4.0 by Joonsoo Kim. Joonsoo had a theory that this is
caused by unexpected merging between MIGRATE_ISOLATE and MIGRATE_CMA
pageblocks:
> CMA isolates MAX_ORDER aligned blocks, but, during the process,
> partialy isolated block exists. If MAX_ORDER is 11 and
> pageblock_order is 9, two pageblocks make up MAX_ORDER
> aligned block and I can think following scenario because pageblock
> (un)isolation would be done one by one.
>
> (each character means one pageblock. 'C', 'I' means MIGRATE_CMA,
> MIGRATE_ISOLATE, respectively.
>
> CC -> IC -> II (Isolation)
> II -> CI -> CC (Un-isolation)
>
> If some pages are freed at this intermediate state such as IC or CI,
> that page could be merged to the other page that is resident on
> different type of pageblock and it will cause wrong freepage count.
This was supposed to be prevented by CMA operating on MAX_ORDER blocks,
but since it doesn't hold the zone->lock between pageblocks, a race
window does exist.
It's also likely that unexpected merging can occur between
MIGRATE_ISOLATE and non-CMA pageblocks. This should be prevented in
__free_one_page() since commit 3c605096d315 ("mm/page_alloc: restrict
max order of merging on isolated pageblock"). However, we only check
the migratetype of the pageblock where buddy merging has been initiated,
not the migratetype of the buddy pageblock (or group of pageblocks)
which can be MIGRATE_ISOLATE.
Joonsoo has suggested checking for buddy migratetype as part of
page_is_buddy(), but that would add extra checks in allocator hotpath
and bloat-o-meter has shown significant code bloat (the function is
inline).
This patch reduces the bloat at some expense of more complicated code.
The buddy-merging while-loop in __free_one_page() is initially bounded
to pageblock_border and without any migratetype checks. The checks are
placed outside, bumping the max_order if merging is allowed, and
returning to the while-loop with a statement which can't be possibly
considered harmful.
This fixes the accounting bug and also removes the arguably weird state
in the original commit 3c605096d315 where buddies could be left
unmerged.
Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
Link: https://lkml.org/lkml/2016/3/2/280
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Debugged-by: Laura Abbott <labbott@redhat.com>
Debugged-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org> [3.18+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-03-25 21:21:50 +00:00
|
|
|
goto continue_merging;
|
|
|
|
}
|
|
|
|
|
|
|
|
done_merging:
|
2020-10-16 03:10:15 +00:00
|
|
|
set_buddy_order(page, order);
|
page allocator: reduce fragmentation in buddy allocator by adding buddies that are merging to the tail of the free lists
In order to reduce fragmentation, this patch classifies freed pages in two
groups according to their probability of being part of a high order merge.
Pages belonging to a compound whose next-highest buddy is free are more
likely to be part of a high order merge in the near future, so they will
be added at the tail of the freelist. The remaining pages are put at the
front of the freelist.
In this way, the pages that are more likely to cause a big merge are kept
free longer. Consequently there is a tendency to aggregate the
long-living allocations on a subset of the compounds, reducing the
fragmentation.
This heuristic was tested on three machines, x86, x86-64 and ppc64 with
3GB of RAM in each machine. The tests were kernbench, netperf, sysbench
and STREAM for performance and a high-order stress test for huge page
allocations.
KernBench X86
Elapsed mean 374.77 ( 0.00%) 375.10 (-0.09%)
User mean 649.53 ( 0.00%) 650.44 (-0.14%)
System mean 54.75 ( 0.00%) 54.18 ( 1.05%)
CPU mean 187.75 ( 0.00%) 187.25 ( 0.27%)
KernBench X86-64
Elapsed mean 94.45 ( 0.00%) 94.01 ( 0.47%)
User mean 323.27 ( 0.00%) 322.66 ( 0.19%)
System mean 36.71 ( 0.00%) 36.50 ( 0.57%)
CPU mean 380.75 ( 0.00%) 381.75 (-0.26%)
KernBench PPC64
Elapsed mean 173.45 ( 0.00%) 173.74 (-0.17%)
User mean 587.99 ( 0.00%) 587.95 ( 0.01%)
System mean 60.60 ( 0.00%) 60.57 ( 0.05%)
CPU mean 373.50 ( 0.00%) 372.75 ( 0.20%)
Nothing notable for kernbench.
NetPerf UDP X86
64 42.68 ( 0.00%) 42.77 ( 0.21%)
128 85.62 ( 0.00%) 85.32 (-0.35%)
256 170.01 ( 0.00%) 168.76 (-0.74%)
1024 655.68 ( 0.00%) 652.33 (-0.51%)
2048 1262.39 ( 0.00%) 1248.61 (-1.10%)
3312 1958.41 ( 0.00%) 1944.61 (-0.71%)
4096 2345.63 ( 0.00%) 2318.83 (-1.16%)
8192 4132.90 ( 0.00%) 4089.50 (-1.06%)
16384 6770.88 ( 0.00%) 6642.05 (-1.94%)*
NetPerf UDP X86-64
64 148.82 ( 0.00%) 154.92 ( 3.94%)
128 298.96 ( 0.00%) 312.95 ( 4.47%)
256 583.67 ( 0.00%) 626.39 ( 6.82%)
1024 2293.18 ( 0.00%) 2371.10 ( 3.29%)
2048 4274.16 ( 0.00%) 4396.83 ( 2.79%)
3312 6356.94 ( 0.00%) 6571.35 ( 3.26%)
4096 7422.68 ( 0.00%) 7635.42 ( 2.79%)*
8192 12114.81 ( 0.00%)* 12346.88 ( 1.88%)
16384 17022.28 ( 0.00%)* 17033.19 ( 0.06%)*
1.64% 2.73%
NetPerf UDP PPC64
64 49.98 ( 0.00%) 50.25 ( 0.54%)
128 98.66 ( 0.00%) 100.95 ( 2.27%)
256 197.33 ( 0.00%) 191.03 (-3.30%)
1024 761.98 ( 0.00%) 785.07 ( 2.94%)
2048 1493.50 ( 0.00%) 1510.85 ( 1.15%)
3312 2303.95 ( 0.00%) 2271.72 (-1.42%)
4096 2774.56 ( 0.00%) 2773.06 (-0.05%)
8192 4918.31 ( 0.00%) 4793.59 (-2.60%)
16384 7497.98 ( 0.00%) 7749.52 ( 3.25%)
The tests are run to have confidence limits within 1%. Results marked
with a * were not confident although in this case, it's only outside by
small amounts. Even with some results that were not confident, the
netperf UDP results were generally positive.
NetPerf TCP X86
64 652.25 ( 0.00%)* 648.12 (-0.64%)*
23.80% 22.82%
128 1229.98 ( 0.00%)* 1220.56 (-0.77%)*
21.03% 18.90%
256 2105.88 ( 0.00%) 1872.03 (-12.49%)*
1.00% 16.46%
1024 3476.46 ( 0.00%)* 3548.28 ( 2.02%)*
13.37% 11.39%
2048 4023.44 ( 0.00%)* 4231.45 ( 4.92%)*
9.76% 12.48%
3312 4348.88 ( 0.00%)* 4396.96 ( 1.09%)*
6.49% 8.75%
4096 4726.56 ( 0.00%)* 4877.71 ( 3.10%)*
9.85% 8.50%
8192 4732.28 ( 0.00%)* 5777.77 (18.10%)*
9.13% 13.04%
16384 5543.05 ( 0.00%)* 5906.24 ( 6.15%)*
7.73% 8.68%
NETPERF TCP X86-64
netperf-tcp-vanilla-netperf netperf-tcp
tcp-vanilla pgalloc-delay
64 1895.87 ( 0.00%)* 1775.07 (-6.81%)*
5.79% 4.78%
128 3571.03 ( 0.00%)* 3342.20 (-6.85%)*
3.68% 6.06%
256 5097.21 ( 0.00%)* 4859.43 (-4.89%)*
3.02% 2.10%
1024 8919.10 ( 0.00%)* 8892.49 (-0.30%)*
5.89% 6.55%
2048 10255.46 ( 0.00%)* 10449.39 ( 1.86%)*
7.08% 7.44%
3312 10839.90 ( 0.00%)* 10740.15 (-0.93%)*
6.87% 7.33%
4096 10814.84 ( 0.00%)* 10766.97 (-0.44%)*
6.86% 8.18%
8192 11606.89 ( 0.00%)* 11189.28 (-3.73%)*
7.49% 5.55%
16384 12554.88 ( 0.00%)* 12361.22 (-1.57%)*
7.36% 6.49%
NETPERF TCP PPC64
netperf-tcp-vanilla-netperf netperf-tcp
tcp-vanilla pgalloc-delay
64 594.17 ( 0.00%) 596.04 ( 0.31%)*
1.00% 2.29%
128 1064.87 ( 0.00%)* 1074.77 ( 0.92%)*
1.30% 1.40%
256 1852.46 ( 0.00%)* 1856.95 ( 0.24%)
1.25% 1.00%
1024 3839.46 ( 0.00%)* 3813.05 (-0.69%)
1.02% 1.00%
2048 4885.04 ( 0.00%)* 4881.97 (-0.06%)*
1.15% 1.04%
3312 5506.90 ( 0.00%) 5459.72 (-0.86%)
4096 6449.19 ( 0.00%) 6345.46 (-1.63%)
8192 7501.17 ( 0.00%) 7508.79 ( 0.10%)
16384 9618.65 ( 0.00%) 9490.10 (-1.35%)
There was a distinct lack of confidence in the X86* figures so I included
what the devation was where the results were not confident. Many of the
results, whether gains or losses were within the standard deviation so no
solid conclusion can be reached on performance impact. Looking at the
figures, only the X86-64 ones look suspicious with a few losses that were
outside the noise. However, the results were so unstable that without
knowing why they vary so much, a solid conclusion cannot be reached.
SYSBENCH X86
sysbench-vanilla pgalloc-delay
1 7722.85 ( 0.00%) 7756.79 ( 0.44%)
2 14901.11 ( 0.00%) 13683.44 (-8.90%)
3 15171.71 ( 0.00%) 14888.25 (-1.90%)
4 14966.98 ( 0.00%) 15029.67 ( 0.42%)
5 14370.47 ( 0.00%) 14865.00 ( 3.33%)
6 14870.33 ( 0.00%) 14845.57 (-0.17%)
7 14429.45 ( 0.00%) 14520.85 ( 0.63%)
8 14354.35 ( 0.00%) 14362.31 ( 0.06%)
SYSBENCH X86-64
1 17448.70 ( 0.00%) 17484.41 ( 0.20%)
2 34276.39 ( 0.00%) 34251.00 (-0.07%)
3 50805.25 ( 0.00%) 50854.80 ( 0.10%)
4 66667.10 ( 0.00%) 66174.69 (-0.74%)
5 66003.91 ( 0.00%) 65685.25 (-0.49%)
6 64981.90 ( 0.00%) 65125.60 ( 0.22%)
7 64933.16 ( 0.00%) 64379.23 (-0.86%)
8 63353.30 ( 0.00%) 63281.22 (-0.11%)
9 63511.84 ( 0.00%) 63570.37 ( 0.09%)
10 62708.27 ( 0.00%) 63166.25 ( 0.73%)
11 62092.81 ( 0.00%) 61787.75 (-0.49%)
12 61330.11 ( 0.00%) 61036.34 (-0.48%)
13 61438.37 ( 0.00%) 61994.47 ( 0.90%)
14 62304.48 ( 0.00%) 62064.90 (-0.39%)
15 63296.48 ( 0.00%) 62875.16 (-0.67%)
16 63951.76 ( 0.00%) 63769.09 (-0.29%)
SYSBENCH PPC64
-sysbench-pgalloc-delay-sysbench
sysbench-vanilla pgalloc-delay
1 7645.08 ( 0.00%) 7467.43 (-2.38%)
2 14856.67 ( 0.00%) 14558.73 (-2.05%)
3 21952.31 ( 0.00%) 21683.64 (-1.24%)
4 27946.09 ( 0.00%) 28623.29 ( 2.37%)
5 28045.11 ( 0.00%) 28143.69 ( 0.35%)
6 27477.10 ( 0.00%) 27337.45 (-0.51%)
7 26489.17 ( 0.00%) 26590.06 ( 0.38%)
8 26642.91 ( 0.00%) 25274.33 (-5.41%)
9 25137.27 ( 0.00%) 24810.06 (-1.32%)
10 24451.99 ( 0.00%) 24275.85 (-0.73%)
11 23262.20 ( 0.00%) 23674.88 ( 1.74%)
12 24234.81 ( 0.00%) 23640.89 (-2.51%)
13 24577.75 ( 0.00%) 24433.50 (-0.59%)
14 25640.19 ( 0.00%) 25116.52 (-2.08%)
15 26188.84 ( 0.00%) 26181.36 (-0.03%)
16 26782.37 ( 0.00%) 26255.99 (-2.00%)
Again, there is little to conclude here. While there are a few losses,
the results vary by +/- 8% in some cases. They are the results of most
concern as there are some large losses but it's also within the variance
typically seen between kernel releases.
The STREAM results varied so little and are so verbose that I didn't
include them here.
The final test stressed how many huge pages can be allocated. The
absolute number of huge pages allocated are the same with or without the
page. However, the "unusability free space index" which is a measure of
external fragmentation was slightly lower (lower is better) throughout the
lifetime of the system. I also measured the latency of how long it took
to successfully allocate a huge page. The latency was slightly lower and
on X86 and PPC64, more huge pages were allocated almost immediately from
the free lists. The improvement is slight but there.
[mel@csn.ul.ie: Tested, reworked for less branches]
[czoccolo@gmail.com: fix oops by checking pfn_valid_within()]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Acked-by: Rik van Riel <riel@redhat.com>
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Corrado Zoccolo <czoccolo@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-05-24 21:31:54 +00:00
|
|
|
|
mm/page_alloc: place pages to tail in __putback_isolated_page()
__putback_isolated_page() already documents that pages will be placed to
the tail of the freelist - this is, however, not the case for "order >=
MAX_ORDER - 2" (see buddy_merge_likely()) - which should be the case for
all existing users.
This change affects two users:
- free page reporting
- page isolation, when undoing the isolation (including memory onlining).
This behavior is desirable for pages that haven't really been touched
lately, so exactly the two users that don't actually read/write page
content, but rather move untouched pages.
The new behavior is especially desirable for memory onlining, where we
allow allocation of newly onlined pages via undo_isolate_page_range() in
online_pages(). Right now, we always place them to the head of the
freelist, resulting in undesireable behavior: Assume we add individual
memory chunks via add_memory() and online them right away to the NORMAL
zone. We create a dependency chain of unmovable allocations e.g., via the
memmap. The memmap of the next chunk will be placed onto previous chunks
- if the last block cannot get offlined+removed, all dependent ones cannot
get offlined+removed. While this can already be observed with individual
DIMMs, it's more of an issue for virtio-mem (and I suspect also ppc
DLPAR).
Document that this should only be used for optimizations, and no code
should rely on this behavior for correction (if the order of the freelists
ever changes).
We won't care about page shuffling: memory onlining already properly
shuffles after onlining. free page reporting doesn't care about
physically contiguous ranges, and there are already cases where page
isolation will simply move (physically close) free pages to (currently)
the head of the freelists via move_freepages_block() instead of shuffling.
If this becomes ever relevant, we should shuffle the whole zone when
undoing isolation of larger ranges, and after free_contig_range().
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-3-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:26 +00:00
|
|
|
if (fpi_flags & FPI_TO_TAIL)
|
|
|
|
to_tail = true;
|
|
|
|
else if (is_shuffle_order(order))
|
2020-04-07 03:04:45 +00:00
|
|
|
to_tail = shuffle_pick_tail();
|
2019-05-14 22:41:35 +00:00
|
|
|
else
|
2020-04-07 03:04:45 +00:00
|
|
|
to_tail = buddy_merge_likely(pfn, buddy_pfn, page, order);
|
2019-05-14 22:41:35 +00:00
|
|
|
|
2020-04-07 03:04:45 +00:00
|
|
|
if (to_tail)
|
2020-04-07 03:04:49 +00:00
|
|
|
add_to_free_list_tail(page, zone, order, migratetype);
|
2020-04-07 03:04:45 +00:00
|
|
|
else
|
2020-04-07 03:04:49 +00:00
|
|
|
add_to_free_list(page, zone, order, migratetype);
|
2020-04-07 03:04:56 +00:00
|
|
|
|
|
|
|
/* Notify page reporting subsystem of freed page */
|
mm/page_alloc: convert "report" flag of __free_one_page() to a proper flag
Patch series "mm: place pages to the freelist tail when onlining and undoing isolation", v2.
When adding separate memory blocks via add_memory*() and onlining them
immediately, the metadata (especially the memmap) of the next block will
be placed onto one of the just added+onlined block. This creates a chain
of unmovable allocations: If the last memory block cannot get
offlined+removed() so will all dependent ones. We directly have unmovable
allocations all over the place.
This can be observed quite easily using virtio-mem, however, it can also
be observed when using DIMMs. The freshly onlined pages will usually be
placed to the head of the freelists, meaning they will be allocated next,
turning the just-added memory usually immediately un-removable. The fresh
pages are cold, prefering to allocate others (that might be hot) also
feels to be the natural thing to do.
It also applies to the hyper-v balloon xen-balloon, and ppc64 dlpar: when
adding separate, successive memory blocks, each memory block will have
unmovable allocations on them - for example gigantic pages will fail to
allocate.
While the ZONE_NORMAL doesn't provide any guarantees that memory can get
offlined+removed again (any kind of fragmentation with unmovable
allocations is possible), there are many scenarios (hotplugging a lot of
memory, running workload, hotunplug some memory/as much as possible) where
we can offline+remove quite a lot with this patchset.
a) To visualize the problem, a very simple example:
Start a VM with 4GB and 8GB of virtio-mem memory:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000033fffffff 9G online yes 32-103
Memory block size: 128M
Total online memory: 12G
Total offline memory: 0B
Then try to unplug as much as possible using virtio-mem. Observe which
memory blocks are still around. Without this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
0x0000000148000000-0x000000014fffffff 128M online yes 41
0x0000000158000000-0x000000015fffffff 128M online yes 43
0x0000000168000000-0x000000016fffffff 128M online yes 45
0x0000000178000000-0x000000017fffffff 128M online yes 47
0x0000000188000000-0x0000000197ffffff 256M online yes 49-50
0x00000001a0000000-0x00000001a7ffffff 128M online yes 52
0x00000001b0000000-0x00000001b7ffffff 128M online yes 54
0x00000001c0000000-0x00000001c7ffffff 128M online yes 56
0x00000001d0000000-0x00000001d7ffffff 128M online yes 58
0x00000001e0000000-0x00000001e7ffffff 128M online yes 60
0x00000001f0000000-0x00000001f7ffffff 128M online yes 62
0x0000000200000000-0x0000000207ffffff 128M online yes 64
0x0000000210000000-0x0000000217ffffff 128M online yes 66
0x0000000220000000-0x0000000227ffffff 128M online yes 68
0x0000000230000000-0x0000000237ffffff 128M online yes 70
0x0000000240000000-0x0000000247ffffff 128M online yes 72
0x0000000250000000-0x0000000257ffffff 128M online yes 74
0x0000000260000000-0x0000000267ffffff 128M online yes 76
0x0000000270000000-0x0000000277ffffff 128M online yes 78
0x0000000280000000-0x0000000287ffffff 128M online yes 80
0x0000000290000000-0x0000000297ffffff 128M online yes 82
0x00000002a0000000-0x00000002a7ffffff 128M online yes 84
0x00000002b0000000-0x00000002b7ffffff 128M online yes 86
0x00000002c0000000-0x00000002c7ffffff 128M online yes 88
0x00000002d0000000-0x00000002d7ffffff 128M online yes 90
0x00000002e0000000-0x00000002e7ffffff 128M online yes 92
0x00000002f0000000-0x00000002f7ffffff 128M online yes 94
0x0000000300000000-0x0000000307ffffff 128M online yes 96
0x0000000310000000-0x0000000317ffffff 128M online yes 98
0x0000000320000000-0x0000000327ffffff 128M online yes 100
0x0000000330000000-0x000000033fffffff 256M online yes 102-103
Memory block size: 128M
Total online memory: 8.1G
Total offline memory: 0B
With this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
Memory block size: 128M
Total online memory: 4G
Total offline memory: 0B
All memory can get unplugged, all memory block can get removed. Of
course, no workload ran and the system was basically idle, but it
highlights the issue - the fairly deterministic chain of unmovable
allocations. When a huge page for the 2MB memmap is needed, a
just-onlined 4MB page will be split. The remaining 2MB page will be used
for the memmap of the next memory block. So one memory block will hold
the memmap of the two following memory blocks. Finally the pages of the
last-onlined memory block will get used for the next bigger allocations -
if any allocation is unmovable, all dependent memory blocks cannot get
unplugged and removed until that allocation is gone.
Note that with bigger memory blocks (e.g., 256MB), *all* memory
blocks are dependent and none can get unplugged again!
b) Experiment with memory intensive workload
I performed an experiment with an older version of this patch set (before
we used undo_isolate_page_range() in online_pages(): Hotplug 56GB to a VM
with an initial 4GB, onlining all memory to ZONE_NORMAL right from the
kernel when adding it. I then run various memory intensive workloads that
consume most system memory for a total of 45 minutes. Once finished, I
try to unplug as much memory as possible.
With this change, I am able to remove via virtio-mem (adding individual
128MB memory blocks) 413 out of 448 added memory blocks. Via individual
(256MB) DIMMs 380 out of 448 added memory blocks. (I don't have any
numbers without this patchset, but looking at the above example, it's at
most half of the 448 memory blocks for virtio-mem, and most probably none
for DIMMs).
Again, there are workloads that might behave very differently due to the
nature of ZONE_NORMAL.
This change also affects (besides memory onlining):
- Other users of undo_isolate_page_range(): Pages are always placed to the
tail.
-- When memory offlining fails
-- When memory isolation fails after having isolated some pageblocks
-- When alloc_contig_range() either succeeds or fails
- Other users of __putback_isolated_page(): Pages are always placed to the
tail.
-- Free page reporting
- Other users of __free_pages_core()
-- AFAIKs, any memory that is getting exposed to the buddy during boot.
IIUC we will now usually allocate memory from lower addresses within
a zone first (especially during boot).
- Other users of generic_online_page()
-- Hyper-V balloon
This patch (of 5):
Let's prepare for additional flags and avoid long parameter lists of
bools. Follow-up patches will also make use of the flags in
__free_pages_ok().
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-1-david@redhat.com
Link: https://lkml.kernel.org/r/20201005121534.15649-2-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:20 +00:00
|
|
|
if (!(fpi_flags & FPI_SKIP_REPORT_NOTIFY))
|
2020-04-07 03:04:56 +00:00
|
|
|
page_reporting_notify_free(order);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:15 +00:00
|
|
|
/*
|
|
|
|
* A bad page could be due to a number of fields. Instead of multiple branches,
|
|
|
|
* try and check multiple fields with one check. The caller must do a detailed
|
|
|
|
* check if necessary.
|
|
|
|
*/
|
|
|
|
static inline bool page_expected_state(struct page *page,
|
|
|
|
unsigned long check_flags)
|
|
|
|
{
|
|
|
|
if (unlikely(atomic_read(&page->_mapcount) != -1))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (unlikely((unsigned long)page->mapping |
|
|
|
|
page_ref_count(page) |
|
|
|
|
#ifdef CONFIG_MEMCG
|
mm: memcontrol: Use helpers to read page's memcg data
Patch series "mm: allow mapping accounted kernel pages to userspace", v6.
Currently a non-slab kernel page which has been charged to a memory cgroup
can't be mapped to userspace. The underlying reason is simple: PageKmemcg
flag is defined as a page type (like buddy, offline, etc), so it takes a
bit from a page->mapped counter. Pages with a type set can't be mapped to
userspace.
But in general the kmemcg flag has nothing to do with mapping to
userspace. It only means that the page has been accounted by the page
allocator, so it has to be properly uncharged on release.
Some bpf maps are mapping the vmalloc-based memory to userspace, and their
memory can't be accounted because of this implementation detail.
This patchset removes this limitation by moving the PageKmemcg flag into
one of the free bits of the page->mem_cgroup pointer. Also it formalizes
accesses to the page->mem_cgroup and page->obj_cgroups using new helpers,
adds several checks and removes a couple of obsolete functions. As the
result the code became more robust with fewer open-coded bit tricks.
This patch (of 4):
Currently there are many open-coded reads of the page->mem_cgroup pointer,
as well as a couple of read helpers, which are barely used.
It creates an obstacle on a way to reuse some bits of the pointer for
storing additional bits of information. In fact, we already do this for
slab pages, where the last bit indicates that a pointer has an attached
vector of objcg pointers instead of a regular memcg pointer.
This commits uses 2 existing helpers and introduces a new helper to
converts all read sides to calls of these helpers:
struct mem_cgroup *page_memcg(struct page *page);
struct mem_cgroup *page_memcg_rcu(struct page *page);
struct mem_cgroup *page_memcg_check(struct page *page);
page_memcg_check() is intended to be used in cases when the page can be a
slab page and have a memcg pointer pointing at objcg vector. It does
check the lowest bit, and if set, returns NULL. page_memcg() contains a
VM_BUG_ON_PAGE() check for the page not being a slab page.
To make sure nobody uses a direct access, struct page's
mem_cgroup/obj_cgroups is converted to unsigned long memcg_data.
Signed-off-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Link: https://lkml.kernel.org/r/20201027001657.3398190-1-guro@fb.com
Link: https://lkml.kernel.org/r/20201027001657.3398190-2-guro@fb.com
Link: https://lore.kernel.org/bpf/20201201215900.3569844-2-guro@fb.com
2020-12-01 21:58:27 +00:00
|
|
|
(unsigned long)page_memcg(page) |
|
2016-05-20 00:14:15 +00:00
|
|
|
#endif
|
|
|
|
(page->flags & check_flags)))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2020-06-03 22:58:39 +00:00
|
|
|
static const char *page_bad_reason(struct page *page, unsigned long flags)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2020-06-03 22:58:29 +00:00
|
|
|
const char *bad_reason = NULL;
|
2014-01-23 23:52:49 +00:00
|
|
|
|
2016-01-16 00:53:42 +00:00
|
|
|
if (unlikely(atomic_read(&page->_mapcount) != -1))
|
2014-01-23 23:52:49 +00:00
|
|
|
bad_reason = "nonzero mapcount";
|
|
|
|
if (unlikely(page->mapping != NULL))
|
|
|
|
bad_reason = "non-NULL mapping";
|
2016-03-17 21:19:26 +00:00
|
|
|
if (unlikely(page_ref_count(page) != 0))
|
2016-05-20 00:10:49 +00:00
|
|
|
bad_reason = "nonzero _refcount";
|
2020-06-03 22:58:39 +00:00
|
|
|
if (unlikely(page->flags & flags)) {
|
|
|
|
if (flags == PAGE_FLAGS_CHECK_AT_PREP)
|
|
|
|
bad_reason = "PAGE_FLAGS_CHECK_AT_PREP flag(s) set";
|
|
|
|
else
|
|
|
|
bad_reason = "PAGE_FLAGS_CHECK_AT_FREE flag(s) set";
|
2014-01-23 23:52:49 +00:00
|
|
|
}
|
2014-12-10 23:44:58 +00:00
|
|
|
#ifdef CONFIG_MEMCG
|
mm: memcontrol: Use helpers to read page's memcg data
Patch series "mm: allow mapping accounted kernel pages to userspace", v6.
Currently a non-slab kernel page which has been charged to a memory cgroup
can't be mapped to userspace. The underlying reason is simple: PageKmemcg
flag is defined as a page type (like buddy, offline, etc), so it takes a
bit from a page->mapped counter. Pages with a type set can't be mapped to
userspace.
But in general the kmemcg flag has nothing to do with mapping to
userspace. It only means that the page has been accounted by the page
allocator, so it has to be properly uncharged on release.
Some bpf maps are mapping the vmalloc-based memory to userspace, and their
memory can't be accounted because of this implementation detail.
This patchset removes this limitation by moving the PageKmemcg flag into
one of the free bits of the page->mem_cgroup pointer. Also it formalizes
accesses to the page->mem_cgroup and page->obj_cgroups using new helpers,
adds several checks and removes a couple of obsolete functions. As the
result the code became more robust with fewer open-coded bit tricks.
This patch (of 4):
Currently there are many open-coded reads of the page->mem_cgroup pointer,
as well as a couple of read helpers, which are barely used.
It creates an obstacle on a way to reuse some bits of the pointer for
storing additional bits of information. In fact, we already do this for
slab pages, where the last bit indicates that a pointer has an attached
vector of objcg pointers instead of a regular memcg pointer.
This commits uses 2 existing helpers and introduces a new helper to
converts all read sides to calls of these helpers:
struct mem_cgroup *page_memcg(struct page *page);
struct mem_cgroup *page_memcg_rcu(struct page *page);
struct mem_cgroup *page_memcg_check(struct page *page);
page_memcg_check() is intended to be used in cases when the page can be a
slab page and have a memcg pointer pointing at objcg vector. It does
check the lowest bit, and if set, returns NULL. page_memcg() contains a
VM_BUG_ON_PAGE() check for the page not being a slab page.
To make sure nobody uses a direct access, struct page's
mem_cgroup/obj_cgroups is converted to unsigned long memcg_data.
Signed-off-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Link: https://lkml.kernel.org/r/20201027001657.3398190-1-guro@fb.com
Link: https://lkml.kernel.org/r/20201027001657.3398190-2-guro@fb.com
Link: https://lore.kernel.org/bpf/20201201215900.3569844-2-guro@fb.com
2020-12-01 21:58:27 +00:00
|
|
|
if (unlikely(page_memcg(page)))
|
2014-12-10 23:44:58 +00:00
|
|
|
bad_reason = "page still charged to cgroup";
|
|
|
|
#endif
|
2020-06-03 22:58:39 +00:00
|
|
|
return bad_reason;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void check_free_page_bad(struct page *page)
|
|
|
|
{
|
|
|
|
bad_page(page,
|
|
|
|
page_bad_reason(page, PAGE_FLAGS_CHECK_AT_FREE));
|
2016-05-20 00:14:18 +00:00
|
|
|
}
|
|
|
|
|
2020-06-03 22:58:36 +00:00
|
|
|
static inline int check_free_page(struct page *page)
|
2016-05-20 00:14:18 +00:00
|
|
|
{
|
2016-05-20 00:14:21 +00:00
|
|
|
if (likely(page_expected_state(page, PAGE_FLAGS_CHECK_AT_FREE)))
|
2016-05-20 00:14:18 +00:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Something has gone sideways, find it */
|
2020-06-03 22:58:33 +00:00
|
|
|
check_free_page_bad(page);
|
2016-05-20 00:14:15 +00:00
|
|
|
return 1;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:32 +00:00
|
|
|
static int free_tail_pages_check(struct page *head_page, struct page *page)
|
|
|
|
{
|
|
|
|
int ret = 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We rely page->lru.next never has bit 0 set, unless the page
|
|
|
|
* is PageTail(). Let's make sure that's true even for poisoned ->lru.
|
|
|
|
*/
|
|
|
|
BUILD_BUG_ON((unsigned long)LIST_POISON1 & 1);
|
|
|
|
|
|
|
|
if (!IS_ENABLED(CONFIG_DEBUG_VM)) {
|
|
|
|
ret = 0;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
switch (page - head_page) {
|
|
|
|
case 1:
|
2018-06-08 00:08:50 +00:00
|
|
|
/* the first tail page: ->mapping may be compound_mapcount() */
|
2016-05-20 00:14:32 +00:00
|
|
|
if (unlikely(compound_mapcount(page))) {
|
2020-06-03 22:58:29 +00:00
|
|
|
bad_page(page, "nonzero compound_mapcount");
|
2016-05-20 00:14:32 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
/*
|
|
|
|
* the second tail page: ->mapping is
|
2018-06-08 00:08:42 +00:00
|
|
|
* deferred_list.next -- ignore value.
|
2016-05-20 00:14:32 +00:00
|
|
|
*/
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
if (page->mapping != TAIL_MAPPING) {
|
2020-06-03 22:58:29 +00:00
|
|
|
bad_page(page, "corrupted mapping in tail page");
|
2016-05-20 00:14:32 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (unlikely(!PageTail(page))) {
|
2020-06-03 22:58:29 +00:00
|
|
|
bad_page(page, "PageTail not set");
|
2016-05-20 00:14:32 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (unlikely(compound_head(page) != head_page)) {
|
2020-06-03 22:58:29 +00:00
|
|
|
bad_page(page, "compound_head not consistent");
|
2016-05-20 00:14:32 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
out:
|
|
|
|
page->mapping = NULL;
|
|
|
|
clear_compound_head(page);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
static void kernel_init_free_pages(struct page *page, int numpages)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2020-08-07 06:25:54 +00:00
|
|
|
/* s390's use of memset() could override KASAN redzones. */
|
|
|
|
kasan_disable_current();
|
2020-12-22 20:02:17 +00:00
|
|
|
for (i = 0; i < numpages; i++) {
|
2021-01-24 05:01:43 +00:00
|
|
|
u8 tag = page_kasan_tag(page + i);
|
2020-12-22 20:02:17 +00:00
|
|
|
page_kasan_tag_reset(page + i);
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
clear_highpage(page + i);
|
2021-01-24 05:01:43 +00:00
|
|
|
page_kasan_tag_set(page + i, tag);
|
2020-12-22 20:02:17 +00:00
|
|
|
}
|
2020-08-07 06:25:54 +00:00
|
|
|
kasan_enable_current();
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:38 +00:00
|
|
|
static __always_inline bool free_pages_prepare(struct page *page,
|
|
|
|
unsigned int order, bool check_free)
|
2016-05-20 00:14:32 +00:00
|
|
|
{
|
2016-05-20 00:14:38 +00:00
|
|
|
int bad = 0;
|
2016-05-20 00:14:32 +00:00
|
|
|
|
|
|
|
VM_BUG_ON_PAGE(PageTail(page), page);
|
|
|
|
|
2016-05-20 00:14:38 +00:00
|
|
|
trace_mm_page_free(page, order);
|
|
|
|
|
mm,hwpoison: rework soft offline for in-use pages
This patch changes the way we set and handle in-use poisoned pages. Until
now, poisoned pages were released to the buddy allocator, trusting that
the checks that take place at allocation time would act as a safe net and
would skip that page.
This has proved to be wrong, as we got some pfn walkers out there, like
compaction, that all they care is the page to be in a buddy freelist.
Although this might not be the only user, having poisoned pages in the
buddy allocator seems a bad idea as we should only have free pages that
are ready and meant to be used as such.
Before explaining the taken approach, let us break down the kind of pages
we can soft offline.
- Anonymous THP (after the split, they end up being 4K pages)
- Hugetlb
- Order-0 pages (that can be either migrated or invalited)
* Normal pages (order-0 and anon-THP)
- If they are clean and unmapped page cache pages, we invalidate
then by means of invalidate_inode_page().
- If they are mapped/dirty, we do the isolate-and-migrate dance.
Either way, do not call put_page directly from those paths. Instead, we
keep the page and send it to page_handle_poison to perform the right
handling.
page_handle_poison sets the HWPoison flag and does the last put_page.
Down the chain, we placed a check for HWPoison page in
free_pages_prepare, that just skips any poisoned page, so those pages
do not end up in any pcplist/freelist.
After that, we set the refcount on the page to 1 and we increment
the poisoned pages counter.
If we see that the check in free_pages_prepare creates trouble, we can
always do what we do for free pages:
- wait until the page hits buddy's freelists
- take it off, and flag it
The downside of the above approach is that we could race with an
allocation, so by the time we want to take the page off the buddy, the
page has been already allocated so we cannot soft offline it.
But the user could always retry it.
* Hugetlb pages
- We isolate-and-migrate them
After the migration has been successful, we call dissolve_free_huge_page,
and we set HWPoison on the page if we succeed.
Hugetlb has a slightly different handling though.
While for non-hugetlb pages we cared about closing the race with an
allocation, doing so for hugetlb pages requires quite some additional
and intrusive code (we would need to hook in free_huge_page and some other
places).
So I decided to not make the code overly complicated and just fail
normally if the page we allocated in the meantime.
We can always build on top of this.
As a bonus, because of the way we handle now in-use pages, we no longer
need the put-as-isolation-migratetype dance, that was guarding for poisoned
pages to end up in pcplists.
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Aristeu Rozanski <aris@ruivo.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Dmitry Yakunin <zeil@yandex-team.ru>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Oscar Salvador <osalvador@suse.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20200922135650.1634-10-osalvador@suse.de
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:07:09 +00:00
|
|
|
if (unlikely(PageHWPoison(page)) && !order) {
|
|
|
|
/*
|
|
|
|
* Do not let hwpoison pages hit pcplists/buddy
|
|
|
|
* Untie memcg state and reset page's owner
|
|
|
|
*/
|
2020-12-01 21:58:30 +00:00
|
|
|
if (memcg_kmem_enabled() && PageMemcgKmem(page))
|
mm,hwpoison: rework soft offline for in-use pages
This patch changes the way we set and handle in-use poisoned pages. Until
now, poisoned pages were released to the buddy allocator, trusting that
the checks that take place at allocation time would act as a safe net and
would skip that page.
This has proved to be wrong, as we got some pfn walkers out there, like
compaction, that all they care is the page to be in a buddy freelist.
Although this might not be the only user, having poisoned pages in the
buddy allocator seems a bad idea as we should only have free pages that
are ready and meant to be used as such.
Before explaining the taken approach, let us break down the kind of pages
we can soft offline.
- Anonymous THP (after the split, they end up being 4K pages)
- Hugetlb
- Order-0 pages (that can be either migrated or invalited)
* Normal pages (order-0 and anon-THP)
- If they are clean and unmapped page cache pages, we invalidate
then by means of invalidate_inode_page().
- If they are mapped/dirty, we do the isolate-and-migrate dance.
Either way, do not call put_page directly from those paths. Instead, we
keep the page and send it to page_handle_poison to perform the right
handling.
page_handle_poison sets the HWPoison flag and does the last put_page.
Down the chain, we placed a check for HWPoison page in
free_pages_prepare, that just skips any poisoned page, so those pages
do not end up in any pcplist/freelist.
After that, we set the refcount on the page to 1 and we increment
the poisoned pages counter.
If we see that the check in free_pages_prepare creates trouble, we can
always do what we do for free pages:
- wait until the page hits buddy's freelists
- take it off, and flag it
The downside of the above approach is that we could race with an
allocation, so by the time we want to take the page off the buddy, the
page has been already allocated so we cannot soft offline it.
But the user could always retry it.
* Hugetlb pages
- We isolate-and-migrate them
After the migration has been successful, we call dissolve_free_huge_page,
and we set HWPoison on the page if we succeed.
Hugetlb has a slightly different handling though.
While for non-hugetlb pages we cared about closing the race with an
allocation, doing so for hugetlb pages requires quite some additional
and intrusive code (we would need to hook in free_huge_page and some other
places).
So I decided to not make the code overly complicated and just fail
normally if the page we allocated in the meantime.
We can always build on top of this.
As a bonus, because of the way we handle now in-use pages, we no longer
need the put-as-isolation-migratetype dance, that was guarding for poisoned
pages to end up in pcplists.
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Aristeu Rozanski <aris@ruivo.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Dmitry Yakunin <zeil@yandex-team.ru>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Oscar Salvador <osalvador@suse.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/20200922135650.1634-10-osalvador@suse.de
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:07:09 +00:00
|
|
|
__memcg_kmem_uncharge_page(page, order);
|
|
|
|
reset_page_owner(page, order);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:38 +00:00
|
|
|
/*
|
|
|
|
* Check tail pages before head page information is cleared to
|
|
|
|
* avoid checking PageCompound for order-0 pages.
|
|
|
|
*/
|
|
|
|
if (unlikely(order)) {
|
|
|
|
bool compound = PageCompound(page);
|
|
|
|
int i;
|
|
|
|
|
|
|
|
VM_BUG_ON_PAGE(compound && compound_order(page) != order, page);
|
2016-05-20 00:14:32 +00:00
|
|
|
|
2016-07-26 22:25:53 +00:00
|
|
|
if (compound)
|
|
|
|
ClearPageDoubleMap(page);
|
2016-05-20 00:14:38 +00:00
|
|
|
for (i = 1; i < (1 << order); i++) {
|
|
|
|
if (compound)
|
|
|
|
bad += free_tail_pages_check(page, page + i);
|
2020-06-03 22:58:36 +00:00
|
|
|
if (unlikely(check_free_page(page + i))) {
|
2016-05-20 00:14:38 +00:00
|
|
|
bad++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
(page + i)->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
|
|
|
|
}
|
|
|
|
}
|
mm: migrate: support non-lru movable page migration
We have allowed migration for only LRU pages until now and it was enough
to make high-order pages. But recently, embedded system(e.g., webOS,
android) uses lots of non-movable pages(e.g., zram, GPU memory) so we
have seen several reports about troubles of small high-order allocation.
For fixing the problem, there were several efforts (e,g,. enhance
compaction algorithm, SLUB fallback to 0-order page, reserved memory,
vmalloc and so on) but if there are lots of non-movable pages in system,
their solutions are void in the long run.
So, this patch is to support facility to change non-movable pages with
movable. For the feature, this patch introduces functions related to
migration to address_space_operations as well as some page flags.
If a driver want to make own pages movable, it should define three
functions which are function pointers of struct
address_space_operations.
1. bool (*isolate_page) (struct page *page, isolate_mode_t mode);
What VM expects on isolate_page function of driver is to return *true*
if driver isolates page successfully. On returing true, VM marks the
page as PG_isolated so concurrent isolation in several CPUs skip the
page for isolation. If a driver cannot isolate the page, it should
return *false*.
Once page is successfully isolated, VM uses page.lru fields so driver
shouldn't expect to preserve values in that fields.
2. int (*migratepage) (struct address_space *mapping,
struct page *newpage, struct page *oldpage, enum migrate_mode);
After isolation, VM calls migratepage of driver with isolated page. The
function of migratepage is to move content of the old page to new page
and set up fields of struct page newpage. Keep in mind that you should
indicate to the VM the oldpage is no longer movable via
__ClearPageMovable() under page_lock if you migrated the oldpage
successfully and returns 0. If driver cannot migrate the page at the
moment, driver can return -EAGAIN. On -EAGAIN, VM will retry page
migration in a short time because VM interprets -EAGAIN as "temporal
migration failure". On returning any error except -EAGAIN, VM will give
up the page migration without retrying in this time.
Driver shouldn't touch page.lru field VM using in the functions.
3. void (*putback_page)(struct page *);
If migration fails on isolated page, VM should return the isolated page
to the driver so VM calls driver's putback_page with migration failed
page. In this function, driver should put the isolated page back to the
own data structure.
4. non-lru movable page flags
There are two page flags for supporting non-lru movable page.
* PG_movable
Driver should use the below function to make page movable under
page_lock.
void __SetPageMovable(struct page *page, struct address_space *mapping)
It needs argument of address_space for registering migration family
functions which will be called by VM. Exactly speaking, PG_movable is
not a real flag of struct page. Rather than, VM reuses page->mapping's
lower bits to represent it.
#define PAGE_MAPPING_MOVABLE 0x2
page->mapping = page->mapping | PAGE_MAPPING_MOVABLE;
so driver shouldn't access page->mapping directly. Instead, driver
should use page_mapping which mask off the low two bits of page->mapping
so it can get right struct address_space.
For testing of non-lru movable page, VM supports __PageMovable function.
However, it doesn't guarantee to identify non-lru movable page because
page->mapping field is unified with other variables in struct page. As
well, if driver releases the page after isolation by VM, page->mapping
doesn't have stable value although it has PAGE_MAPPING_MOVABLE (Look at
__ClearPageMovable). But __PageMovable is cheap to catch whether page
is LRU or non-lru movable once the page has been isolated. Because LRU
pages never can have PAGE_MAPPING_MOVABLE in page->mapping. It is also
good for just peeking to test non-lru movable pages before more
expensive checking with lock_page in pfn scanning to select victim.
For guaranteeing non-lru movable page, VM provides PageMovable function.
Unlike __PageMovable, PageMovable functions validates page->mapping and
mapping->a_ops->isolate_page under lock_page. The lock_page prevents
sudden destroying of page->mapping.
Driver using __SetPageMovable should clear the flag via
__ClearMovablePage under page_lock before the releasing the page.
* PG_isolated
To prevent concurrent isolation among several CPUs, VM marks isolated
page as PG_isolated under lock_page. So if a CPU encounters PG_isolated
non-lru movable page, it can skip it. Driver doesn't need to manipulate
the flag because VM will set/clear it automatically. Keep in mind that
if driver sees PG_isolated page, it means the page have been isolated by
VM so it shouldn't touch page.lru field. PG_isolated is alias with
PG_reclaim flag so driver shouldn't use the flag for own purpose.
[opensource.ganesh@gmail.com: mm/compaction: remove local variable is_lru]
Link: http://lkml.kernel.org/r/20160618014841.GA7422@leo-test
Link: http://lkml.kernel.org/r/1464736881-24886-3-git-send-email-minchan@kernel.org
Signed-off-by: Gioh Kim <gi-oh.kim@profitbricks.com>
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Rafael Aquini <aquini@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: John Einar Reitan <john.reitan@foss.arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:05 +00:00
|
|
|
if (PageMappingFlags(page))
|
2016-05-20 00:14:32 +00:00
|
|
|
page->mapping = NULL;
|
2020-12-01 21:58:30 +00:00
|
|
|
if (memcg_kmem_enabled() && PageMemcgKmem(page))
|
2020-04-02 04:06:46 +00:00
|
|
|
__memcg_kmem_uncharge_page(page, order);
|
2016-05-20 00:14:38 +00:00
|
|
|
if (check_free)
|
2020-06-03 22:58:36 +00:00
|
|
|
bad += check_free_page(page);
|
2016-05-20 00:14:38 +00:00
|
|
|
if (bad)
|
|
|
|
return false;
|
2016-05-20 00:14:32 +00:00
|
|
|
|
2016-05-20 00:14:38 +00:00
|
|
|
page_cpupid_reset_last(page);
|
|
|
|
page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
|
|
|
|
reset_page_owner(page, order);
|
2016-05-20 00:14:32 +00:00
|
|
|
|
|
|
|
if (!PageHighMem(page)) {
|
|
|
|
debug_check_no_locks_freed(page_address(page),
|
2016-05-20 00:14:38 +00:00
|
|
|
PAGE_SIZE << order);
|
2016-05-20 00:14:32 +00:00
|
|
|
debug_check_no_obj_freed(page_address(page),
|
2016-05-20 00:14:38 +00:00
|
|
|
PAGE_SIZE << order);
|
2016-05-20 00:14:32 +00:00
|
|
|
}
|
mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options
Patch series "add init_on_alloc/init_on_free boot options", v10.
Provide init_on_alloc and init_on_free boot options.
These are aimed at preventing possible information leaks and making the
control-flow bugs that depend on uninitialized values more deterministic.
Enabling either of the options guarantees that the memory returned by the
page allocator and SL[AU]B is initialized with zeroes. SLOB allocator
isn't supported at the moment, as its emulation of kmem caches complicates
handling of SLAB_TYPESAFE_BY_RCU caches correctly.
Enabling init_on_free also guarantees that pages and heap objects are
initialized right after they're freed, so it won't be possible to access
stale data by using a dangling pointer.
As suggested by Michal Hocko, right now we don't let the heap users to
disable initialization for certain allocations. There's not enough
evidence that doing so can speed up real-life cases, and introducing ways
to opt-out may result in things going out of control.
This patch (of 2):
The new options are needed to prevent possible information leaks and make
control-flow bugs that depend on uninitialized values more deterministic.
This is expected to be on-by-default on Android and Chrome OS. And it
gives the opportunity for anyone else to use it under distros too via the
boot args. (The init_on_free feature is regularly requested by folks
where memory forensics is included in their threat models.)
init_on_alloc=1 makes the kernel initialize newly allocated pages and heap
objects with zeroes. Initialization is done at allocation time at the
places where checks for __GFP_ZERO are performed.
init_on_free=1 makes the kernel initialize freed pages and heap objects
with zeroes upon their deletion. This helps to ensure sensitive data
doesn't leak via use-after-free accesses.
Both init_on_alloc=1 and init_on_free=1 guarantee that the allocator
returns zeroed memory. The two exceptions are slab caches with
constructors and SLAB_TYPESAFE_BY_RCU flag. Those are never
zero-initialized to preserve their semantics.
Both init_on_alloc and init_on_free default to zero, but those defaults
can be overridden with CONFIG_INIT_ON_ALLOC_DEFAULT_ON and
CONFIG_INIT_ON_FREE_DEFAULT_ON.
If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.
Slowdown for the new features compared to init_on_free=0, init_on_alloc=0:
hackbench, init_on_free=1: +7.62% sys time (st.err 0.74%)
hackbench, init_on_alloc=1: +7.75% sys time (st.err 2.14%)
Linux build with -j12, init_on_free=1: +8.38% wall time (st.err 0.39%)
Linux build with -j12, init_on_free=1: +24.42% sys time (st.err 0.52%)
Linux build with -j12, init_on_alloc=1: -0.13% wall time (st.err 0.42%)
Linux build with -j12, init_on_alloc=1: +0.57% sys time (st.err 0.40%)
The slowdown for init_on_free=0, init_on_alloc=0 compared to the baseline
is within the standard error.
The new features are also going to pave the way for hardware memory
tagging (e.g. arm64's MTE), which will require both on_alloc and on_free
hooks to set the tags for heap objects. With MTE, tagging will have the
same cost as memory initialization.
Although init_on_free is rather costly, there are paranoid use-cases where
in-memory data lifetime is desired to be minimized. There are various
arguments for/against the realism of the associated threat models, but
given that we'll need the infrastructure for MTE anyway, and there are
people who want wipe-on-free behavior no matter what the performance cost,
it seems reasonable to include it in this series.
[glider@google.com: v8]
Link: http://lkml.kernel.org/r/20190626121943.131390-2-glider@google.com
[glider@google.com: v9]
Link: http://lkml.kernel.org/r/20190627130316.254309-2-glider@google.com
[glider@google.com: v10]
Link: http://lkml.kernel.org/r/20190628093131.199499-2-glider@google.com
Link: http://lkml.kernel.org/r/20190617151050.92663-2-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.cz> [page and dmapool parts
Acked-by: James Morris <jamorris@linux.microsoft.com>]
Cc: Christoph Lameter <cl@linux.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Sandeep Patil <sspatil@android.com>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:59:19 +00:00
|
|
|
if (want_init_on_free())
|
|
|
|
kernel_init_free_pages(page, 1 << order);
|
|
|
|
|
2020-12-15 03:13:34 +00:00
|
|
|
kernel_poison_pages(page, 1 << order);
|
|
|
|
|
mm/page_alloc.c: fix a crash in free_pages_prepare()
On architectures like s390, arch_free_page() could mark the page unused
(set_page_unused()) and any access later would trigger a kernel panic.
Fix it by moving arch_free_page() after all possible accessing calls.
Hardware name: IBM 2964 N96 400 (z/VM 6.4.0)
Krnl PSW : 0404e00180000000 0000000026c2b96e (__free_pages_ok+0x34e/0x5d8)
R:0 T:1 IO:0 EX:0 Key:0 M:1 W:0 P:0 AS:3 CC:2 PM:0 RI:0 EA:3
Krnl GPRS: 0000000088d43af7 0000000000484000 000000000000007c 000000000000000f
000003d080012100 000003d080013fc0 0000000000000000 0000000000100000
00000000275cca48 0000000000000100 0000000000000008 000003d080010000
00000000000001d0 000003d000000000 0000000026c2b78a 000000002717fdb0
Krnl Code: 0000000026c2b95c: ec1100b30659 risbgn %r1,%r1,0,179,6
0000000026c2b962: e32014000036 pfd 2,1024(%r1)
#0000000026c2b968: d7ff10001000 xc 0(256,%r1),0(%r1)
>0000000026c2b96e: 41101100 la %r1,256(%r1)
0000000026c2b972: a737fff8 brctg %r3,26c2b962
0000000026c2b976: d7ff10001000 xc 0(256,%r1),0(%r1)
0000000026c2b97c: e31003400004 lg %r1,832
0000000026c2b982: ebff1430016a asi 5168(%r1),-1
Call Trace:
__free_pages_ok+0x16a/0x5d8)
memblock_free_all+0x206/0x290
mem_init+0x58/0x120
start_kernel+0x2b0/0x570
startup_continue+0x6a/0xc0
INFO: lockdep is turned off.
Last Breaking-Event-Address:
__free_pages_ok+0x372/0x5d8
Kernel panic - not syncing: Fatal exception: panic_on_oops
00: HCPGIR450W CP entered; disabled wait PSW 00020001 80000000 00000000 26A2379C
In the past, only kernel_poison_pages() would trigger this but it needs
"page_poison=on" kernel cmdline, and I suspect nobody tested that on
s390. Recently, kernel_init_free_pages() (commit 6471384af2a6 ("mm:
security: introduce init_on_alloc=1 and init_on_free=1 boot options"))
was added and could trigger this as well.
[akpm@linux-foundation.org: add comment]
Link: http://lkml.kernel.org/r/1569613623-16820-1-git-send-email-cai@lca.pw
Fixes: 8823b1dbc05f ("mm/page_poison.c: enable PAGE_POISONING as a separate option")
Fixes: 6471384af2a6 ("mm: security: introduce init_on_alloc=1 and init_on_free=1 boot options")
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Cc: <stable@vger.kernel.org> [5.3+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-10-07 00:58:25 +00:00
|
|
|
/*
|
|
|
|
* arch_free_page() can make the page's contents inaccessible. s390
|
|
|
|
* does this. So nothing which can access the page's contents should
|
|
|
|
* happen after this.
|
|
|
|
*/
|
|
|
|
arch_free_page(page, order);
|
|
|
|
|
mm: introduce debug_pagealloc_{map,unmap}_pages() helpers
Patch series "arch, mm: improve robustness of direct map manipulation", v7.
During recent discussion about KVM protected memory, David raised a
concern about usage of __kernel_map_pages() outside of DEBUG_PAGEALLOC
scope [1].
Indeed, for architectures that define CONFIG_ARCH_HAS_SET_DIRECT_MAP it is
possible that __kernel_map_pages() would fail, but since this function is
void, the failure will go unnoticed.
Moreover, there's lack of consistency of __kernel_map_pages() semantics
across architectures as some guard this function with #ifdef
DEBUG_PAGEALLOC, some refuse to update the direct map if page allocation
debugging is disabled at run time and some allow modifying the direct map
regardless of DEBUG_PAGEALLOC settings.
This set straightens this out by restoring dependency of
__kernel_map_pages() on DEBUG_PAGEALLOC and updating the call sites
accordingly.
Since currently the only user of __kernel_map_pages() outside
DEBUG_PAGEALLOC is hibernation, it is updated to make direct map accesses
there more explicit.
[1] https://lore.kernel.org/lkml/2759b4bf-e1e3-d006-7d86-78a40348269d@redhat.com
This patch (of 4):
When CONFIG_DEBUG_PAGEALLOC is enabled, it unmaps pages from the kernel
direct mapping after free_pages(). The pages than need to be mapped back
before they could be used. Theese mapping operations use
__kernel_map_pages() guarded with with debug_pagealloc_enabled().
The only place that calls __kernel_map_pages() without checking whether
DEBUG_PAGEALLOC is enabled is the hibernation code that presumes
availability of this function when ARCH_HAS_SET_DIRECT_MAP is set. Still,
on arm64, __kernel_map_pages() will bail out when DEBUG_PAGEALLOC is not
enabled but set_direct_map_invalid_noflush() may render some pages not
present in the direct map and hibernation code won't be able to save such
pages.
To make page allocation debugging and hibernation interaction more robust,
the dependency on DEBUG_PAGEALLOC or ARCH_HAS_SET_DIRECT_MAP has to be
made more explicit.
Start with combining the guard condition and the call to
__kernel_map_pages() into debug_pagealloc_map_pages() and
debug_pagealloc_unmap_pages() functions to emphasize that
__kernel_map_pages() should not be called without DEBUG_PAGEALLOC and use
these new functions to map/unmap pages when page allocation debugging is
enabled.
Link: https://lkml.kernel.org/r/20201109192128.960-1-rppt@kernel.org
Link: https://lkml.kernel.org/r/20201109192128.960-2-rppt@kernel.org
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: "Edgecombe, Rick P" <rick.p.edgecombe@intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:20 +00:00
|
|
|
debug_pagealloc_unmap_pages(page, 1 << order);
|
2019-04-26 00:11:35 +00:00
|
|
|
|
mm/page_alloc.c: don't call kasan_free_pages() at deferred mem init
When CONFIG_KASAN is enabled on large memory SMP systems, the deferrred
pages initialization can take a long time. Below were the reported init
times on a 8-socket 96-core 4TB IvyBridge system.
1) Non-debug kernel without CONFIG_KASAN
[ 8.764222] node 1 initialised, 132086516 pages in 7027ms
2) Debug kernel with CONFIG_KASAN
[ 146.288115] node 1 initialised, 132075466 pages in 143052ms
So the page init time in a debug kernel was 20X of the non-debug kernel.
The long init time can be problematic as the page initialization is done
with interrupt disabled. In this particular case, it caused the
appearance of following warning messages as well as NMI backtraces of all
the cores that were doing the initialization.
[ 68.240049] rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
[ 68.241000] rcu: 25-...0: (100 ticks this GP) idle=b72/1/0x4000000000000000 softirq=915/915 fqs=16252
[ 68.241000] rcu: 44-...0: (95 ticks this GP) idle=49a/1/0x4000000000000000 softirq=788/788 fqs=16253
[ 68.241000] rcu: 54-...0: (104 ticks this GP) idle=03a/1/0x4000000000000000 softirq=721/825 fqs=16253
[ 68.241000] rcu: 60-...0: (103 ticks this GP) idle=cbe/1/0x4000000000000000 softirq=637/740 fqs=16253
[ 68.241000] rcu: 72-...0: (105 ticks this GP) idle=786/1/0x4000000000000000 softirq=536/641 fqs=16253
[ 68.241000] rcu: 84-...0: (99 ticks this GP) idle=292/1/0x4000000000000000 softirq=537/537 fqs=16253
[ 68.241000] rcu: 111-...0: (104 ticks this GP) idle=bde/1/0x4000000000000000 softirq=474/476 fqs=16253
[ 68.241000] rcu: (detected by 13, t=65018 jiffies, g=249, q=2)
The long init time was mainly caused by the call to kasan_free_pages() to
poison the newly initialized pages. On a 4TB system, we are talking about
almost 500GB of memory probably on the same node.
In reality, we may not need to poison the newly initialized pages before
they are ever allocated. So KASAN poisoning of freed pages before the
completion of deferred memory initialization is now disabled. Those pages
will be properly poisoned when they are allocated or freed after deferred
pages initialization is done.
With this change, the new page initialization time became:
[ 21.948010] node 1 initialised, 132075466 pages in 18702ms
This was still about double the non-debug kernel time, but was much
better than before.
Link: http://lkml.kernel.org/r/1544459388-8736-1-git-send-email-longman@redhat.com
Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:38:51 +00:00
|
|
|
kasan_free_nondeferred_pages(page, order);
|
2016-05-20 00:14:32 +00:00
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:38 +00:00
|
|
|
#ifdef CONFIG_DEBUG_VM
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
/*
|
|
|
|
* With DEBUG_VM enabled, order-0 pages are checked immediately when being freed
|
|
|
|
* to pcp lists. With debug_pagealloc also enabled, they are also rechecked when
|
|
|
|
* moved from pcp lists to free lists.
|
|
|
|
*/
|
|
|
|
static bool free_pcp_prepare(struct page *page)
|
2016-05-20 00:14:38 +00:00
|
|
|
{
|
|
|
|
return free_pages_prepare(page, 0, true);
|
|
|
|
}
|
|
|
|
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
static bool bulkfree_pcp_prepare(struct page *page)
|
2016-05-20 00:14:38 +00:00
|
|
|
{
|
mm, debug_pagealloc: don't rely on static keys too early
Commit 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable
debugging") has introduced a static key to reduce overhead when
debug_pagealloc is compiled in but not enabled. It relied on the
assumption that jump_label_init() is called before parse_early_param()
as in start_kernel(), so when the "debug_pagealloc=on" option is parsed,
it is safe to enable the static key.
However, it turns out multiple architectures call parse_early_param()
earlier from their setup_arch(). x86 also calls jump_label_init() even
earlier, so no issue was found while testing the commit, but same is not
true for e.g. ppc64 and s390 where the kernel would not boot with
debug_pagealloc=on as found by our QA.
To fix this without tricky changes to init code of multiple
architectures, this patch partially reverts the static key conversion
from 96a2b03f281d. Init-time and non-fastpath calls (such as in arch
code) of debug_pagealloc_enabled() will again test a simple bool
variable. Fastpath mm code is converted to a new
debug_pagealloc_enabled_static() variant that relies on the static key,
which is enabled in a well-defined point in mm_init() where it's
guaranteed that jump_label_init() has been called, regardless of
architecture.
[sfr@canb.auug.org.au: export _debug_pagealloc_enabled_early]
Link: http://lkml.kernel.org/r/20200106164944.063ac07b@canb.auug.org.au
Link: http://lkml.kernel.org/r/20191219130612.23171-1-vbabka@suse.cz
Fixes: 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable debugging")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Qian Cai <cai@lca.pw>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-14 00:29:20 +00:00
|
|
|
if (debug_pagealloc_enabled_static())
|
2020-06-03 22:58:36 +00:00
|
|
|
return check_free_page(page);
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
else
|
|
|
|
return false;
|
2016-05-20 00:14:38 +00:00
|
|
|
}
|
|
|
|
#else
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
/*
|
|
|
|
* With DEBUG_VM disabled, order-0 pages being freed are checked only when
|
|
|
|
* moving from pcp lists to free list in order to reduce overhead. With
|
|
|
|
* debug_pagealloc enabled, they are checked also immediately when being freed
|
|
|
|
* to the pcp lists.
|
|
|
|
*/
|
2016-05-20 00:14:38 +00:00
|
|
|
static bool free_pcp_prepare(struct page *page)
|
|
|
|
{
|
mm, debug_pagealloc: don't rely on static keys too early
Commit 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable
debugging") has introduced a static key to reduce overhead when
debug_pagealloc is compiled in but not enabled. It relied on the
assumption that jump_label_init() is called before parse_early_param()
as in start_kernel(), so when the "debug_pagealloc=on" option is parsed,
it is safe to enable the static key.
However, it turns out multiple architectures call parse_early_param()
earlier from their setup_arch(). x86 also calls jump_label_init() even
earlier, so no issue was found while testing the commit, but same is not
true for e.g. ppc64 and s390 where the kernel would not boot with
debug_pagealloc=on as found by our QA.
To fix this without tricky changes to init code of multiple
architectures, this patch partially reverts the static key conversion
from 96a2b03f281d. Init-time and non-fastpath calls (such as in arch
code) of debug_pagealloc_enabled() will again test a simple bool
variable. Fastpath mm code is converted to a new
debug_pagealloc_enabled_static() variant that relies on the static key,
which is enabled in a well-defined point in mm_init() where it's
guaranteed that jump_label_init() has been called, regardless of
architecture.
[sfr@canb.auug.org.au: export _debug_pagealloc_enabled_early]
Link: http://lkml.kernel.org/r/20200106164944.063ac07b@canb.auug.org.au
Link: http://lkml.kernel.org/r/20191219130612.23171-1-vbabka@suse.cz
Fixes: 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable debugging")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Qian Cai <cai@lca.pw>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-14 00:29:20 +00:00
|
|
|
if (debug_pagealloc_enabled_static())
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
return free_pages_prepare(page, 0, true);
|
|
|
|
else
|
|
|
|
return free_pages_prepare(page, 0, false);
|
2016-05-20 00:14:38 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:32 +00:00
|
|
|
static bool bulkfree_pcp_prepare(struct page *page)
|
|
|
|
{
|
2020-06-03 22:58:36 +00:00
|
|
|
return check_free_page(page);
|
2016-05-20 00:14:32 +00:00
|
|
|
}
|
|
|
|
#endif /* CONFIG_DEBUG_VM */
|
|
|
|
|
mm/free_pcppages_bulk: prefetch buddy while not holding lock
When a page is freed back to the global pool, its buddy will be checked
to see if it's possible to do a merge. This requires accessing buddy's
page structure and that access could take a long time if it's cache
cold.
This patch adds a prefetch to the to-be-freed page's buddy outside of
zone->lock in hope of accessing buddy's page structure later under
zone->lock will be faster. Since we *always* do buddy merging and check
an order-0 page's buddy to try to merge it when it goes into the main
allocator, the cacheline will always come in, i.e. the prefetched data
will never be unused.
Normally, the number of prefetch will be pcp->batch(default=31 and has
an upper limit of (PAGE_SHIFT * 8)=96 on x86_64) but in the case of
pcp's pages get all drained, it will be pcp->count which has an upper
limit of pcp->high. pcp->high, although has a default value of 186
(pcp->batch=31 * 6), can be changed by user through
/proc/sys/vm/percpu_pagelist_fraction and there is no software upper
limit so could be large, like several thousand. For this reason, only
the first pcp->batch number of page's buddy structure is prefetched to
avoid excessive prefetching.
In the meantime, there are two concerns:
1. the prefetch could potentially evict existing cachelines, especially
for L1D cache since it is not huge
2. there is some additional instruction overhead, namely calculating
buddy pfn twice
For 1, it's hard to say, this microbenchmark though shows good result
but the actual benefit of this patch will be workload/CPU dependant;
For 2, since the calculation is a XOR on two local variables, it's
expected in many cases that cycles spent will be offset by reduced
memory latency later. This is especially true for NUMA machines where
multiple CPUs are contending on zone->lock and the most time consuming
part under zone->lock is the wait of 'struct page' cacheline of the
to-be-freed pages and their buddies.
Test with will-it-scale/page_fault1 full load:
kernel Broadwell(2S) Skylake(2S) Broadwell(4S) Skylake(4S)
v4.16-rc2+ 9034215 7971818 13667135 15677465
patch2/3 9536374 +5.6% 8314710 +4.3% 14070408 +3.0% 16675866 +6.4%
this patch 10180856 +6.8% 8506369 +2.3% 14756865 +4.9% 17325324 +3.9%
Note: this patch's performance improvement percent is against patch2/3.
(Changelog stolen from Dave Hansen and Mel Gorman's comments at
http://lkml.kernel.org/r/148a42d8-8306-2f2f-7f7c-86bc118f8ccd@intel.com)
[aaron.lu@intel.com: use helper function, avoid disordering pages]
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Link: http://lkml.kernel.org/r/20180320113146.GB24737@intel.com
[aaron.lu@intel.com: v4]
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Link: http://lkml.kernel.org/r/20180309082431.GB30868@intel.com
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Suggested-by: Ying Huang <ying.huang@intel.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:24:14 +00:00
|
|
|
static inline void prefetch_buddy(struct page *page)
|
|
|
|
{
|
|
|
|
unsigned long pfn = page_to_pfn(page);
|
|
|
|
unsigned long buddy_pfn = __find_buddy_pfn(pfn, 0);
|
|
|
|
struct page *buddy = page + (buddy_pfn - pfn);
|
|
|
|
|
|
|
|
prefetch(buddy);
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
* Frees a number of pages from the PCP lists
|
2005-04-16 22:20:36 +00:00
|
|
|
* Assumes all pages on list are in same zone, and of same order.
|
2005-09-10 07:26:59 +00:00
|
|
|
* count is the number of pages to free.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
|
|
|
* If the zone was previously in an "all pages pinned" state then look to
|
|
|
|
* see if this freeing clears that state.
|
|
|
|
*
|
|
|
|
* And clear the zone's pages_scanned counter, to hold off the "all pages are
|
|
|
|
* pinned" detection logic.
|
|
|
|
*/
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
static void free_pcppages_bulk(struct zone *zone, int count,
|
|
|
|
struct per_cpu_pages *pcp)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
int migratetype = 0;
|
2009-09-22 00:03:20 +00:00
|
|
|
int batch_free = 0;
|
2020-12-15 03:10:50 +00:00
|
|
|
int prefetch_nr = READ_ONCE(pcp->batch);
|
2016-05-20 00:13:58 +00:00
|
|
|
bool isolated_pageblocks;
|
2018-04-05 23:24:10 +00:00
|
|
|
struct page *page, *tmp;
|
|
|
|
LIST_HEAD(head);
|
2009-06-16 22:32:13 +00:00
|
|
|
|
mm, page_alloc: fix core hung in free_pcppages_bulk()
The following race is observed with the repeated online, offline and a
delay between two successive online of memory blocks of movable zone.
P1 P2
Online the first memory block in
the movable zone. The pcp struct
values are initialized to default
values,i.e., pcp->high = 0 &
pcp->batch = 1.
Allocate the pages from the
movable zone.
Try to Online the second memory
block in the movable zone thus it
entered the online_pages() but yet
to call zone_pcp_update().
This process is entered into
the exit path thus it tries
to release the order-0 pages
to pcp lists through
free_unref_page_commit().
As pcp->high = 0, pcp->count = 1
proceed to call the function
free_pcppages_bulk().
Update the pcp values thus the
new pcp values are like, say,
pcp->high = 378, pcp->batch = 63.
Read the pcp's batch value using
READ_ONCE() and pass the same to
free_pcppages_bulk(), pcp values
passed here are, batch = 63,
count = 1.
Since num of pages in the pcp
lists are less than ->batch,
then it will stuck in
while(list_empty(list)) loop
with interrupts disabled thus
a core hung.
Avoid this by ensuring free_pcppages_bulk() is called with proper count of
pcp list pages.
The mentioned race is some what easily reproducible without [1] because
pcp's are not updated for the first memory block online and thus there is
a enough race window for P2 between alloc+free and pcp struct values
update through onlining of second memory block.
With [1], the race still exists but it is very narrow as we update the pcp
struct values for the first memory block online itself.
This is not limited to the movable zone, it could also happen in cases
with the normal zone (e.g., hotplug to a node that only has DMA memory, or
no other memory yet).
[1]: https://patchwork.kernel.org/patch/11696389/
Fixes: 5f8dcc21211a ("page-allocator: split per-cpu list into one-list-per-migrate-type")
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: David Hildenbrand <david@redhat.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: <stable@vger.kernel.org> [2.6+]
Link: http://lkml.kernel.org/r/1597150703-19003-1-git-send-email-charante@codeaurora.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-21 00:42:27 +00:00
|
|
|
/*
|
|
|
|
* Ensure proper count is passed which otherwise would stuck in the
|
|
|
|
* below while (list_empty(list)) loop.
|
|
|
|
*/
|
|
|
|
count = min(pcp->count, count);
|
2016-05-20 00:14:24 +00:00
|
|
|
while (count) {
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
struct list_head *list;
|
|
|
|
|
|
|
|
/*
|
2009-09-22 00:03:20 +00:00
|
|
|
* Remove pages from lists in a round-robin fashion. A
|
|
|
|
* batch_free count is maintained that is incremented when an
|
|
|
|
* empty list is encountered. This is so more pages are freed
|
|
|
|
* off fuller lists instead of spinning excessively around empty
|
|
|
|
* lists
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
*/
|
|
|
|
do {
|
2009-09-22 00:03:20 +00:00
|
|
|
batch_free++;
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
if (++migratetype == MIGRATE_PCPTYPES)
|
|
|
|
migratetype = 0;
|
|
|
|
list = &pcp->lists[migratetype];
|
|
|
|
} while (list_empty(list));
|
2006-01-08 09:00:42 +00:00
|
|
|
|
2011-03-22 23:32:45 +00:00
|
|
|
/* This is the only non-empty list. Free them all. */
|
|
|
|
if (batch_free == MIGRATE_PCPTYPES)
|
2016-05-20 00:14:24 +00:00
|
|
|
batch_free = count;
|
2011-03-22 23:32:45 +00:00
|
|
|
|
2009-09-22 00:03:20 +00:00
|
|
|
do {
|
2016-01-14 23:20:30 +00:00
|
|
|
page = list_last_entry(list, struct page, lru);
|
2018-04-05 23:24:10 +00:00
|
|
|
/* must delete to avoid corrupting pcp list */
|
2009-09-22 00:03:20 +00:00
|
|
|
list_del(&page->lru);
|
2018-04-05 23:24:06 +00:00
|
|
|
pcp->count--;
|
mm, page_isolation: remove bogus tests for isolated pages
The __test_page_isolated_in_pageblock() is used to verify whether all
pages in pageblock were either successfully isolated, or are hwpoisoned.
Two of the possible state of pages, that are tested, are however bogus
and misleading.
Both tests rely on get_freepage_migratetype(page), which however has no
guarantees about pages on freelists. Specifically, it doesn't guarantee
that the migratetype returned by the function actually matches the
migratetype of the freelist that the page is on. Such guarantee is not
its purpose and would have negative impact on allocator performance.
The first test checks whether the freepage_migratetype equals
MIGRATE_ISOLATE, supposedly to catch races between page isolation and
allocator activity. These races should be fixed nowadays with
51bb1a4093 ("mm/page_alloc: add freepage on isolate pageblock to correct
buddy list") and related patches. As explained above, the check
wouldn't be able to catch them reliably anyway. For the same reason
false positives can happen, although they are harmless, as the
move_freepages() call would just move the page to the same freelist it's
already on. So removing the test is not a bug fix, just cleanup. After
this patch, we assume that all PageBuddy pages are on the correct
freelist and that the races were really fixed. A truly reliable
verification in the form of e.g. VM_BUG_ON() would be complicated and
is arguably not needed.
The second test (page_count(page) == 0 && get_freepage_migratetype(page)
== MIGRATE_ISOLATE) is probably supposed (the code comes from a big
memory isolation patch from 2007) to catch pages on MIGRATE_ISOLATE
pcplists. However, pcplists don't contain MIGRATE_ISOLATE freepages
nowadays, those are freed directly to free lists, so the check is
obsolete. Remove it as well.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Minchan Kim <minchan@kernel.org>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: Laura Abbott <lauraa@codeaurora.org>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Seungho Park <seungho1.park@lge.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 22:01:22 +00:00
|
|
|
|
2016-05-20 00:14:32 +00:00
|
|
|
if (bulkfree_pcp_prepare(page))
|
|
|
|
continue;
|
|
|
|
|
2018-04-05 23:24:10 +00:00
|
|
|
list_add_tail(&page->lru, &head);
|
mm/free_pcppages_bulk: prefetch buddy while not holding lock
When a page is freed back to the global pool, its buddy will be checked
to see if it's possible to do a merge. This requires accessing buddy's
page structure and that access could take a long time if it's cache
cold.
This patch adds a prefetch to the to-be-freed page's buddy outside of
zone->lock in hope of accessing buddy's page structure later under
zone->lock will be faster. Since we *always* do buddy merging and check
an order-0 page's buddy to try to merge it when it goes into the main
allocator, the cacheline will always come in, i.e. the prefetched data
will never be unused.
Normally, the number of prefetch will be pcp->batch(default=31 and has
an upper limit of (PAGE_SHIFT * 8)=96 on x86_64) but in the case of
pcp's pages get all drained, it will be pcp->count which has an upper
limit of pcp->high. pcp->high, although has a default value of 186
(pcp->batch=31 * 6), can be changed by user through
/proc/sys/vm/percpu_pagelist_fraction and there is no software upper
limit so could be large, like several thousand. For this reason, only
the first pcp->batch number of page's buddy structure is prefetched to
avoid excessive prefetching.
In the meantime, there are two concerns:
1. the prefetch could potentially evict existing cachelines, especially
for L1D cache since it is not huge
2. there is some additional instruction overhead, namely calculating
buddy pfn twice
For 1, it's hard to say, this microbenchmark though shows good result
but the actual benefit of this patch will be workload/CPU dependant;
For 2, since the calculation is a XOR on two local variables, it's
expected in many cases that cycles spent will be offset by reduced
memory latency later. This is especially true for NUMA machines where
multiple CPUs are contending on zone->lock and the most time consuming
part under zone->lock is the wait of 'struct page' cacheline of the
to-be-freed pages and their buddies.
Test with will-it-scale/page_fault1 full load:
kernel Broadwell(2S) Skylake(2S) Broadwell(4S) Skylake(4S)
v4.16-rc2+ 9034215 7971818 13667135 15677465
patch2/3 9536374 +5.6% 8314710 +4.3% 14070408 +3.0% 16675866 +6.4%
this patch 10180856 +6.8% 8506369 +2.3% 14756865 +4.9% 17325324 +3.9%
Note: this patch's performance improvement percent is against patch2/3.
(Changelog stolen from Dave Hansen and Mel Gorman's comments at
http://lkml.kernel.org/r/148a42d8-8306-2f2f-7f7c-86bc118f8ccd@intel.com)
[aaron.lu@intel.com: use helper function, avoid disordering pages]
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Link: http://lkml.kernel.org/r/20180320113146.GB24737@intel.com
[aaron.lu@intel.com: v4]
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Link: http://lkml.kernel.org/r/20180309082431.GB30868@intel.com
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Suggested-by: Ying Huang <ying.huang@intel.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:24:14 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We are going to put the page back to the global
|
|
|
|
* pool, prefetch its buddy to speed up later access
|
|
|
|
* under zone->lock. It is believed the overhead of
|
|
|
|
* an additional test and calculating buddy_pfn here
|
|
|
|
* can be offset by reduced memory latency later. To
|
|
|
|
* avoid excessive prefetching due to large count, only
|
|
|
|
* prefetch buddy for the first pcp->batch nr of pages.
|
|
|
|
*/
|
2020-12-15 03:10:50 +00:00
|
|
|
if (prefetch_nr) {
|
mm/free_pcppages_bulk: prefetch buddy while not holding lock
When a page is freed back to the global pool, its buddy will be checked
to see if it's possible to do a merge. This requires accessing buddy's
page structure and that access could take a long time if it's cache
cold.
This patch adds a prefetch to the to-be-freed page's buddy outside of
zone->lock in hope of accessing buddy's page structure later under
zone->lock will be faster. Since we *always* do buddy merging and check
an order-0 page's buddy to try to merge it when it goes into the main
allocator, the cacheline will always come in, i.e. the prefetched data
will never be unused.
Normally, the number of prefetch will be pcp->batch(default=31 and has
an upper limit of (PAGE_SHIFT * 8)=96 on x86_64) but in the case of
pcp's pages get all drained, it will be pcp->count which has an upper
limit of pcp->high. pcp->high, although has a default value of 186
(pcp->batch=31 * 6), can be changed by user through
/proc/sys/vm/percpu_pagelist_fraction and there is no software upper
limit so could be large, like several thousand. For this reason, only
the first pcp->batch number of page's buddy structure is prefetched to
avoid excessive prefetching.
In the meantime, there are two concerns:
1. the prefetch could potentially evict existing cachelines, especially
for L1D cache since it is not huge
2. there is some additional instruction overhead, namely calculating
buddy pfn twice
For 1, it's hard to say, this microbenchmark though shows good result
but the actual benefit of this patch will be workload/CPU dependant;
For 2, since the calculation is a XOR on two local variables, it's
expected in many cases that cycles spent will be offset by reduced
memory latency later. This is especially true for NUMA machines where
multiple CPUs are contending on zone->lock and the most time consuming
part under zone->lock is the wait of 'struct page' cacheline of the
to-be-freed pages and their buddies.
Test with will-it-scale/page_fault1 full load:
kernel Broadwell(2S) Skylake(2S) Broadwell(4S) Skylake(4S)
v4.16-rc2+ 9034215 7971818 13667135 15677465
patch2/3 9536374 +5.6% 8314710 +4.3% 14070408 +3.0% 16675866 +6.4%
this patch 10180856 +6.8% 8506369 +2.3% 14756865 +4.9% 17325324 +3.9%
Note: this patch's performance improvement percent is against patch2/3.
(Changelog stolen from Dave Hansen and Mel Gorman's comments at
http://lkml.kernel.org/r/148a42d8-8306-2f2f-7f7c-86bc118f8ccd@intel.com)
[aaron.lu@intel.com: use helper function, avoid disordering pages]
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Link: http://lkml.kernel.org/r/20180320113146.GB24737@intel.com
[aaron.lu@intel.com: v4]
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Link: http://lkml.kernel.org/r/20180309082431.GB30868@intel.com
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Suggested-by: Ying Huang <ying.huang@intel.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:24:14 +00:00
|
|
|
prefetch_buddy(page);
|
2020-12-15 03:10:50 +00:00
|
|
|
prefetch_nr--;
|
|
|
|
}
|
2016-05-20 00:14:24 +00:00
|
|
|
} while (--count && --batch_free && !list_empty(list));
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2018-04-05 23:24:10 +00:00
|
|
|
|
|
|
|
spin_lock(&zone->lock);
|
|
|
|
isolated_pageblocks = has_isolate_pageblock(zone);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Use safe version since after __free_one_page(),
|
|
|
|
* page->lru.next will not point to original list.
|
|
|
|
*/
|
|
|
|
list_for_each_entry_safe(page, tmp, &head, lru) {
|
|
|
|
int mt = get_pcppage_migratetype(page);
|
|
|
|
/* MIGRATE_ISOLATE page should not go to pcplists */
|
|
|
|
VM_BUG_ON_PAGE(is_migrate_isolate(mt), page);
|
|
|
|
/* Pageblock could have been isolated meanwhile */
|
|
|
|
if (unlikely(isolated_pageblocks))
|
|
|
|
mt = get_pageblock_migratetype(page);
|
|
|
|
|
mm/page_alloc: convert "report" flag of __free_one_page() to a proper flag
Patch series "mm: place pages to the freelist tail when onlining and undoing isolation", v2.
When adding separate memory blocks via add_memory*() and onlining them
immediately, the metadata (especially the memmap) of the next block will
be placed onto one of the just added+onlined block. This creates a chain
of unmovable allocations: If the last memory block cannot get
offlined+removed() so will all dependent ones. We directly have unmovable
allocations all over the place.
This can be observed quite easily using virtio-mem, however, it can also
be observed when using DIMMs. The freshly onlined pages will usually be
placed to the head of the freelists, meaning they will be allocated next,
turning the just-added memory usually immediately un-removable. The fresh
pages are cold, prefering to allocate others (that might be hot) also
feels to be the natural thing to do.
It also applies to the hyper-v balloon xen-balloon, and ppc64 dlpar: when
adding separate, successive memory blocks, each memory block will have
unmovable allocations on them - for example gigantic pages will fail to
allocate.
While the ZONE_NORMAL doesn't provide any guarantees that memory can get
offlined+removed again (any kind of fragmentation with unmovable
allocations is possible), there are many scenarios (hotplugging a lot of
memory, running workload, hotunplug some memory/as much as possible) where
we can offline+remove quite a lot with this patchset.
a) To visualize the problem, a very simple example:
Start a VM with 4GB and 8GB of virtio-mem memory:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000033fffffff 9G online yes 32-103
Memory block size: 128M
Total online memory: 12G
Total offline memory: 0B
Then try to unplug as much as possible using virtio-mem. Observe which
memory blocks are still around. Without this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
0x0000000148000000-0x000000014fffffff 128M online yes 41
0x0000000158000000-0x000000015fffffff 128M online yes 43
0x0000000168000000-0x000000016fffffff 128M online yes 45
0x0000000178000000-0x000000017fffffff 128M online yes 47
0x0000000188000000-0x0000000197ffffff 256M online yes 49-50
0x00000001a0000000-0x00000001a7ffffff 128M online yes 52
0x00000001b0000000-0x00000001b7ffffff 128M online yes 54
0x00000001c0000000-0x00000001c7ffffff 128M online yes 56
0x00000001d0000000-0x00000001d7ffffff 128M online yes 58
0x00000001e0000000-0x00000001e7ffffff 128M online yes 60
0x00000001f0000000-0x00000001f7ffffff 128M online yes 62
0x0000000200000000-0x0000000207ffffff 128M online yes 64
0x0000000210000000-0x0000000217ffffff 128M online yes 66
0x0000000220000000-0x0000000227ffffff 128M online yes 68
0x0000000230000000-0x0000000237ffffff 128M online yes 70
0x0000000240000000-0x0000000247ffffff 128M online yes 72
0x0000000250000000-0x0000000257ffffff 128M online yes 74
0x0000000260000000-0x0000000267ffffff 128M online yes 76
0x0000000270000000-0x0000000277ffffff 128M online yes 78
0x0000000280000000-0x0000000287ffffff 128M online yes 80
0x0000000290000000-0x0000000297ffffff 128M online yes 82
0x00000002a0000000-0x00000002a7ffffff 128M online yes 84
0x00000002b0000000-0x00000002b7ffffff 128M online yes 86
0x00000002c0000000-0x00000002c7ffffff 128M online yes 88
0x00000002d0000000-0x00000002d7ffffff 128M online yes 90
0x00000002e0000000-0x00000002e7ffffff 128M online yes 92
0x00000002f0000000-0x00000002f7ffffff 128M online yes 94
0x0000000300000000-0x0000000307ffffff 128M online yes 96
0x0000000310000000-0x0000000317ffffff 128M online yes 98
0x0000000320000000-0x0000000327ffffff 128M online yes 100
0x0000000330000000-0x000000033fffffff 256M online yes 102-103
Memory block size: 128M
Total online memory: 8.1G
Total offline memory: 0B
With this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
Memory block size: 128M
Total online memory: 4G
Total offline memory: 0B
All memory can get unplugged, all memory block can get removed. Of
course, no workload ran and the system was basically idle, but it
highlights the issue - the fairly deterministic chain of unmovable
allocations. When a huge page for the 2MB memmap is needed, a
just-onlined 4MB page will be split. The remaining 2MB page will be used
for the memmap of the next memory block. So one memory block will hold
the memmap of the two following memory blocks. Finally the pages of the
last-onlined memory block will get used for the next bigger allocations -
if any allocation is unmovable, all dependent memory blocks cannot get
unplugged and removed until that allocation is gone.
Note that with bigger memory blocks (e.g., 256MB), *all* memory
blocks are dependent and none can get unplugged again!
b) Experiment with memory intensive workload
I performed an experiment with an older version of this patch set (before
we used undo_isolate_page_range() in online_pages(): Hotplug 56GB to a VM
with an initial 4GB, onlining all memory to ZONE_NORMAL right from the
kernel when adding it. I then run various memory intensive workloads that
consume most system memory for a total of 45 minutes. Once finished, I
try to unplug as much memory as possible.
With this change, I am able to remove via virtio-mem (adding individual
128MB memory blocks) 413 out of 448 added memory blocks. Via individual
(256MB) DIMMs 380 out of 448 added memory blocks. (I don't have any
numbers without this patchset, but looking at the above example, it's at
most half of the 448 memory blocks for virtio-mem, and most probably none
for DIMMs).
Again, there are workloads that might behave very differently due to the
nature of ZONE_NORMAL.
This change also affects (besides memory onlining):
- Other users of undo_isolate_page_range(): Pages are always placed to the
tail.
-- When memory offlining fails
-- When memory isolation fails after having isolated some pageblocks
-- When alloc_contig_range() either succeeds or fails
- Other users of __putback_isolated_page(): Pages are always placed to the
tail.
-- Free page reporting
- Other users of __free_pages_core()
-- AFAIKs, any memory that is getting exposed to the buddy during boot.
IIUC we will now usually allocate memory from lower addresses within
a zone first (especially during boot).
- Other users of generic_online_page()
-- Hyper-V balloon
This patch (of 5):
Let's prepare for additional flags and avoid long parameter lists of
bools. Follow-up patches will also make use of the flags in
__free_pages_ok().
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-1-david@redhat.com
Link: https://lkml.kernel.org/r/20201005121534.15649-2-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:20 +00:00
|
|
|
__free_one_page(page, page_to_pfn(page), zone, 0, mt, FPI_NONE);
|
2018-04-05 23:24:10 +00:00
|
|
|
trace_mm_page_pcpu_drain(page, 0, mt);
|
|
|
|
}
|
2017-04-20 21:37:43 +00:00
|
|
|
spin_unlock(&zone->lock);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2014-06-04 23:10:17 +00:00
|
|
|
static void free_one_page(struct zone *zone,
|
|
|
|
struct page *page, unsigned long pfn,
|
2014-06-04 23:10:21 +00:00
|
|
|
unsigned int order,
|
2020-10-16 03:09:35 +00:00
|
|
|
int migratetype, fpi_t fpi_flags)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2017-04-20 21:37:43 +00:00
|
|
|
spin_lock(&zone->lock);
|
mm/page_alloc: fix incorrect isolation behavior by rechecking migratetype
Before describing bugs itself, I first explain definition of freepage.
1. pages on buddy list are counted as freepage.
2. pages on isolate migratetype buddy list are *not* counted as freepage.
3. pages on cma buddy list are counted as CMA freepage, too.
Now, I describe problems and related patch.
Patch 1: There is race conditions on getting pageblock migratetype that
it results in misplacement of freepages on buddy list, incorrect
freepage count and un-availability of freepage.
Patch 2: Freepages on pcp list could have stale cached information to
determine migratetype of buddy list to go. This causes misplacement of
freepages on buddy list and incorrect freepage count.
Patch 4: Merging between freepages on different migratetype of
pageblocks will cause freepages accouting problem. This patch fixes it.
Without patchset [3], above problem doesn't happens on my CMA allocation
test, because CMA reserved pages aren't used at all. So there is no
chance for above race.
With patchset [3], I did simple CMA allocation test and get below
result:
- Virtual machine, 4 cpus, 1024 MB memory, 256 MB CMA reservation
- run kernel build (make -j16) on background
- 30 times CMA allocation(8MB * 30 = 240MB) attempts in 5 sec interval
- Result: more than 5000 freepage count are missed
With patchset [3] and this patchset, I found that no freepage count are
missed so that I conclude that problems are solved.
On my simple memory offlining test, these problems also occur on that
environment, too.
This patch (of 4):
There are two paths to reach core free function of buddy allocator,
__free_one_page(), one is free_one_page()->__free_one_page() and the
other is free_hot_cold_page()->free_pcppages_bulk()->__free_one_page().
Each paths has race condition causing serious problems. At first, this
patch is focused on first type of freepath. And then, following patch
will solve the problem in second type of freepath.
In the first type of freepath, we got migratetype of freeing page
without holding the zone lock, so it could be racy. There are two cases
of this race.
1. pages are added to isolate buddy list after restoring orignal
migratetype
CPU1 CPU2
get migratetype => return MIGRATE_ISOLATE
call free_one_page() with MIGRATE_ISOLATE
grab the zone lock
unisolate pageblock
release the zone lock
grab the zone lock
call __free_one_page() with MIGRATE_ISOLATE
freepage go into isolate buddy list,
although pageblock is already unisolated
This may cause two problems. One is that we can't use this page anymore
until next isolation attempt of this pageblock, because freepage is on
isolate buddy list. The other is that freepage accouting could be wrong
due to merging between different buddy list. Freepages on isolate buddy
list aren't counted as freepage, but ones on normal buddy list are
counted as freepage. If merge happens, buddy freepage on normal buddy
list is inevitably moved to isolate buddy list without any consideration
of freepage accouting so it could be incorrect.
2. pages are added to normal buddy list while pageblock is isolated.
It is similar with above case.
This also may cause two problems. One is that we can't keep these
freepages from being allocated. Although this pageblock is isolated,
freepage would be added to normal buddy list so that it could be
allocated without any restriction. And the other problem is same as
case 1, that it, incorrect freepage accouting.
This race condition would be prevented by checking migratetype again
with holding the zone lock. Because it is somewhat heavy operation and
it isn't needed in common case, we want to avoid rechecking as much as
possible. So this patch introduce new variable, nr_isolate_pageblock in
struct zone to check if there is isolated pageblock. With this, we can
avoid to re-check migratetype in common case and do it only if there is
isolated pageblock or migratetype is MIGRATE_ISOLATE. This solve above
mentioned problems.
Changes from v3:
Add one more check in free_one_page() that checks whether migratetype is
MIGRATE_ISOLATE or not. Without this, abovementioned case 1 could happens.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Laura Abbott <lauraa@codeaurora.org>
Cc: Heesub Shin <heesub.shin@samsung.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Ritesh Harjani <ritesh.list@gmail.com>
Cc: Gioh Kim <gioh.kim@lge.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-11-13 23:19:11 +00:00
|
|
|
if (unlikely(has_isolate_pageblock(zone) ||
|
|
|
|
is_migrate_isolate(migratetype))) {
|
|
|
|
migratetype = get_pfnblock_migratetype(page, pfn);
|
|
|
|
}
|
2020-10-16 03:09:35 +00:00
|
|
|
__free_one_page(page, pfn, zone, order, migratetype, fpi_flags);
|
2017-04-20 21:37:43 +00:00
|
|
|
spin_unlock(&zone->lock);
|
2006-01-08 09:00:42 +00:00
|
|
|
}
|
|
|
|
|
2015-06-30 21:56:45 +00:00
|
|
|
static void __meminit __init_single_page(struct page *page, unsigned long pfn,
|
2018-04-05 23:23:00 +00:00
|
|
|
unsigned long zone, int nid)
|
2015-06-30 21:56:45 +00:00
|
|
|
{
|
2018-04-05 23:23:00 +00:00
|
|
|
mm_zero_struct_page(page);
|
2015-06-30 21:56:45 +00:00
|
|
|
set_page_links(page, zone, nid, pfn);
|
|
|
|
init_page_count(page);
|
|
|
|
page_mapcount_reset(page);
|
|
|
|
page_cpupid_reset_last(page);
|
2018-12-28 08:30:57 +00:00
|
|
|
page_kasan_tag_reset(page);
|
2015-06-30 21:56:45 +00:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&page->lru);
|
|
|
|
#ifdef WANT_PAGE_VIRTUAL
|
|
|
|
/* The shift won't overflow because ZONE_NORMAL is below 4G. */
|
|
|
|
if (!is_highmem_idx(zone))
|
|
|
|
set_page_address(page, __va(pfn << PAGE_SHIFT));
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2015-06-30 21:57:05 +00:00
|
|
|
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
|
2017-10-03 23:15:10 +00:00
|
|
|
static void __meminit init_reserved_page(unsigned long pfn)
|
2015-06-30 21:57:05 +00:00
|
|
|
{
|
|
|
|
pg_data_t *pgdat;
|
|
|
|
int nid, zid;
|
|
|
|
|
|
|
|
if (!early_page_uninitialised(pfn))
|
|
|
|
return;
|
|
|
|
|
|
|
|
nid = early_pfn_to_nid(pfn);
|
|
|
|
pgdat = NODE_DATA(nid);
|
|
|
|
|
|
|
|
for (zid = 0; zid < MAX_NR_ZONES; zid++) {
|
|
|
|
struct zone *zone = &pgdat->node_zones[zid];
|
|
|
|
|
|
|
|
if (pfn >= zone->zone_start_pfn && pfn < zone_end_pfn(zone))
|
|
|
|
break;
|
|
|
|
}
|
2018-04-05 23:23:00 +00:00
|
|
|
__init_single_page(pfn_to_page(pfn), pfn, zid, nid);
|
2015-06-30 21:57:05 +00:00
|
|
|
}
|
|
|
|
#else
|
|
|
|
static inline void init_reserved_page(unsigned long pfn)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
|
|
|
|
|
2015-06-30 21:56:48 +00:00
|
|
|
/*
|
|
|
|
* Initialised pages do not have PageReserved set. This function is
|
|
|
|
* called for each range allocated by the bootmem allocator and
|
|
|
|
* marks the pages PageReserved. The remaining valid pages are later
|
|
|
|
* sent to the buddy page allocator.
|
|
|
|
*/
|
2016-05-20 23:58:38 +00:00
|
|
|
void __meminit reserve_bootmem_region(phys_addr_t start, phys_addr_t end)
|
2015-06-30 21:56:48 +00:00
|
|
|
{
|
|
|
|
unsigned long start_pfn = PFN_DOWN(start);
|
|
|
|
unsigned long end_pfn = PFN_UP(end);
|
|
|
|
|
2015-06-30 21:57:05 +00:00
|
|
|
for (; start_pfn < end_pfn; start_pfn++) {
|
|
|
|
if (pfn_valid(start_pfn)) {
|
|
|
|
struct page *page = pfn_to_page(start_pfn);
|
|
|
|
|
|
|
|
init_reserved_page(start_pfn);
|
2015-11-07 00:29:54 +00:00
|
|
|
|
|
|
|
/* Avoid false-positive PageTail() */
|
|
|
|
INIT_LIST_HEAD(&page->lru);
|
|
|
|
|
2018-10-26 22:07:48 +00:00
|
|
|
/*
|
|
|
|
* no need for atomic set_bit because the struct
|
|
|
|
* page is not visible yet so nobody should
|
|
|
|
* access it yet.
|
|
|
|
*/
|
|
|
|
__SetPageReserved(page);
|
2015-06-30 21:57:05 +00:00
|
|
|
}
|
|
|
|
}
|
2015-06-30 21:56:48 +00:00
|
|
|
}
|
|
|
|
|
2020-10-16 03:09:35 +00:00
|
|
|
static void __free_pages_ok(struct page *page, unsigned int order,
|
|
|
|
fpi_t fpi_flags)
|
2010-05-24 21:32:38 +00:00
|
|
|
{
|
2017-04-20 21:37:43 +00:00
|
|
|
unsigned long flags;
|
2012-10-08 23:32:11 +00:00
|
|
|
int migratetype;
|
2014-06-04 23:10:17 +00:00
|
|
|
unsigned long pfn = page_to_pfn(page);
|
2010-05-24 21:32:38 +00:00
|
|
|
|
2016-05-20 00:14:38 +00:00
|
|
|
if (!free_pages_prepare(page, order, true))
|
2010-05-24 21:32:38 +00:00
|
|
|
return;
|
|
|
|
|
2014-06-04 23:10:19 +00:00
|
|
|
migratetype = get_pfnblock_migratetype(page, pfn);
|
2017-04-20 21:37:43 +00:00
|
|
|
local_irq_save(flags);
|
|
|
|
__count_vm_events(PGFREE, 1 << order);
|
2020-10-16 03:09:35 +00:00
|
|
|
free_one_page(page_zone(page), page, pfn, order, migratetype,
|
|
|
|
fpi_flags);
|
2017-04-20 21:37:43 +00:00
|
|
|
local_irq_restore(flags);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2019-03-05 23:42:14 +00:00
|
|
|
void __free_pages_core(struct page *page, unsigned int order)
|
2006-01-06 08:11:08 +00:00
|
|
|
{
|
2012-01-10 23:08:10 +00:00
|
|
|
unsigned int nr_pages = 1 << order;
|
2013-09-11 21:20:37 +00:00
|
|
|
struct page *p = page;
|
2012-01-10 23:08:10 +00:00
|
|
|
unsigned int loop;
|
2006-01-06 08:11:08 +00:00
|
|
|
|
2020-10-16 03:09:35 +00:00
|
|
|
/*
|
|
|
|
* When initializing the memmap, __init_single_page() sets the refcount
|
|
|
|
* of all pages to 1 ("allocated"/"not free"). We have to set the
|
|
|
|
* refcount of all involved pages to 0.
|
|
|
|
*/
|
2013-09-11 21:20:37 +00:00
|
|
|
prefetchw(p);
|
|
|
|
for (loop = 0; loop < (nr_pages - 1); loop++, p++) {
|
|
|
|
prefetchw(p + 1);
|
2012-01-10 23:08:10 +00:00
|
|
|
__ClearPageReserved(p);
|
|
|
|
set_page_count(p, 0);
|
2006-01-06 08:11:08 +00:00
|
|
|
}
|
2013-09-11 21:20:37 +00:00
|
|
|
__ClearPageReserved(p);
|
|
|
|
set_page_count(p, 0);
|
2012-01-10 23:08:10 +00:00
|
|
|
|
2018-12-28 08:34:24 +00:00
|
|
|
atomic_long_add(nr_pages, &page_zone(page)->managed_pages);
|
2020-10-16 03:09:35 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Bypass PCP and place fresh pages right to the tail, primarily
|
|
|
|
* relevant for memory onlining.
|
|
|
|
*/
|
|
|
|
__free_pages_ok(page, order, FPI_TO_TAIL);
|
2006-01-06 08:11:08 +00:00
|
|
|
}
|
|
|
|
|
2020-06-03 22:57:02 +00:00
|
|
|
#ifdef CONFIG_NEED_MULTIPLE_NODES
|
2015-08-06 22:46:13 +00:00
|
|
|
|
2020-12-15 03:09:32 +00:00
|
|
|
/*
|
|
|
|
* During memory init memblocks map pfns to nids. The search is expensive and
|
|
|
|
* this caches recent lookups. The implementation of __early_pfn_to_nid
|
|
|
|
* treats start/end as pfns.
|
|
|
|
*/
|
|
|
|
struct mminit_pfnnid_cache {
|
|
|
|
unsigned long last_start;
|
|
|
|
unsigned long last_end;
|
|
|
|
int last_nid;
|
|
|
|
};
|
2015-06-30 21:56:59 +00:00
|
|
|
|
2020-12-15 03:09:32 +00:00
|
|
|
static struct mminit_pfnnid_cache early_pfnnid_cache __meminitdata;
|
2020-06-03 22:56:57 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Required by SPARSEMEM. Given a PFN, return what node the PFN is on.
|
|
|
|
*/
|
2020-12-15 03:09:32 +00:00
|
|
|
static int __meminit __early_pfn_to_nid(unsigned long pfn,
|
2020-06-03 22:56:57 +00:00
|
|
|
struct mminit_pfnnid_cache *state)
|
2015-06-30 21:56:59 +00:00
|
|
|
{
|
2020-06-03 22:56:57 +00:00
|
|
|
unsigned long start_pfn, end_pfn;
|
2015-06-30 21:56:59 +00:00
|
|
|
int nid;
|
|
|
|
|
2020-06-03 22:56:57 +00:00
|
|
|
if (state->last_start <= pfn && pfn < state->last_end)
|
|
|
|
return state->last_nid;
|
|
|
|
|
|
|
|
nid = memblock_search_pfn_nid(pfn, &start_pfn, &end_pfn);
|
|
|
|
if (nid != NUMA_NO_NODE) {
|
|
|
|
state->last_start = start_pfn;
|
|
|
|
state->last_end = end_pfn;
|
|
|
|
state->last_nid = nid;
|
|
|
|
}
|
2015-08-06 22:46:13 +00:00
|
|
|
|
|
|
|
return nid;
|
2015-06-30 21:56:59 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int __meminit early_pfn_to_nid(unsigned long pfn)
|
|
|
|
{
|
2015-08-06 22:46:13 +00:00
|
|
|
static DEFINE_SPINLOCK(early_pfn_lock);
|
2015-06-30 21:56:59 +00:00
|
|
|
int nid;
|
|
|
|
|
2015-08-06 22:46:13 +00:00
|
|
|
spin_lock(&early_pfn_lock);
|
2019-05-14 00:21:13 +00:00
|
|
|
nid = __early_pfn_to_nid(pfn, &early_pfnnid_cache);
|
2015-08-06 22:46:13 +00:00
|
|
|
if (nid < 0)
|
2016-07-14 19:07:20 +00:00
|
|
|
nid = first_online_node;
|
2015-08-06 22:46:13 +00:00
|
|
|
spin_unlock(&early_pfn_lock);
|
2015-06-30 21:56:59 +00:00
|
|
|
|
2015-08-06 22:46:13 +00:00
|
|
|
return nid;
|
2015-06-30 21:56:59 +00:00
|
|
|
}
|
2020-06-03 22:57:02 +00:00
|
|
|
#endif /* CONFIG_NEED_MULTIPLE_NODES */
|
2015-06-30 21:56:59 +00:00
|
|
|
|
2018-10-30 22:09:36 +00:00
|
|
|
void __init memblock_free_pages(struct page *page, unsigned long pfn,
|
2015-06-30 21:57:02 +00:00
|
|
|
unsigned int order)
|
|
|
|
{
|
|
|
|
if (early_page_uninitialised(pfn))
|
|
|
|
return;
|
2019-03-05 23:42:14 +00:00
|
|
|
__free_pages_core(page, order);
|
2015-06-30 21:57:02 +00:00
|
|
|
}
|
|
|
|
|
2016-03-15 21:57:51 +00:00
|
|
|
/*
|
|
|
|
* Check that the whole (or subset of) a pageblock given by the interval of
|
|
|
|
* [start_pfn, end_pfn) is valid and within the same zone, before scanning it
|
|
|
|
* with the migration of free compaction scanner. The scanners then need to
|
|
|
|
* use only pfn_valid_within() check for arches that allow holes within
|
|
|
|
* pageblocks.
|
|
|
|
*
|
|
|
|
* Return struct page pointer of start_pfn, or NULL if checks were not passed.
|
|
|
|
*
|
|
|
|
* It's possible on some configurations to have a setup like node0 node1 node0
|
|
|
|
* i.e. it's possible that all pages within a zones range of pages do not
|
|
|
|
* belong to a single zone. We assume that a border between node0 and node1
|
|
|
|
* can occur within a single pageblock, but not a node0 node1 node0
|
|
|
|
* interleaving within a single pageblock. It is therefore sufficient to check
|
|
|
|
* the first and last page of a pageblock and avoid checking each individual
|
|
|
|
* page in a pageblock.
|
|
|
|
*/
|
|
|
|
struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
|
|
|
|
unsigned long end_pfn, struct zone *zone)
|
|
|
|
{
|
|
|
|
struct page *start_page;
|
|
|
|
struct page *end_page;
|
|
|
|
|
|
|
|
/* end_pfn is one past the range we are checking */
|
|
|
|
end_pfn--;
|
|
|
|
|
|
|
|
if (!pfn_valid(start_pfn) || !pfn_valid(end_pfn))
|
|
|
|
return NULL;
|
|
|
|
|
2017-07-06 22:37:56 +00:00
|
|
|
start_page = pfn_to_online_page(start_pfn);
|
|
|
|
if (!start_page)
|
|
|
|
return NULL;
|
2016-03-15 21:57:51 +00:00
|
|
|
|
|
|
|
if (page_zone(start_page) != zone)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
end_page = pfn_to_page(end_pfn);
|
|
|
|
|
|
|
|
/* This gives a shorter code than deriving page_zone(end_page) */
|
|
|
|
if (page_zone_id(start_page) != page_zone_id(end_page))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
return start_page;
|
|
|
|
}
|
|
|
|
|
|
|
|
void set_zone_contiguous(struct zone *zone)
|
|
|
|
{
|
|
|
|
unsigned long block_start_pfn = zone->zone_start_pfn;
|
|
|
|
unsigned long block_end_pfn;
|
|
|
|
|
|
|
|
block_end_pfn = ALIGN(block_start_pfn + 1, pageblock_nr_pages);
|
|
|
|
for (; block_start_pfn < zone_end_pfn(zone);
|
|
|
|
block_start_pfn = block_end_pfn,
|
|
|
|
block_end_pfn += pageblock_nr_pages) {
|
|
|
|
|
|
|
|
block_end_pfn = min(block_end_pfn, zone_end_pfn(zone));
|
|
|
|
|
|
|
|
if (!__pageblock_pfn_to_page(block_start_pfn,
|
|
|
|
block_end_pfn, zone))
|
|
|
|
return;
|
2020-05-08 01:35:46 +00:00
|
|
|
cond_resched();
|
2016-03-15 21:57:51 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* We confirm that there is no hole */
|
|
|
|
zone->contiguous = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
void clear_zone_contiguous(struct zone *zone)
|
|
|
|
{
|
|
|
|
zone->contiguous = false;
|
|
|
|
}
|
|
|
|
|
2015-06-30 21:57:05 +00:00
|
|
|
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
static void __init deferred_free_range(unsigned long pfn,
|
|
|
|
unsigned long nr_pages)
|
2015-06-30 21:57:16 +00:00
|
|
|
{
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
struct page *page;
|
|
|
|
unsigned long i;
|
2015-06-30 21:57:16 +00:00
|
|
|
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
if (!nr_pages)
|
2015-06-30 21:57:16 +00:00
|
|
|
return;
|
|
|
|
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
page = pfn_to_page(pfn);
|
|
|
|
|
2015-06-30 21:57:16 +00:00
|
|
|
/* Free a large naturally-aligned chunk if possible */
|
2016-10-07 23:58:09 +00:00
|
|
|
if (nr_pages == pageblock_nr_pages &&
|
|
|
|
(pfn & (pageblock_nr_pages - 1)) == 0) {
|
2015-06-30 21:57:20 +00:00
|
|
|
set_pageblock_migratetype(page, MIGRATE_MOVABLE);
|
2019-03-05 23:42:14 +00:00
|
|
|
__free_pages_core(page, pageblock_order);
|
2015-06-30 21:57:16 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2016-10-07 23:58:09 +00:00
|
|
|
for (i = 0; i < nr_pages; i++, page++, pfn++) {
|
|
|
|
if ((pfn & (pageblock_nr_pages - 1)) == 0)
|
|
|
|
set_pageblock_migratetype(page, MIGRATE_MOVABLE);
|
2019-03-05 23:42:14 +00:00
|
|
|
__free_pages_core(page, 0);
|
2016-10-07 23:58:09 +00:00
|
|
|
}
|
2015-06-30 21:57:16 +00:00
|
|
|
}
|
|
|
|
|
2015-08-06 22:46:16 +00:00
|
|
|
/* Completion tracking for deferred_init_memmap() threads */
|
|
|
|
static atomic_t pgdat_init_n_undone __initdata;
|
|
|
|
static __initdata DECLARE_COMPLETION(pgdat_init_all_done_comp);
|
|
|
|
|
|
|
|
static inline void __init pgdat_init_report_one_done(void)
|
|
|
|
{
|
|
|
|
if (atomic_dec_and_test(&pgdat_init_n_undone))
|
|
|
|
complete(&pgdat_init_all_done_comp);
|
|
|
|
}
|
2015-06-30 21:57:27 +00:00
|
|
|
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
/*
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
* Returns true if page needs to be initialized or freed to buddy allocator.
|
|
|
|
*
|
|
|
|
* First we check if pfn is valid on architectures where it is possible to have
|
|
|
|
* holes within pageblock_nr_pages. On systems where it is not possible, this
|
|
|
|
* function is optimized out.
|
|
|
|
*
|
|
|
|
* Then, we check if a current large page is valid by only checking the validity
|
|
|
|
* of the head pfn.
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
*/
|
2019-05-14 00:21:13 +00:00
|
|
|
static inline bool __init deferred_pfn_valid(unsigned long pfn)
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
{
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
if (!pfn_valid_within(pfn))
|
|
|
|
return false;
|
|
|
|
if (!(pfn & (pageblock_nr_pages - 1)) && !pfn_valid(pfn))
|
|
|
|
return false;
|
|
|
|
return true;
|
|
|
|
}
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
/*
|
|
|
|
* Free pages to buddy allocator. Try to free aligned pages in
|
|
|
|
* pageblock_nr_pages sizes.
|
|
|
|
*/
|
2019-05-14 00:21:13 +00:00
|
|
|
static void __init deferred_free_pages(unsigned long pfn,
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
unsigned long end_pfn)
|
|
|
|
{
|
|
|
|
unsigned long nr_pgmask = pageblock_nr_pages - 1;
|
|
|
|
unsigned long nr_free = 0;
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
for (; pfn < end_pfn; pfn++) {
|
2019-05-14 00:21:13 +00:00
|
|
|
if (!deferred_pfn_valid(pfn)) {
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
deferred_free_range(pfn - nr_free, nr_free);
|
|
|
|
nr_free = 0;
|
|
|
|
} else if (!(pfn & nr_pgmask)) {
|
|
|
|
deferred_free_range(pfn - nr_free, nr_free);
|
|
|
|
nr_free = 1;
|
|
|
|
} else {
|
|
|
|
nr_free++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
/* Free the last block of pages to allocator */
|
|
|
|
deferred_free_range(pfn - nr_free, nr_free);
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
}
|
|
|
|
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
/*
|
|
|
|
* Initialize struct pages. We minimize pfn page lookups and scheduler checks
|
|
|
|
* by performing it only once every pageblock_nr_pages.
|
|
|
|
* Return number of pages initialized.
|
|
|
|
*/
|
2019-05-14 00:21:13 +00:00
|
|
|
static unsigned long __init deferred_init_pages(struct zone *zone,
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
unsigned long pfn,
|
|
|
|
unsigned long end_pfn)
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
{
|
|
|
|
unsigned long nr_pgmask = pageblock_nr_pages - 1;
|
2019-05-14 00:21:13 +00:00
|
|
|
int nid = zone_to_nid(zone);
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
unsigned long nr_pages = 0;
|
2019-05-14 00:21:13 +00:00
|
|
|
int zid = zone_idx(zone);
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
struct page *page = NULL;
|
|
|
|
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
for (; pfn < end_pfn; pfn++) {
|
2019-05-14 00:21:13 +00:00
|
|
|
if (!deferred_pfn_valid(pfn)) {
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
page = NULL;
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
continue;
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
} else if (!page || !(pfn & nr_pgmask)) {
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
page = pfn_to_page(pfn);
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
} else {
|
|
|
|
page++;
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
}
|
2018-04-05 23:23:00 +00:00
|
|
|
__init_single_page(page, pfn, zid, nid);
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
nr_pages++;
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
}
|
mm: split deferred_init_range into initializing and freeing parts
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-02-01 00:16:30 +00:00
|
|
|
return (nr_pages);
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
}
|
|
|
|
|
2019-05-14 00:21:20 +00:00
|
|
|
/*
|
|
|
|
* This function is meant to pre-load the iterator for the zone init.
|
|
|
|
* Specifically it walks through the ranges until we are caught up to the
|
|
|
|
* first_init_pfn value and exits there. If we never encounter the value we
|
|
|
|
* return false indicating there are no valid ranges left.
|
|
|
|
*/
|
|
|
|
static bool __init
|
|
|
|
deferred_init_mem_pfn_range_in_zone(u64 *i, struct zone *zone,
|
|
|
|
unsigned long *spfn, unsigned long *epfn,
|
|
|
|
unsigned long first_init_pfn)
|
|
|
|
{
|
|
|
|
u64 j;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Start out by walking through the ranges in this zone that have
|
|
|
|
* already been initialized. We don't need to do anything with them
|
|
|
|
* so we just need to flush them out of the system.
|
|
|
|
*/
|
|
|
|
for_each_free_mem_pfn_range_in_zone(j, zone, spfn, epfn) {
|
|
|
|
if (*epfn <= first_init_pfn)
|
|
|
|
continue;
|
|
|
|
if (*spfn < first_init_pfn)
|
|
|
|
*spfn = first_init_pfn;
|
|
|
|
*i = j;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Initialize and free pages. We do it in two loops: first we initialize
|
|
|
|
* struct page, then free to buddy allocator, because while we are
|
|
|
|
* freeing pages we can access pages that are ahead (computing buddy
|
|
|
|
* page in __free_one_page()).
|
|
|
|
*
|
|
|
|
* In order to try and keep some memory in the cache we have the loop
|
|
|
|
* broken along max page order boundaries. This way we will not cause
|
|
|
|
* any issues with the buddy page computation.
|
|
|
|
*/
|
|
|
|
static unsigned long __init
|
|
|
|
deferred_init_maxorder(u64 *i, struct zone *zone, unsigned long *start_pfn,
|
|
|
|
unsigned long *end_pfn)
|
|
|
|
{
|
|
|
|
unsigned long mo_pfn = ALIGN(*start_pfn + 1, MAX_ORDER_NR_PAGES);
|
|
|
|
unsigned long spfn = *start_pfn, epfn = *end_pfn;
|
|
|
|
unsigned long nr_pages = 0;
|
|
|
|
u64 j = *i;
|
|
|
|
|
|
|
|
/* First we loop through and initialize the page values */
|
|
|
|
for_each_free_mem_pfn_range_in_zone_from(j, zone, start_pfn, end_pfn) {
|
|
|
|
unsigned long t;
|
|
|
|
|
|
|
|
if (mo_pfn <= *start_pfn)
|
|
|
|
break;
|
|
|
|
|
|
|
|
t = min(mo_pfn, *end_pfn);
|
|
|
|
nr_pages += deferred_init_pages(zone, *start_pfn, t);
|
|
|
|
|
|
|
|
if (mo_pfn < *end_pfn) {
|
|
|
|
*start_pfn = mo_pfn;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Reset values and now loop through freeing pages as needed */
|
|
|
|
swap(j, *i);
|
|
|
|
|
|
|
|
for_each_free_mem_pfn_range_in_zone_from(j, zone, &spfn, &epfn) {
|
|
|
|
unsigned long t;
|
|
|
|
|
|
|
|
if (mo_pfn <= spfn)
|
|
|
|
break;
|
|
|
|
|
|
|
|
t = min(mo_pfn, epfn);
|
|
|
|
deferred_free_pages(spfn, t);
|
|
|
|
|
|
|
|
if (mo_pfn <= epfn)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return nr_pages;
|
|
|
|
}
|
|
|
|
|
mm: parallelize deferred_init_memmap()
Deferred struct page init is a significant bottleneck in kernel boot.
Optimizing it maximizes availability for large-memory systems and allows
spinning up short-lived VMs as needed without having to leave them
running. It also benefits bare metal machines hosting VMs that are
sensitive to downtime. In projects such as VMM Fast Restart[1], where
guest state is preserved across kexec reboot, it helps prevent application
and network timeouts in the guests.
Multithread to take full advantage of system memory bandwidth.
The maximum number of threads is capped at the number of CPUs on the node
because speedups always improve with additional threads on every system
tested, and at this phase of boot, the system is otherwise idle and
waiting on page init to finish.
Helper threads operate on section-aligned ranges to both avoid false
sharing when setting the pageblock's migrate type and to avoid accessing
uninitialized buddy pages, though max order alignment is enough for the
latter.
The minimum chunk size is also a section. There was benefit to using
multiple threads even on relatively small memory (1G) systems, and this is
the smallest size that the alignment allows.
The time (milliseconds) is the slowest node to initialize since boot
blocks until all nodes finish. intel_pstate is loaded in active mode
without hwp and with turbo enabled, and intel_idle is active as well.
Intel(R) Xeon(R) Platinum 8167M CPU @ 2.00GHz (Skylake, bare metal)
2 nodes * 26 cores * 2 threads = 104 CPUs
384G/node = 768G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 4089.7 ( 8.1) -- 1785.7 ( 7.6)
2% ( 1) 1.7% 4019.3 ( 1.5) 3.8% 1717.7 ( 11.8)
12% ( 6) 34.9% 2662.7 ( 2.9) 79.9% 359.3 ( 0.6)
25% ( 13) 39.9% 2459.0 ( 3.6) 91.2% 157.0 ( 0.0)
37% ( 19) 39.2% 2485.0 ( 29.7) 90.4% 172.0 ( 28.6)
50% ( 26) 39.3% 2482.7 ( 25.7) 90.3% 173.7 ( 30.0)
75% ( 39) 39.0% 2495.7 ( 5.5) 89.4% 190.0 ( 1.0)
100% ( 52) 40.2% 2443.7 ( 3.8) 92.3% 138.0 ( 1.0)
Intel(R) Xeon(R) CPU E5-2699C v4 @ 2.20GHz (Broadwell, kvm guest)
1 node * 16 cores * 2 threads = 32 CPUs
192G/node = 192G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1988.7 ( 9.6) -- 1096.0 ( 11.5)
3% ( 1) 1.1% 1967.0 ( 17.6) 0.3% 1092.7 ( 11.0)
12% ( 4) 41.1% 1170.3 ( 14.2) 73.8% 287.0 ( 3.6)
25% ( 8) 47.1% 1052.7 ( 21.9) 83.9% 177.0 ( 13.5)
38% ( 12) 48.9% 1016.3 ( 12.1) 86.8% 144.7 ( 1.5)
50% ( 16) 48.9% 1015.7 ( 8.1) 87.8% 134.0 ( 4.4)
75% ( 24) 49.1% 1012.3 ( 3.1) 88.1% 130.3 ( 2.3)
100% ( 32) 49.5% 1004.0 ( 5.3) 88.5% 125.7 ( 2.1)
Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz (Haswell, bare metal)
2 nodes * 18 cores * 2 threads = 72 CPUs
128G/node = 256G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1680.0 ( 4.6) -- 627.0 ( 4.0)
3% ( 1) 0.3% 1675.7 ( 4.5) -0.2% 628.0 ( 3.6)
11% ( 4) 25.6% 1250.7 ( 2.1) 67.9% 201.0 ( 0.0)
25% ( 9) 30.7% 1164.0 ( 17.3) 81.8% 114.3 ( 17.7)
36% ( 13) 31.4% 1152.7 ( 10.8) 84.0% 100.3 ( 17.9)
50% ( 18) 31.5% 1150.7 ( 9.3) 83.9% 101.0 ( 14.1)
75% ( 27) 31.7% 1148.0 ( 5.6) 84.5% 97.3 ( 6.4)
100% ( 36) 32.0% 1142.3 ( 4.0) 85.6% 90.0 ( 1.0)
AMD EPYC 7551 32-Core Processor (Zen, kvm guest)
1 node * 8 cores * 2 threads = 16 CPUs
64G/node = 64G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1029.3 ( 25.1) -- 240.7 ( 1.5)
6% ( 1) -0.6% 1036.0 ( 7.8) -2.2% 246.0 ( 0.0)
12% ( 2) 11.8% 907.7 ( 8.6) 44.7% 133.0 ( 1.0)
25% ( 4) 13.9% 886.0 ( 10.6) 62.6% 90.0 ( 6.0)
38% ( 6) 17.8% 845.7 ( 14.2) 69.1% 74.3 ( 3.8)
50% ( 8) 16.8% 856.0 ( 22.1) 72.9% 65.3 ( 5.7)
75% ( 12) 15.4% 871.0 ( 29.2) 79.8% 48.7 ( 7.4)
100% ( 16) 21.0% 813.7 ( 21.0) 80.5% 47.0 ( 5.2)
Server-oriented distros that enable deferred page init sometimes run in
small VMs, and they still benefit even though the fraction of boot time
saved is smaller:
AMD EPYC 7551 32-Core Processor (Zen, kvm guest)
1 node * 2 cores * 2 threads = 4 CPUs
16G/node = 16G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 716.0 ( 14.0) -- 49.7 ( 0.6)
25% ( 1) 1.8% 703.0 ( 5.3) -4.0% 51.7 ( 0.6)
50% ( 2) 1.6% 704.7 ( 1.2) 43.0% 28.3 ( 0.6)
75% ( 3) 2.7% 696.7 ( 13.1) 49.7% 25.0 ( 0.0)
100% ( 4) 4.1% 687.0 ( 10.4) 55.7% 22.0 ( 0.0)
Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz (Haswell, kvm guest)
1 node * 2 cores * 2 threads = 4 CPUs
14G/node = 14G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 787.7 ( 6.4) -- 122.3 ( 0.6)
25% ( 1) 0.2% 786.3 ( 10.8) -2.5% 125.3 ( 2.1)
50% ( 2) 5.9% 741.0 ( 13.9) 37.6% 76.3 ( 19.7)
75% ( 3) 8.3% 722.0 ( 19.0) 49.9% 61.3 ( 3.2)
100% ( 4) 9.3% 714.7 ( 9.5) 56.4% 53.3 ( 1.5)
On Josh's 96-CPU and 192G memory system:
Without this patch series:
[ 0.487132] node 0 initialised, 23398907 pages in 292ms
[ 0.499132] node 1 initialised, 24189223 pages in 304ms
...
[ 0.629376] Run /sbin/init as init process
With this patch series:
[ 0.231435] node 1 initialised, 24189223 pages in 32ms
[ 0.236718] node 0 initialised, 23398907 pages in 36ms
[1] https://static.sched.com/hosted_files/kvmforum2019/66/VMM-fast-restart_kvmforum2019.pdf
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Robert Elliott <elliott@hpe.com>
Cc: Shile Zhang <shile.zhang@linux.alibaba.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Zi Yan <ziy@nvidia.com>
Link: http://lkml.kernel.org/r/20200527173608.2885243-7-daniel.m.jordan@oracle.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-03 22:59:51 +00:00
|
|
|
static void __init
|
|
|
|
deferred_init_memmap_chunk(unsigned long start_pfn, unsigned long end_pfn,
|
|
|
|
void *arg)
|
|
|
|
{
|
|
|
|
unsigned long spfn, epfn;
|
|
|
|
struct zone *zone = arg;
|
|
|
|
u64 i;
|
|
|
|
|
|
|
|
deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn, start_pfn);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Initialize and free pages in MAX_ORDER sized increments so that we
|
|
|
|
* can avoid introducing any issues with the buddy allocator.
|
|
|
|
*/
|
|
|
|
while (spfn < end_pfn) {
|
|
|
|
deferred_init_maxorder(&i, zone, &spfn, &epfn);
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-06-03 22:59:55 +00:00
|
|
|
/* An arch may override for more concurrency. */
|
|
|
|
__weak int __init
|
|
|
|
deferred_page_init_max_threads(const struct cpumask *node_cpumask)
|
|
|
|
{
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2015-06-30 21:57:05 +00:00
|
|
|
/* Initialise remaining memory on a node */
|
2015-06-30 21:57:27 +00:00
|
|
|
static int __init deferred_init_memmap(void *data)
|
2015-06-30 21:57:05 +00:00
|
|
|
{
|
2015-06-30 21:57:27 +00:00
|
|
|
pg_data_t *pgdat = data;
|
2019-05-14 00:21:20 +00:00
|
|
|
const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
|
2020-06-03 22:59:47 +00:00
|
|
|
unsigned long spfn = 0, epfn = 0;
|
2019-05-14 00:21:20 +00:00
|
|
|
unsigned long first_init_pfn, flags;
|
2015-06-30 21:57:05 +00:00
|
|
|
unsigned long start = jiffies;
|
|
|
|
struct zone *zone;
|
mm: parallelize deferred_init_memmap()
Deferred struct page init is a significant bottleneck in kernel boot.
Optimizing it maximizes availability for large-memory systems and allows
spinning up short-lived VMs as needed without having to leave them
running. It also benefits bare metal machines hosting VMs that are
sensitive to downtime. In projects such as VMM Fast Restart[1], where
guest state is preserved across kexec reboot, it helps prevent application
and network timeouts in the guests.
Multithread to take full advantage of system memory bandwidth.
The maximum number of threads is capped at the number of CPUs on the node
because speedups always improve with additional threads on every system
tested, and at this phase of boot, the system is otherwise idle and
waiting on page init to finish.
Helper threads operate on section-aligned ranges to both avoid false
sharing when setting the pageblock's migrate type and to avoid accessing
uninitialized buddy pages, though max order alignment is enough for the
latter.
The minimum chunk size is also a section. There was benefit to using
multiple threads even on relatively small memory (1G) systems, and this is
the smallest size that the alignment allows.
The time (milliseconds) is the slowest node to initialize since boot
blocks until all nodes finish. intel_pstate is loaded in active mode
without hwp and with turbo enabled, and intel_idle is active as well.
Intel(R) Xeon(R) Platinum 8167M CPU @ 2.00GHz (Skylake, bare metal)
2 nodes * 26 cores * 2 threads = 104 CPUs
384G/node = 768G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 4089.7 ( 8.1) -- 1785.7 ( 7.6)
2% ( 1) 1.7% 4019.3 ( 1.5) 3.8% 1717.7 ( 11.8)
12% ( 6) 34.9% 2662.7 ( 2.9) 79.9% 359.3 ( 0.6)
25% ( 13) 39.9% 2459.0 ( 3.6) 91.2% 157.0 ( 0.0)
37% ( 19) 39.2% 2485.0 ( 29.7) 90.4% 172.0 ( 28.6)
50% ( 26) 39.3% 2482.7 ( 25.7) 90.3% 173.7 ( 30.0)
75% ( 39) 39.0% 2495.7 ( 5.5) 89.4% 190.0 ( 1.0)
100% ( 52) 40.2% 2443.7 ( 3.8) 92.3% 138.0 ( 1.0)
Intel(R) Xeon(R) CPU E5-2699C v4 @ 2.20GHz (Broadwell, kvm guest)
1 node * 16 cores * 2 threads = 32 CPUs
192G/node = 192G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1988.7 ( 9.6) -- 1096.0 ( 11.5)
3% ( 1) 1.1% 1967.0 ( 17.6) 0.3% 1092.7 ( 11.0)
12% ( 4) 41.1% 1170.3 ( 14.2) 73.8% 287.0 ( 3.6)
25% ( 8) 47.1% 1052.7 ( 21.9) 83.9% 177.0 ( 13.5)
38% ( 12) 48.9% 1016.3 ( 12.1) 86.8% 144.7 ( 1.5)
50% ( 16) 48.9% 1015.7 ( 8.1) 87.8% 134.0 ( 4.4)
75% ( 24) 49.1% 1012.3 ( 3.1) 88.1% 130.3 ( 2.3)
100% ( 32) 49.5% 1004.0 ( 5.3) 88.5% 125.7 ( 2.1)
Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz (Haswell, bare metal)
2 nodes * 18 cores * 2 threads = 72 CPUs
128G/node = 256G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1680.0 ( 4.6) -- 627.0 ( 4.0)
3% ( 1) 0.3% 1675.7 ( 4.5) -0.2% 628.0 ( 3.6)
11% ( 4) 25.6% 1250.7 ( 2.1) 67.9% 201.0 ( 0.0)
25% ( 9) 30.7% 1164.0 ( 17.3) 81.8% 114.3 ( 17.7)
36% ( 13) 31.4% 1152.7 ( 10.8) 84.0% 100.3 ( 17.9)
50% ( 18) 31.5% 1150.7 ( 9.3) 83.9% 101.0 ( 14.1)
75% ( 27) 31.7% 1148.0 ( 5.6) 84.5% 97.3 ( 6.4)
100% ( 36) 32.0% 1142.3 ( 4.0) 85.6% 90.0 ( 1.0)
AMD EPYC 7551 32-Core Processor (Zen, kvm guest)
1 node * 8 cores * 2 threads = 16 CPUs
64G/node = 64G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1029.3 ( 25.1) -- 240.7 ( 1.5)
6% ( 1) -0.6% 1036.0 ( 7.8) -2.2% 246.0 ( 0.0)
12% ( 2) 11.8% 907.7 ( 8.6) 44.7% 133.0 ( 1.0)
25% ( 4) 13.9% 886.0 ( 10.6) 62.6% 90.0 ( 6.0)
38% ( 6) 17.8% 845.7 ( 14.2) 69.1% 74.3 ( 3.8)
50% ( 8) 16.8% 856.0 ( 22.1) 72.9% 65.3 ( 5.7)
75% ( 12) 15.4% 871.0 ( 29.2) 79.8% 48.7 ( 7.4)
100% ( 16) 21.0% 813.7 ( 21.0) 80.5% 47.0 ( 5.2)
Server-oriented distros that enable deferred page init sometimes run in
small VMs, and they still benefit even though the fraction of boot time
saved is smaller:
AMD EPYC 7551 32-Core Processor (Zen, kvm guest)
1 node * 2 cores * 2 threads = 4 CPUs
16G/node = 16G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 716.0 ( 14.0) -- 49.7 ( 0.6)
25% ( 1) 1.8% 703.0 ( 5.3) -4.0% 51.7 ( 0.6)
50% ( 2) 1.6% 704.7 ( 1.2) 43.0% 28.3 ( 0.6)
75% ( 3) 2.7% 696.7 ( 13.1) 49.7% 25.0 ( 0.0)
100% ( 4) 4.1% 687.0 ( 10.4) 55.7% 22.0 ( 0.0)
Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz (Haswell, kvm guest)
1 node * 2 cores * 2 threads = 4 CPUs
14G/node = 14G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 787.7 ( 6.4) -- 122.3 ( 0.6)
25% ( 1) 0.2% 786.3 ( 10.8) -2.5% 125.3 ( 2.1)
50% ( 2) 5.9% 741.0 ( 13.9) 37.6% 76.3 ( 19.7)
75% ( 3) 8.3% 722.0 ( 19.0) 49.9% 61.3 ( 3.2)
100% ( 4) 9.3% 714.7 ( 9.5) 56.4% 53.3 ( 1.5)
On Josh's 96-CPU and 192G memory system:
Without this patch series:
[ 0.487132] node 0 initialised, 23398907 pages in 292ms
[ 0.499132] node 1 initialised, 24189223 pages in 304ms
...
[ 0.629376] Run /sbin/init as init process
With this patch series:
[ 0.231435] node 1 initialised, 24189223 pages in 32ms
[ 0.236718] node 0 initialised, 23398907 pages in 36ms
[1] https://static.sched.com/hosted_files/kvmforum2019/66/VMM-fast-restart_kvmforum2019.pdf
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Robert Elliott <elliott@hpe.com>
Cc: Shile Zhang <shile.zhang@linux.alibaba.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Zi Yan <ziy@nvidia.com>
Link: http://lkml.kernel.org/r/20200527173608.2885243-7-daniel.m.jordan@oracle.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-03 22:59:51 +00:00
|
|
|
int zid, max_threads;
|
mm: deferred_init_memmap improvements
Patch series "complete deferred page initialization", v12.
SMP machines can benefit from the DEFERRED_STRUCT_PAGE_INIT config
option, which defers initializing struct pages until all cpus have been
started so it can be done in parallel.
However, this feature is sub-optimal, because the deferred page
initialization code expects that the struct pages have already been
zeroed, and the zeroing is done early in boot with a single thread only.
Also, we access that memory and set flags before struct pages are
initialized. All of this is fixed in this patchset.
In this work we do the following:
- Never read access struct page until it was initialized
- Never set any fields in struct pages before they are initialized
- Zero struct page at the beginning of struct page initialization
==========================================================================
Performance improvements on x86 machine with 8 nodes:
Intel(R) Xeon(R) CPU E7-8895 v3 @ 2.60GHz and 1T of memory:
TIME SPEED UP
base no deferred: 95.796233s
fix no deferred: 79.978956s 19.77%
base deferred: 77.254713s
fix deferred: 55.050509s 40.34%
==========================================================================
SPARC M6 3600 MHz with 15T of memory
TIME SPEED UP
base no deferred: 358.335727s
fix no deferred: 302.320936s 18.52%
base deferred: 237.534603s
fix deferred: 182.103003s 30.44%
==========================================================================
Raw dmesg output with timestamps:
x86 base no deferred: https://hastebin.com/ofunepurit.scala
x86 base deferred: https://hastebin.com/ifazegeyas.scala
x86 fix no deferred: https://hastebin.com/pegocohevo.scala
x86 fix deferred: https://hastebin.com/ofupevikuk.scala
sparc base no deferred: https://hastebin.com/ibobeteken.go
sparc base deferred: https://hastebin.com/fariqimiyu.go
sparc fix no deferred: https://hastebin.com/muhegoheyi.go
sparc fix deferred: https://hastebin.com/xadinobutu.go
This patch (of 11):
deferred_init_memmap() is called when struct pages are initialized later
in boot by slave CPUs. This patch simplifies and optimizes this
function, and also fixes a couple issues (described below).
The main change is that now we are iterating through free memblock areas
instead of all configured memory. Thus, we do not have to check if the
struct page has already been initialized.
=====
In deferred_init_memmap() where all deferred struct pages are
initialized we have a check like this:
if (page->flags) {
VM_BUG_ON(page_zone(page) != zone);
goto free_range;
}
This way we are checking if the current deferred page has already been
initialized. It works, because memory for struct pages has been zeroed,
and the only way flags are not zero if it went through
__init_single_page() before. But, once we change the current behavior
and won't zero the memory in memblock allocator, we cannot trust
anything inside "struct page"es until they are initialized. This patch
fixes this.
The deferred_init_memmap() is re-written to loop through only free
memory ranges provided by memblock.
Note, this first issue is relevant only when the following change is
merged:
=====
This patch fixes another existing issue on systems that have holes in
zones i.e CONFIG_HOLES_IN_ZONE is defined.
In for_each_mem_pfn_range() we have code like this:
if (!pfn_valid_within(pfn)
goto free_range;
Note: 'page' is not set to NULL and is not incremented but 'pfn'
advances. Thus means if deferred struct pages are enabled on systems
with these kind of holes, linux would get memory corruptions. I have
fixed this issue by defining a new macro that performs all the necessary
operations when we free the current set of pages.
[pasha.tatashin@oracle.com: buddy page accessed before initialized]
Link: http://lkml.kernel.org/r/20171102170221.7401-2-pasha.tatashin@oracle.com
Link: http://lkml.kernel.org/r/20171013173214.27300-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:09 +00:00
|
|
|
u64 i;
|
2015-06-30 21:57:05 +00:00
|
|
|
|
2018-04-05 23:22:27 +00:00
|
|
|
/* Bind memory initialisation thread to a local node if possible */
|
|
|
|
if (!cpumask_empty(cpumask))
|
|
|
|
set_cpus_allowed_ptr(current, cpumask);
|
|
|
|
|
|
|
|
pgdat_resize_lock(pgdat, &flags);
|
|
|
|
first_init_pfn = pgdat->first_deferred_pfn;
|
2015-06-30 21:57:27 +00:00
|
|
|
if (first_init_pfn == ULONG_MAX) {
|
2018-04-05 23:22:27 +00:00
|
|
|
pgdat_resize_unlock(pgdat, &flags);
|
2015-08-06 22:46:16 +00:00
|
|
|
pgdat_init_report_one_done();
|
2015-06-30 21:57:27 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2015-06-30 21:57:05 +00:00
|
|
|
/* Sanity check boundaries */
|
|
|
|
BUG_ON(pgdat->first_deferred_pfn < pgdat->node_start_pfn);
|
|
|
|
BUG_ON(pgdat->first_deferred_pfn > pgdat_end_pfn(pgdat));
|
|
|
|
pgdat->first_deferred_pfn = ULONG_MAX;
|
|
|
|
|
2020-06-03 22:59:24 +00:00
|
|
|
/*
|
|
|
|
* Once we unlock here, the zone cannot be grown anymore, thus if an
|
|
|
|
* interrupt thread must allocate this early in boot, zone must be
|
|
|
|
* pre-grown prior to start of deferred page initialization.
|
|
|
|
*/
|
|
|
|
pgdat_resize_unlock(pgdat, &flags);
|
|
|
|
|
2015-06-30 21:57:05 +00:00
|
|
|
/* Only the highest zone is deferred so find it */
|
|
|
|
for (zid = 0; zid < MAX_NR_ZONES; zid++) {
|
|
|
|
zone = pgdat->node_zones + zid;
|
|
|
|
if (first_init_pfn < zone_end_pfn(zone))
|
|
|
|
break;
|
|
|
|
}
|
2019-05-14 00:21:20 +00:00
|
|
|
|
|
|
|
/* If the zone is empty somebody else may have cleared out the zone */
|
|
|
|
if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
|
|
|
|
first_init_pfn))
|
|
|
|
goto zone_empty;
|
2015-06-30 21:57:05 +00:00
|
|
|
|
2020-06-03 22:59:55 +00:00
|
|
|
max_threads = deferred_page_init_max_threads(cpumask);
|
2015-06-30 21:57:05 +00:00
|
|
|
|
2020-06-03 22:59:20 +00:00
|
|
|
while (spfn < epfn) {
|
mm: parallelize deferred_init_memmap()
Deferred struct page init is a significant bottleneck in kernel boot.
Optimizing it maximizes availability for large-memory systems and allows
spinning up short-lived VMs as needed without having to leave them
running. It also benefits bare metal machines hosting VMs that are
sensitive to downtime. In projects such as VMM Fast Restart[1], where
guest state is preserved across kexec reboot, it helps prevent application
and network timeouts in the guests.
Multithread to take full advantage of system memory bandwidth.
The maximum number of threads is capped at the number of CPUs on the node
because speedups always improve with additional threads on every system
tested, and at this phase of boot, the system is otherwise idle and
waiting on page init to finish.
Helper threads operate on section-aligned ranges to both avoid false
sharing when setting the pageblock's migrate type and to avoid accessing
uninitialized buddy pages, though max order alignment is enough for the
latter.
The minimum chunk size is also a section. There was benefit to using
multiple threads even on relatively small memory (1G) systems, and this is
the smallest size that the alignment allows.
The time (milliseconds) is the slowest node to initialize since boot
blocks until all nodes finish. intel_pstate is loaded in active mode
without hwp and with turbo enabled, and intel_idle is active as well.
Intel(R) Xeon(R) Platinum 8167M CPU @ 2.00GHz (Skylake, bare metal)
2 nodes * 26 cores * 2 threads = 104 CPUs
384G/node = 768G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 4089.7 ( 8.1) -- 1785.7 ( 7.6)
2% ( 1) 1.7% 4019.3 ( 1.5) 3.8% 1717.7 ( 11.8)
12% ( 6) 34.9% 2662.7 ( 2.9) 79.9% 359.3 ( 0.6)
25% ( 13) 39.9% 2459.0 ( 3.6) 91.2% 157.0 ( 0.0)
37% ( 19) 39.2% 2485.0 ( 29.7) 90.4% 172.0 ( 28.6)
50% ( 26) 39.3% 2482.7 ( 25.7) 90.3% 173.7 ( 30.0)
75% ( 39) 39.0% 2495.7 ( 5.5) 89.4% 190.0 ( 1.0)
100% ( 52) 40.2% 2443.7 ( 3.8) 92.3% 138.0 ( 1.0)
Intel(R) Xeon(R) CPU E5-2699C v4 @ 2.20GHz (Broadwell, kvm guest)
1 node * 16 cores * 2 threads = 32 CPUs
192G/node = 192G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1988.7 ( 9.6) -- 1096.0 ( 11.5)
3% ( 1) 1.1% 1967.0 ( 17.6) 0.3% 1092.7 ( 11.0)
12% ( 4) 41.1% 1170.3 ( 14.2) 73.8% 287.0 ( 3.6)
25% ( 8) 47.1% 1052.7 ( 21.9) 83.9% 177.0 ( 13.5)
38% ( 12) 48.9% 1016.3 ( 12.1) 86.8% 144.7 ( 1.5)
50% ( 16) 48.9% 1015.7 ( 8.1) 87.8% 134.0 ( 4.4)
75% ( 24) 49.1% 1012.3 ( 3.1) 88.1% 130.3 ( 2.3)
100% ( 32) 49.5% 1004.0 ( 5.3) 88.5% 125.7 ( 2.1)
Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz (Haswell, bare metal)
2 nodes * 18 cores * 2 threads = 72 CPUs
128G/node = 256G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1680.0 ( 4.6) -- 627.0 ( 4.0)
3% ( 1) 0.3% 1675.7 ( 4.5) -0.2% 628.0 ( 3.6)
11% ( 4) 25.6% 1250.7 ( 2.1) 67.9% 201.0 ( 0.0)
25% ( 9) 30.7% 1164.0 ( 17.3) 81.8% 114.3 ( 17.7)
36% ( 13) 31.4% 1152.7 ( 10.8) 84.0% 100.3 ( 17.9)
50% ( 18) 31.5% 1150.7 ( 9.3) 83.9% 101.0 ( 14.1)
75% ( 27) 31.7% 1148.0 ( 5.6) 84.5% 97.3 ( 6.4)
100% ( 36) 32.0% 1142.3 ( 4.0) 85.6% 90.0 ( 1.0)
AMD EPYC 7551 32-Core Processor (Zen, kvm guest)
1 node * 8 cores * 2 threads = 16 CPUs
64G/node = 64G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 1029.3 ( 25.1) -- 240.7 ( 1.5)
6% ( 1) -0.6% 1036.0 ( 7.8) -2.2% 246.0 ( 0.0)
12% ( 2) 11.8% 907.7 ( 8.6) 44.7% 133.0 ( 1.0)
25% ( 4) 13.9% 886.0 ( 10.6) 62.6% 90.0 ( 6.0)
38% ( 6) 17.8% 845.7 ( 14.2) 69.1% 74.3 ( 3.8)
50% ( 8) 16.8% 856.0 ( 22.1) 72.9% 65.3 ( 5.7)
75% ( 12) 15.4% 871.0 ( 29.2) 79.8% 48.7 ( 7.4)
100% ( 16) 21.0% 813.7 ( 21.0) 80.5% 47.0 ( 5.2)
Server-oriented distros that enable deferred page init sometimes run in
small VMs, and they still benefit even though the fraction of boot time
saved is smaller:
AMD EPYC 7551 32-Core Processor (Zen, kvm guest)
1 node * 2 cores * 2 threads = 4 CPUs
16G/node = 16G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 716.0 ( 14.0) -- 49.7 ( 0.6)
25% ( 1) 1.8% 703.0 ( 5.3) -4.0% 51.7 ( 0.6)
50% ( 2) 1.6% 704.7 ( 1.2) 43.0% 28.3 ( 0.6)
75% ( 3) 2.7% 696.7 ( 13.1) 49.7% 25.0 ( 0.0)
100% ( 4) 4.1% 687.0 ( 10.4) 55.7% 22.0 ( 0.0)
Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz (Haswell, kvm guest)
1 node * 2 cores * 2 threads = 4 CPUs
14G/node = 14G memory
kernel boot deferred init
------------------------ ------------------------
node% (thr) speedup time_ms (stdev) speedup time_ms (stdev)
( 0) -- 787.7 ( 6.4) -- 122.3 ( 0.6)
25% ( 1) 0.2% 786.3 ( 10.8) -2.5% 125.3 ( 2.1)
50% ( 2) 5.9% 741.0 ( 13.9) 37.6% 76.3 ( 19.7)
75% ( 3) 8.3% 722.0 ( 19.0) 49.9% 61.3 ( 3.2)
100% ( 4) 9.3% 714.7 ( 9.5) 56.4% 53.3 ( 1.5)
On Josh's 96-CPU and 192G memory system:
Without this patch series:
[ 0.487132] node 0 initialised, 23398907 pages in 292ms
[ 0.499132] node 1 initialised, 24189223 pages in 304ms
...
[ 0.629376] Run /sbin/init as init process
With this patch series:
[ 0.231435] node 1 initialised, 24189223 pages in 32ms
[ 0.236718] node 0 initialised, 23398907 pages in 36ms
[1] https://static.sched.com/hosted_files/kvmforum2019/66/VMM-fast-restart_kvmforum2019.pdf
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Robert Elliott <elliott@hpe.com>
Cc: Shile Zhang <shile.zhang@linux.alibaba.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Zi Yan <ziy@nvidia.com>
Link: http://lkml.kernel.org/r/20200527173608.2885243-7-daniel.m.jordan@oracle.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-03 22:59:51 +00:00
|
|
|
unsigned long epfn_align = ALIGN(epfn, PAGES_PER_SECTION);
|
|
|
|
struct padata_mt_job job = {
|
|
|
|
.thread_fn = deferred_init_memmap_chunk,
|
|
|
|
.fn_arg = zone,
|
|
|
|
.start = spfn,
|
|
|
|
.size = epfn_align - spfn,
|
|
|
|
.align = PAGES_PER_SECTION,
|
|
|
|
.min_chunk = PAGES_PER_SECTION,
|
|
|
|
.max_threads = max_threads,
|
|
|
|
};
|
|
|
|
|
|
|
|
padata_do_multithreaded(&job);
|
|
|
|
deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
|
|
|
|
epfn_align);
|
2020-06-03 22:59:20 +00:00
|
|
|
}
|
2019-05-14 00:21:20 +00:00
|
|
|
zone_empty:
|
2015-06-30 21:57:05 +00:00
|
|
|
/* Sanity check that the next zone really is unpopulated */
|
|
|
|
WARN_ON(++zid < MAX_NR_ZONES && populated_zone(++zone));
|
|
|
|
|
2020-06-03 22:59:47 +00:00
|
|
|
pr_info("node %d deferred pages initialised in %ums\n",
|
|
|
|
pgdat->node_id, jiffies_to_msecs(jiffies - start));
|
2015-08-06 22:46:16 +00:00
|
|
|
|
|
|
|
pgdat_init_report_one_done();
|
2015-06-30 21:57:27 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2018-04-05 23:22:31 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If this zone has deferred pages, try to grow it by initializing enough
|
|
|
|
* deferred pages to satisfy the allocation specified by order, rounded up to
|
|
|
|
* the nearest PAGES_PER_SECTION boundary. So we're adding memory in increments
|
|
|
|
* of SECTION_SIZE bytes by initializing struct pages in increments of
|
|
|
|
* PAGES_PER_SECTION * sizeof(struct page) bytes.
|
|
|
|
*
|
|
|
|
* Return true when zone was grown, otherwise return false. We return true even
|
|
|
|
* when we grow less than requested, to let the caller decide if there are
|
|
|
|
* enough pages to satisfy the allocation.
|
|
|
|
*
|
|
|
|
* Note: We use noinline because this function is needed only during boot, and
|
|
|
|
* it is called from a __ref function _deferred_grow_zone. This way we are
|
|
|
|
* making sure that it is not inlined into permanent text section.
|
|
|
|
*/
|
|
|
|
static noinline bool __init
|
|
|
|
deferred_grow_zone(struct zone *zone, unsigned int order)
|
|
|
|
{
|
|
|
|
unsigned long nr_pages_needed = ALIGN(1 << order, PAGES_PER_SECTION);
|
2019-05-14 00:21:17 +00:00
|
|
|
pg_data_t *pgdat = zone->zone_pgdat;
|
2018-04-05 23:22:31 +00:00
|
|
|
unsigned long first_deferred_pfn = pgdat->first_deferred_pfn;
|
2019-05-14 00:21:20 +00:00
|
|
|
unsigned long spfn, epfn, flags;
|
|
|
|
unsigned long nr_pages = 0;
|
2018-04-05 23:22:31 +00:00
|
|
|
u64 i;
|
|
|
|
|
|
|
|
/* Only the last zone may have deferred pages */
|
|
|
|
if (zone_end_pfn(zone) != pgdat_end_pfn(pgdat))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
pgdat_resize_lock(pgdat, &flags);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If someone grew this zone while we were waiting for spinlock, return
|
|
|
|
* true, as there might be enough pages already.
|
|
|
|
*/
|
|
|
|
if (first_deferred_pfn != pgdat->first_deferred_pfn) {
|
|
|
|
pgdat_resize_unlock(pgdat, &flags);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2019-05-14 00:21:20 +00:00
|
|
|
/* If the zone is empty somebody else may have cleared out the zone */
|
|
|
|
if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
|
|
|
|
first_deferred_pfn)) {
|
|
|
|
pgdat->first_deferred_pfn = ULONG_MAX;
|
2018-04-05 23:22:31 +00:00
|
|
|
pgdat_resize_unlock(pgdat, &flags);
|
2019-07-04 22:14:36 +00:00
|
|
|
/* Retry only once. */
|
|
|
|
return first_deferred_pfn != ULONG_MAX;
|
2018-04-05 23:22:31 +00:00
|
|
|
}
|
|
|
|
|
2019-05-14 00:21:20 +00:00
|
|
|
/*
|
|
|
|
* Initialize and free pages in MAX_ORDER sized increments so
|
|
|
|
* that we can avoid introducing any issues with the buddy
|
|
|
|
* allocator.
|
|
|
|
*/
|
|
|
|
while (spfn < epfn) {
|
|
|
|
/* update our first deferred PFN for this section */
|
|
|
|
first_deferred_pfn = spfn;
|
|
|
|
|
|
|
|
nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
|
2020-06-03 22:59:20 +00:00
|
|
|
touch_nmi_watchdog();
|
2018-04-05 23:22:31 +00:00
|
|
|
|
2019-05-14 00:21:20 +00:00
|
|
|
/* We should only stop along section boundaries */
|
|
|
|
if ((first_deferred_pfn ^ spfn) < PAGES_PER_SECTION)
|
|
|
|
continue;
|
2018-04-05 23:22:31 +00:00
|
|
|
|
2019-05-14 00:21:20 +00:00
|
|
|
/* If our quota has been met we can stop here */
|
2018-04-05 23:22:31 +00:00
|
|
|
if (nr_pages >= nr_pages_needed)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2019-05-14 00:21:20 +00:00
|
|
|
pgdat->first_deferred_pfn = spfn;
|
2018-04-05 23:22:31 +00:00
|
|
|
pgdat_resize_unlock(pgdat, &flags);
|
|
|
|
|
|
|
|
return nr_pages > 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* deferred_grow_zone() is __init, but it is called from
|
|
|
|
* get_page_from_freelist() during early boot until deferred_pages permanently
|
|
|
|
* disables this call. This is why we have refdata wrapper to avoid warning,
|
|
|
|
* and to ensure that the function body gets unloaded.
|
|
|
|
*/
|
|
|
|
static bool __ref
|
|
|
|
_deferred_grow_zone(struct zone *zone, unsigned int order)
|
|
|
|
{
|
|
|
|
return deferred_grow_zone(zone, order);
|
|
|
|
}
|
|
|
|
|
2016-03-15 21:57:51 +00:00
|
|
|
#endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
|
2015-06-30 21:57:27 +00:00
|
|
|
|
|
|
|
void __init page_alloc_init_late(void)
|
|
|
|
{
|
2016-03-15 21:57:51 +00:00
|
|
|
struct zone *zone;
|
mm: shuffle initial free memory to improve memory-side-cache utilization
Patch series "mm: Randomize free memory", v10.
This patch (of 3):
Randomization of the page allocator improves the average utilization of
a direct-mapped memory-side-cache. Memory side caching is a platform
capability that Linux has been previously exposed to in HPC
(high-performance computing) environments on specialty platforms. In
that instance it was a smaller pool of high-bandwidth-memory relative to
higher-capacity / lower-bandwidth DRAM. Now, this capability is going
to be found on general purpose server platforms where DRAM is a cache in
front of higher latency persistent memory [1].
Robert offered an explanation of the state of the art of Linux
interactions with memory-side-caches [2], and I copy it here:
It's been a problem in the HPC space:
http://www.nersc.gov/research-and-development/knl-cache-mode-performance-coe/
A kernel module called zonesort is available to try to help:
https://software.intel.com/en-us/articles/xeon-phi-software
and this abandoned patch series proposed that for the kernel:
https://lkml.kernel.org/r/20170823100205.17311-1-lukasz.daniluk@intel.com
Dan's patch series doesn't attempt to ensure buffers won't conflict, but
also reduces the chance that the buffers will. This will make performance
more consistent, albeit slower than "optimal" (which is near impossible
to attain in a general-purpose kernel). That's better than forcing
users to deploy remedies like:
"To eliminate this gradual degradation, we have added a Stream
measurement to the Node Health Check that follows each job;
nodes are rebooted whenever their measured memory bandwidth
falls below 300 GB/s."
A replacement for zonesort was merged upstream in commit cc9aec03e58f
("x86/numa_emulation: Introduce uniform split capability"). With this
numa_emulation capability, memory can be split into cache sized
("near-memory" sized) numa nodes. A bind operation to such a node, and
disabling workloads on other nodes, enables full cache performance.
However, once the workload exceeds the cache size then cache conflicts
are unavoidable. While HPC environments might be able to tolerate
time-scheduling of cache sized workloads, for general purpose server
platforms, the oversubscribed cache case will be the common case.
The worst case scenario is that a server system owner benchmarks a
workload at boot with an un-contended cache only to see that performance
degrade over time, even below the average cache performance due to
excessive conflicts. Randomization clips the peaks and fills in the
valleys of cache utilization to yield steady average performance.
Here are some performance impact details of the patches:
1/ An Intel internal synthetic memory bandwidth measurement tool, saw a
3X speedup in a contrived case that tries to force cache conflicts.
The contrived cased used the numa_emulation capability to force an
instance of the benchmark to be run in two of the near-memory sized
numa nodes. If both instances were placed on the same emulated they
would fit and cause zero conflicts. While on separate emulated nodes
without randomization they underutilized the cache and conflicted
unnecessarily due to the in-order allocation per node.
2/ A well known Java server application benchmark was run with a heap
size that exceeded cache size by 3X. The cache conflict rate was 8%
for the first run and degraded to 21% after page allocator aging. With
randomization enabled the rate levelled out at 11%.
3/ A MongoDB workload did not observe measurable difference in
cache-conflict rates, but the overall throughput dropped by 7% with
randomization in one case.
4/ Mel Gorman ran his suite of performance workloads with randomization
enabled on platforms without a memory-side-cache and saw a mix of some
improvements and some losses [3].
While there is potentially significant improvement for applications that
depend on low latency access across a wide working-set, the performance
may be negligible to negative for other workloads. For this reason the
shuffle capability defaults to off unless a direct-mapped
memory-side-cache is detected. Even then, the page_alloc.shuffle=0
parameter can be specified to disable the randomization on those systems.
Outside of memory-side-cache utilization concerns there is potentially
security benefit from randomization. Some data exfiltration and
return-oriented-programming attacks rely on the ability to infer the
location of sensitive data objects. The kernel page allocator, especially
early in system boot, has predictable first-in-first out behavior for
physical pages. Pages are freed in physical address order when first
onlined.
Quoting Kees:
"While we already have a base-address randomization
(CONFIG_RANDOMIZE_MEMORY), attacks against the same hardware and
memory layouts would certainly be using the predictability of
allocation ordering (i.e. for attacks where the base address isn't
important: only the relative positions between allocated memory).
This is common in lots of heap-style attacks. They try to gain
control over ordering by spraying allocations, etc.
I'd really like to see this because it gives us something similar
to CONFIG_SLAB_FREELIST_RANDOM but for the page allocator."
While SLAB_FREELIST_RANDOM reduces the predictability of some local slab
caches it leaves vast bulk of memory to be predictably in order allocated.
However, it should be noted, the concrete security benefits are hard to
quantify, and no known CVE is mitigated by this randomization.
Introduce shuffle_free_memory(), and its helper shuffle_zone(), to perform
a Fisher-Yates shuffle of the page allocator 'free_area' lists when they
are initially populated with free memory at boot and at hotplug time. Do
this based on either the presence of a page_alloc.shuffle=Y command line
parameter, or autodetection of a memory-side-cache (to be added in a
follow-on patch).
The shuffling is done in terms of CONFIG_SHUFFLE_PAGE_ORDER sized free
pages where the default CONFIG_SHUFFLE_PAGE_ORDER is MAX_ORDER-1 i.e. 10,
4MB this trades off randomization granularity for time spent shuffling.
MAX_ORDER-1 was chosen to be minimally invasive to the page allocator
while still showing memory-side cache behavior improvements, and the
expectation that the security implications of finer granularity
randomization is mitigated by CONFIG_SLAB_FREELIST_RANDOM. The
performance impact of the shuffling appears to be in the noise compared to
other memory initialization work.
This initial randomization can be undone over time so a follow-on patch is
introduced to inject entropy on page free decisions. It is reasonable to
ask if the page free entropy is sufficient, but it is not enough due to
the in-order initial freeing of pages. At the start of that process
putting page1 in front or behind page0 still keeps them close together,
page2 is still near page1 and has a high chance of being adjacent. As
more pages are added ordering diversity improves, but there is still high
page locality for the low address pages and this leads to no significant
impact to the cache conflict rate.
[1]: https://itpeernetwork.intel.com/intel-optane-dc-persistent-memory-operating-modes/
[2]: https://lkml.kernel.org/r/AT5PR8401MB1169D656C8B5E121752FC0F8AB120@AT5PR8401MB1169.NAMPRD84.PROD.OUTLOOK.COM
[3]: https://lkml.org/lkml/2018/10/12/309
[dan.j.williams@intel.com: fix shuffle enable]
Link: http://lkml.kernel.org/r/154943713038.3858443.4125180191382062871.stgit@dwillia2-desk3.amr.corp.intel.com
[cai@lca.pw: fix SHUFFLE_PAGE_ALLOCATOR help texts]
Link: http://lkml.kernel.org/r/20190425201300.75650-1-cai@lca.pw
Link: http://lkml.kernel.org/r/154899811738.3165233.12325692939590944259.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Robert Elliott <elliott@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 22:41:28 +00:00
|
|
|
int nid;
|
2016-03-15 21:57:51 +00:00
|
|
|
|
|
|
|
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
|
2015-06-30 21:57:27 +00:00
|
|
|
|
2015-08-06 22:46:16 +00:00
|
|
|
/* There will be num_node_state(N_MEMORY) threads */
|
|
|
|
atomic_set(&pgdat_init_n_undone, num_node_state(N_MEMORY));
|
2015-06-30 21:57:27 +00:00
|
|
|
for_each_node_state(nid, N_MEMORY) {
|
|
|
|
kthread_run(deferred_init_memmap, NODE_DATA(nid), "pgdatinit%d", nid);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Block until all are initialised */
|
2015-08-06 22:46:16 +00:00
|
|
|
wait_for_completion(&pgdat_init_all_done_comp);
|
2015-08-06 22:46:20 +00:00
|
|
|
|
2019-11-06 05:16:27 +00:00
|
|
|
/*
|
|
|
|
* The number of managed pages has changed due to the initialisation
|
|
|
|
* so the pcpu batch and high limits needs to be updated or the limits
|
|
|
|
* will be artificially small.
|
|
|
|
*/
|
|
|
|
for_each_populated_zone(zone)
|
|
|
|
zone_pcp_update(zone);
|
|
|
|
|
2018-04-05 23:22:31 +00:00
|
|
|
/*
|
|
|
|
* We initialized the rest of the deferred pages. Permanently disable
|
|
|
|
* on-demand struct page initialization.
|
|
|
|
*/
|
|
|
|
static_branch_disable(&deferred_pages);
|
|
|
|
|
2015-08-06 22:46:20 +00:00
|
|
|
/* Reinit limits that are based on free pages after the kernel is up */
|
|
|
|
files_maxfiles_init();
|
2016-03-15 21:57:51 +00:00
|
|
|
#endif
|
2019-05-14 00:22:59 +00:00
|
|
|
|
init/main: fix broken buffer_init when DEFERRED_STRUCT_PAGE_INIT set
In the booting phase if CONFIG_DEFERRED_STRUCT_PAGE_INIT is set,
we have following callchain:
start_kernel
...
mm_init
mem_init
memblock_free_all
reset_all_zones_managed_pages
free_low_memory_core_early
...
buffer_init
nr_free_buffer_pages
zone->managed_pages
...
rest_init
kernel_init
kernel_init_freeable
page_alloc_init_late
kthread_run(deferred_init_memmap, NODE_DATA(nid), "pgdatinit%d", nid);
wait_for_completion(&pgdat_init_all_done_comp);
...
files_maxfiles_init
It's clear that buffer_init depends on zone->managed_pages, but it's reset
in reset_all_zones_managed_pages after that pages are readded into
zone->managed_pages, but when buffer_init runs this process is half done
and most of them will finally be added till deferred_init_memmap done. In
large memory couting of nr_free_buffer_pages drifts too much, also
drifting from kernels to kernels on same hardware.
Fix is simple, it delays buffer_init run till deferred_init_memmap all
done.
But as corrected by this patch, max_buffer_heads becomes very large, the
value is roughly as many as 4 times of totalram_pages, formula:
max_buffer_heads = nrpages * (10%) * (PAGE_SIZE / sizeof(struct
buffer_head));
Say in a 64GB memory box we have 16777216 pages, then max_buffer_heads
turns out to be roughly 67,108,864. In common cases, should a buffer_head
be mapped to one page/block(4KB)? So max_buffer_heads never exceeds
totalram_pages. IMO it's likely to make buffer_heads_over_limit bool
value alwasy false, then make codes 'if (buffer_heads_over_limit)' test in
vmscan unnecessary.
So this patch will change the original behavior related to
buffer_heads_over_limit in vmscan since we used a half done value of
zone->managed_pages before, or should we use a smaller factor(<10%) in
previous formula.
akpm: I think this is OK - the max_buffer_heads code is only needed on
highmem machines, to prevent ZONE_NORMAL from being consumed by large
amounts of buffer_heads attached to highmem pagecache. This problem will
not occur on 64-bit machines, so this feature's non-functionality on such
machines is a feature, not a bug.
Link: https://lkml.kernel.org/r/20201123110500.103523-1-linf@wangsu.com
Signed-off-by: Lin Feng <linf@wangsu.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:11:19 +00:00
|
|
|
buffer_init();
|
|
|
|
|
2017-08-18 22:16:05 +00:00
|
|
|
/* Discard memblock private memory */
|
|
|
|
memblock_discard();
|
2016-03-15 21:57:51 +00:00
|
|
|
|
mm: shuffle initial free memory to improve memory-side-cache utilization
Patch series "mm: Randomize free memory", v10.
This patch (of 3):
Randomization of the page allocator improves the average utilization of
a direct-mapped memory-side-cache. Memory side caching is a platform
capability that Linux has been previously exposed to in HPC
(high-performance computing) environments on specialty platforms. In
that instance it was a smaller pool of high-bandwidth-memory relative to
higher-capacity / lower-bandwidth DRAM. Now, this capability is going
to be found on general purpose server platforms where DRAM is a cache in
front of higher latency persistent memory [1].
Robert offered an explanation of the state of the art of Linux
interactions with memory-side-caches [2], and I copy it here:
It's been a problem in the HPC space:
http://www.nersc.gov/research-and-development/knl-cache-mode-performance-coe/
A kernel module called zonesort is available to try to help:
https://software.intel.com/en-us/articles/xeon-phi-software
and this abandoned patch series proposed that for the kernel:
https://lkml.kernel.org/r/20170823100205.17311-1-lukasz.daniluk@intel.com
Dan's patch series doesn't attempt to ensure buffers won't conflict, but
also reduces the chance that the buffers will. This will make performance
more consistent, albeit slower than "optimal" (which is near impossible
to attain in a general-purpose kernel). That's better than forcing
users to deploy remedies like:
"To eliminate this gradual degradation, we have added a Stream
measurement to the Node Health Check that follows each job;
nodes are rebooted whenever their measured memory bandwidth
falls below 300 GB/s."
A replacement for zonesort was merged upstream in commit cc9aec03e58f
("x86/numa_emulation: Introduce uniform split capability"). With this
numa_emulation capability, memory can be split into cache sized
("near-memory" sized) numa nodes. A bind operation to such a node, and
disabling workloads on other nodes, enables full cache performance.
However, once the workload exceeds the cache size then cache conflicts
are unavoidable. While HPC environments might be able to tolerate
time-scheduling of cache sized workloads, for general purpose server
platforms, the oversubscribed cache case will be the common case.
The worst case scenario is that a server system owner benchmarks a
workload at boot with an un-contended cache only to see that performance
degrade over time, even below the average cache performance due to
excessive conflicts. Randomization clips the peaks and fills in the
valleys of cache utilization to yield steady average performance.
Here are some performance impact details of the patches:
1/ An Intel internal synthetic memory bandwidth measurement tool, saw a
3X speedup in a contrived case that tries to force cache conflicts.
The contrived cased used the numa_emulation capability to force an
instance of the benchmark to be run in two of the near-memory sized
numa nodes. If both instances were placed on the same emulated they
would fit and cause zero conflicts. While on separate emulated nodes
without randomization they underutilized the cache and conflicted
unnecessarily due to the in-order allocation per node.
2/ A well known Java server application benchmark was run with a heap
size that exceeded cache size by 3X. The cache conflict rate was 8%
for the first run and degraded to 21% after page allocator aging. With
randomization enabled the rate levelled out at 11%.
3/ A MongoDB workload did not observe measurable difference in
cache-conflict rates, but the overall throughput dropped by 7% with
randomization in one case.
4/ Mel Gorman ran his suite of performance workloads with randomization
enabled on platforms without a memory-side-cache and saw a mix of some
improvements and some losses [3].
While there is potentially significant improvement for applications that
depend on low latency access across a wide working-set, the performance
may be negligible to negative for other workloads. For this reason the
shuffle capability defaults to off unless a direct-mapped
memory-side-cache is detected. Even then, the page_alloc.shuffle=0
parameter can be specified to disable the randomization on those systems.
Outside of memory-side-cache utilization concerns there is potentially
security benefit from randomization. Some data exfiltration and
return-oriented-programming attacks rely on the ability to infer the
location of sensitive data objects. The kernel page allocator, especially
early in system boot, has predictable first-in-first out behavior for
physical pages. Pages are freed in physical address order when first
onlined.
Quoting Kees:
"While we already have a base-address randomization
(CONFIG_RANDOMIZE_MEMORY), attacks against the same hardware and
memory layouts would certainly be using the predictability of
allocation ordering (i.e. for attacks where the base address isn't
important: only the relative positions between allocated memory).
This is common in lots of heap-style attacks. They try to gain
control over ordering by spraying allocations, etc.
I'd really like to see this because it gives us something similar
to CONFIG_SLAB_FREELIST_RANDOM but for the page allocator."
While SLAB_FREELIST_RANDOM reduces the predictability of some local slab
caches it leaves vast bulk of memory to be predictably in order allocated.
However, it should be noted, the concrete security benefits are hard to
quantify, and no known CVE is mitigated by this randomization.
Introduce shuffle_free_memory(), and its helper shuffle_zone(), to perform
a Fisher-Yates shuffle of the page allocator 'free_area' lists when they
are initially populated with free memory at boot and at hotplug time. Do
this based on either the presence of a page_alloc.shuffle=Y command line
parameter, or autodetection of a memory-side-cache (to be added in a
follow-on patch).
The shuffling is done in terms of CONFIG_SHUFFLE_PAGE_ORDER sized free
pages where the default CONFIG_SHUFFLE_PAGE_ORDER is MAX_ORDER-1 i.e. 10,
4MB this trades off randomization granularity for time spent shuffling.
MAX_ORDER-1 was chosen to be minimally invasive to the page allocator
while still showing memory-side cache behavior improvements, and the
expectation that the security implications of finer granularity
randomization is mitigated by CONFIG_SLAB_FREELIST_RANDOM. The
performance impact of the shuffling appears to be in the noise compared to
other memory initialization work.
This initial randomization can be undone over time so a follow-on patch is
introduced to inject entropy on page free decisions. It is reasonable to
ask if the page free entropy is sufficient, but it is not enough due to
the in-order initial freeing of pages. At the start of that process
putting page1 in front or behind page0 still keeps them close together,
page2 is still near page1 and has a high chance of being adjacent. As
more pages are added ordering diversity improves, but there is still high
page locality for the low address pages and this leads to no significant
impact to the cache conflict rate.
[1]: https://itpeernetwork.intel.com/intel-optane-dc-persistent-memory-operating-modes/
[2]: https://lkml.kernel.org/r/AT5PR8401MB1169D656C8B5E121752FC0F8AB120@AT5PR8401MB1169.NAMPRD84.PROD.OUTLOOK.COM
[3]: https://lkml.org/lkml/2018/10/12/309
[dan.j.williams@intel.com: fix shuffle enable]
Link: http://lkml.kernel.org/r/154943713038.3858443.4125180191382062871.stgit@dwillia2-desk3.amr.corp.intel.com
[cai@lca.pw: fix SHUFFLE_PAGE_ALLOCATOR help texts]
Link: http://lkml.kernel.org/r/20190425201300.75650-1-cai@lca.pw
Link: http://lkml.kernel.org/r/154899811738.3165233.12325692939590944259.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Robert Elliott <elliott@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 22:41:28 +00:00
|
|
|
for_each_node_state(nid, N_MEMORY)
|
|
|
|
shuffle_free_memory(NODE_DATA(nid));
|
|
|
|
|
2016-03-15 21:57:51 +00:00
|
|
|
for_each_populated_zone(zone)
|
|
|
|
set_zone_contiguous(zone);
|
2015-06-30 21:57:05 +00:00
|
|
|
}
|
|
|
|
|
2011-12-29 12:09:50 +00:00
|
|
|
#ifdef CONFIG_CMA
|
2013-08-23 05:52:52 +00:00
|
|
|
/* Free whole pageblock and set its migration type to MIGRATE_CMA. */
|
2011-12-29 12:09:50 +00:00
|
|
|
void __init init_cma_reserved_pageblock(struct page *page)
|
|
|
|
{
|
|
|
|
unsigned i = pageblock_nr_pages;
|
|
|
|
struct page *p = page;
|
|
|
|
|
|
|
|
do {
|
|
|
|
__ClearPageReserved(p);
|
|
|
|
set_page_count(p, 0);
|
2018-05-23 01:18:21 +00:00
|
|
|
} while (++p, --i);
|
2011-12-29 12:09:50 +00:00
|
|
|
|
|
|
|
set_pageblock_migratetype(page, MIGRATE_CMA);
|
2014-07-02 22:22:35 +00:00
|
|
|
|
|
|
|
if (pageblock_order >= MAX_ORDER) {
|
|
|
|
i = pageblock_nr_pages;
|
|
|
|
p = page;
|
|
|
|
do {
|
|
|
|
set_page_refcounted(p);
|
|
|
|
__free_pages(p, MAX_ORDER - 1);
|
|
|
|
p += MAX_ORDER_NR_PAGES;
|
|
|
|
} while (i -= MAX_ORDER_NR_PAGES);
|
|
|
|
} else {
|
|
|
|
set_page_refcounted(page);
|
|
|
|
__free_pages(page, pageblock_order);
|
|
|
|
}
|
|
|
|
|
2013-07-03 22:03:21 +00:00
|
|
|
adjust_managed_page_count(page, pageblock_nr_pages);
|
2021-02-26 01:16:40 +00:00
|
|
|
page_zone(page)->cma_pages += pageblock_nr_pages;
|
2011-12-29 12:09:50 +00:00
|
|
|
}
|
|
|
|
#endif
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The order of subdivision here is critical for the IO subsystem.
|
|
|
|
* Please do not alter this order without good reasons and regression
|
|
|
|
* testing. Specifically, as large blocks of memory are subdivided,
|
|
|
|
* the order in which smaller blocks are delivered depends on the order
|
|
|
|
* they're subdivided in this function. This is the primary factor
|
|
|
|
* influencing the order in which pages are delivered to the IO
|
|
|
|
* subsystem according to empirical testing, and this is also justified
|
|
|
|
* by considering the behavior of a buddy system containing a single
|
|
|
|
* large block of memory acted on by a series of small allocations.
|
|
|
|
* This behavior is a critical factor in sglist merging's success.
|
|
|
|
*
|
2012-12-06 09:39:54 +00:00
|
|
|
* -- nyc
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2006-01-06 08:11:01 +00:00
|
|
|
static inline void expand(struct zone *zone, struct page *page,
|
2020-04-07 03:04:49 +00:00
|
|
|
int low, int high, int migratetype)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
unsigned long size = 1 << high;
|
|
|
|
|
|
|
|
while (high > low) {
|
|
|
|
high--;
|
|
|
|
size >>= 1;
|
2014-01-23 23:52:54 +00:00
|
|
|
VM_BUG_ON_PAGE(bad_range(zone, &page[size]), &page[size]);
|
2012-01-10 23:07:28 +00:00
|
|
|
|
2016-10-07 23:58:15 +00:00
|
|
|
/*
|
|
|
|
* Mark as guard pages (or page), that will allow to
|
|
|
|
* merge back to allocator when buddy will be freed.
|
|
|
|
* Corresponding page table entries will not be touched,
|
|
|
|
* pages will stay not present in virtual address space
|
|
|
|
*/
|
|
|
|
if (set_page_guard(zone, &page[size], high, migratetype))
|
2012-01-10 23:07:28 +00:00
|
|
|
continue;
|
2016-10-07 23:58:15 +00:00
|
|
|
|
2020-04-07 03:04:49 +00:00
|
|
|
add_to_free_list(&page[size], zone, high, migratetype);
|
2020-10-16 03:10:15 +00:00
|
|
|
set_buddy_order(&page[size], high);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:41 +00:00
|
|
|
static void check_new_page_bad(struct page *page)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
mm: check __PG_HWPOISON separately from PAGE_FLAGS_CHECK_AT_*
The race condition addressed in commit add05cecef80 ("mm: soft-offline:
don't free target page in successful page migration") was not closed
completely, because that can happen not only for soft-offline, but also
for hard-offline. Consider that a slab page is about to be freed into
buddy pool, and then an uncorrected memory error hits the page just
after entering __free_one_page(), then VM_BUG_ON_PAGE(page->flags &
PAGE_FLAGS_CHECK_AT_PREP) is triggered, despite the fact that it's not
necessary because the data on the affected page is not consumed.
To solve it, this patch drops __PG_HWPOISON from page flag checks at
allocation/free time. I think it's justified because __PG_HWPOISON
flags is defined to prevent the page from being reused, and setting it
outside the page's alloc-free cycle is a designed behavior (not a bug.)
For recent months, I was annoyed about BUG_ON when soft-offlined page
remains on lru cache list for a while, which is avoided by calling
put_page() instead of putback_lru_page() in page migration's success
path. This means that this patch reverts a major change from commit
add05cecef80 about the new refcounting rule of soft-offlined pages, so
"reuse window" revives. This will be closed by a subsequent patch.
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Dean Nelson <dnelson@redhat.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-08-06 22:47:08 +00:00
|
|
|
if (unlikely(page->flags & __PG_HWPOISON)) {
|
2016-05-20 23:58:50 +00:00
|
|
|
/* Don't complain about hwpoisoned pages */
|
|
|
|
page_mapcount_reset(page); /* remove PageBuddy */
|
|
|
|
return;
|
mm: check __PG_HWPOISON separately from PAGE_FLAGS_CHECK_AT_*
The race condition addressed in commit add05cecef80 ("mm: soft-offline:
don't free target page in successful page migration") was not closed
completely, because that can happen not only for soft-offline, but also
for hard-offline. Consider that a slab page is about to be freed into
buddy pool, and then an uncorrected memory error hits the page just
after entering __free_one_page(), then VM_BUG_ON_PAGE(page->flags &
PAGE_FLAGS_CHECK_AT_PREP) is triggered, despite the fact that it's not
necessary because the data on the affected page is not consumed.
To solve it, this patch drops __PG_HWPOISON from page flag checks at
allocation/free time. I think it's justified because __PG_HWPOISON
flags is defined to prevent the page from being reused, and setting it
outside the page's alloc-free cycle is a designed behavior (not a bug.)
For recent months, I was annoyed about BUG_ON when soft-offlined page
remains on lru cache list for a while, which is avoided by calling
put_page() instead of putback_lru_page() in page migration's success
path. This means that this patch reverts a major change from commit
add05cecef80 about the new refcounting rule of soft-offlined pages, so
"reuse window" revives. This will be closed by a subsequent patch.
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Dean Nelson <dnelson@redhat.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-08-06 22:47:08 +00:00
|
|
|
}
|
2020-06-03 22:58:39 +00:00
|
|
|
|
|
|
|
bad_page(page,
|
|
|
|
page_bad_reason(page, PAGE_FLAGS_CHECK_AT_PREP));
|
2016-05-20 00:14:41 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This page is about to be returned from the page allocator
|
|
|
|
*/
|
|
|
|
static inline int check_new_page(struct page *page)
|
|
|
|
{
|
|
|
|
if (likely(page_expected_state(page,
|
|
|
|
PAGE_FLAGS_CHECK_AT_PREP|__PG_HWPOISON)))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
check_new_page_bad(page);
|
|
|
|
return 1;
|
2009-09-16 09:50:12 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:35 +00:00
|
|
|
#ifdef CONFIG_DEBUG_VM
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
/*
|
|
|
|
* With DEBUG_VM enabled, order-0 pages are checked for expected state when
|
|
|
|
* being allocated from pcp lists. With debug_pagealloc also enabled, they are
|
|
|
|
* also checked when pcp lists are refilled from the free lists.
|
|
|
|
*/
|
|
|
|
static inline bool check_pcp_refill(struct page *page)
|
2016-05-20 00:14:35 +00:00
|
|
|
{
|
mm, debug_pagealloc: don't rely on static keys too early
Commit 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable
debugging") has introduced a static key to reduce overhead when
debug_pagealloc is compiled in but not enabled. It relied on the
assumption that jump_label_init() is called before parse_early_param()
as in start_kernel(), so when the "debug_pagealloc=on" option is parsed,
it is safe to enable the static key.
However, it turns out multiple architectures call parse_early_param()
earlier from their setup_arch(). x86 also calls jump_label_init() even
earlier, so no issue was found while testing the commit, but same is not
true for e.g. ppc64 and s390 where the kernel would not boot with
debug_pagealloc=on as found by our QA.
To fix this without tricky changes to init code of multiple
architectures, this patch partially reverts the static key conversion
from 96a2b03f281d. Init-time and non-fastpath calls (such as in arch
code) of debug_pagealloc_enabled() will again test a simple bool
variable. Fastpath mm code is converted to a new
debug_pagealloc_enabled_static() variant that relies on the static key,
which is enabled in a well-defined point in mm_init() where it's
guaranteed that jump_label_init() has been called, regardless of
architecture.
[sfr@canb.auug.org.au: export _debug_pagealloc_enabled_early]
Link: http://lkml.kernel.org/r/20200106164944.063ac07b@canb.auug.org.au
Link: http://lkml.kernel.org/r/20191219130612.23171-1-vbabka@suse.cz
Fixes: 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable debugging")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Qian Cai <cai@lca.pw>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-14 00:29:20 +00:00
|
|
|
if (debug_pagealloc_enabled_static())
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
return check_new_page(page);
|
|
|
|
else
|
|
|
|
return false;
|
2016-05-20 00:14:35 +00:00
|
|
|
}
|
|
|
|
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
static inline bool check_new_pcp(struct page *page)
|
2016-05-20 00:14:35 +00:00
|
|
|
{
|
|
|
|
return check_new_page(page);
|
|
|
|
}
|
|
|
|
#else
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
/*
|
|
|
|
* With DEBUG_VM disabled, free order-0 pages are checked for expected state
|
|
|
|
* when pcp lists are being refilled from the free lists. With debug_pagealloc
|
|
|
|
* enabled, they are also checked when being allocated from the pcp lists.
|
|
|
|
*/
|
|
|
|
static inline bool check_pcp_refill(struct page *page)
|
2016-05-20 00:14:35 +00:00
|
|
|
{
|
|
|
|
return check_new_page(page);
|
|
|
|
}
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
static inline bool check_new_pcp(struct page *page)
|
2016-05-20 00:14:35 +00:00
|
|
|
{
|
mm, debug_pagealloc: don't rely on static keys too early
Commit 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable
debugging") has introduced a static key to reduce overhead when
debug_pagealloc is compiled in but not enabled. It relied on the
assumption that jump_label_init() is called before parse_early_param()
as in start_kernel(), so when the "debug_pagealloc=on" option is parsed,
it is safe to enable the static key.
However, it turns out multiple architectures call parse_early_param()
earlier from their setup_arch(). x86 also calls jump_label_init() even
earlier, so no issue was found while testing the commit, but same is not
true for e.g. ppc64 and s390 where the kernel would not boot with
debug_pagealloc=on as found by our QA.
To fix this without tricky changes to init code of multiple
architectures, this patch partially reverts the static key conversion
from 96a2b03f281d. Init-time and non-fastpath calls (such as in arch
code) of debug_pagealloc_enabled() will again test a simple bool
variable. Fastpath mm code is converted to a new
debug_pagealloc_enabled_static() variant that relies on the static key,
which is enabled in a well-defined point in mm_init() where it's
guaranteed that jump_label_init() has been called, regardless of
architecture.
[sfr@canb.auug.org.au: export _debug_pagealloc_enabled_early]
Link: http://lkml.kernel.org/r/20200106164944.063ac07b@canb.auug.org.au
Link: http://lkml.kernel.org/r/20191219130612.23171-1-vbabka@suse.cz
Fixes: 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable debugging")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Qian Cai <cai@lca.pw>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-14 00:29:20 +00:00
|
|
|
if (debug_pagealloc_enabled_static())
|
mm, page_alloc: more extensive free page checking with debug_pagealloc
The page allocator checks struct pages for expected state (mapcount,
flags etc) as pages are being allocated (check_new_page()) and freed
(free_pages_check()) to provide some defense against errors in page
allocator users.
Prior commits 479f854a207c ("mm, page_alloc: defer debugging checks of
pages allocated from the PCP") and 4db7548ccbd9 ("mm, page_alloc: defer
debugging checks of freed pages until a PCP drain") this has happened
for order-0 pages as they were allocated from or freed to the per-cpu
caches (pcplists). Since those are fast paths, the checks are now
performed only when pages are moved between pcplists and global free
lists. This however lowers the chances of catching errors soon enough.
In order to increase the chances of the checks to catch errors, the
kernel has to be rebuilt with CONFIG_DEBUG_VM, which also enables
multiple other internal debug checks (VM_BUG_ON() etc), which is
suboptimal when the goal is to catch errors in mm users, not in mm code
itself.
To catch some wrong users of the page allocator we have
CONFIG_DEBUG_PAGEALLOC, which is designed to have virtually no overhead
unless enabled at boot time. Memory corruptions when writing to freed
pages have often the same underlying errors (use-after-free, double free)
as corrupting the corresponding struct pages, so this existing debugging
functionality is a good fit to extend by also perform struct page checks
at least as often as if CONFIG_DEBUG_VM was enabled.
Specifically, after this patch, when debug_pagealloc is enabled on boot,
and CONFIG_DEBUG_VM disabled, pages are checked when allocated from or
freed to the pcplists *in addition* to being moved between pcplists and
free lists. When both debug_pagealloc and CONFIG_DEBUG_VM are enabled,
pages are checked when being moved between pcplists and free lists *in
addition* to when allocated from or freed to the pcplists.
When debug_pagealloc is not enabled on boot, the overhead in fast paths
should be virtually none thanks to the use of static key.
Link: http://lkml.kernel.org/r/20190603143451.27353-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-07-12 03:55:09 +00:00
|
|
|
return check_new_page(page);
|
|
|
|
else
|
|
|
|
return false;
|
2016-05-20 00:14:35 +00:00
|
|
|
}
|
|
|
|
#endif /* CONFIG_DEBUG_VM */
|
|
|
|
|
|
|
|
static bool check_new_pages(struct page *page, unsigned int order)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
for (i = 0; i < (1 << order); i++) {
|
|
|
|
struct page *p = page + i;
|
|
|
|
|
|
|
|
if (unlikely(check_new_page(p)))
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2016-07-26 22:23:58 +00:00
|
|
|
inline void post_alloc_hook(struct page *page, unsigned int order,
|
|
|
|
gfp_t gfp_flags)
|
|
|
|
{
|
|
|
|
set_page_private(page, 0);
|
|
|
|
set_page_refcounted(page);
|
|
|
|
|
|
|
|
arch_alloc_page(page, order);
|
mm: introduce debug_pagealloc_{map,unmap}_pages() helpers
Patch series "arch, mm: improve robustness of direct map manipulation", v7.
During recent discussion about KVM protected memory, David raised a
concern about usage of __kernel_map_pages() outside of DEBUG_PAGEALLOC
scope [1].
Indeed, for architectures that define CONFIG_ARCH_HAS_SET_DIRECT_MAP it is
possible that __kernel_map_pages() would fail, but since this function is
void, the failure will go unnoticed.
Moreover, there's lack of consistency of __kernel_map_pages() semantics
across architectures as some guard this function with #ifdef
DEBUG_PAGEALLOC, some refuse to update the direct map if page allocation
debugging is disabled at run time and some allow modifying the direct map
regardless of DEBUG_PAGEALLOC settings.
This set straightens this out by restoring dependency of
__kernel_map_pages() on DEBUG_PAGEALLOC and updating the call sites
accordingly.
Since currently the only user of __kernel_map_pages() outside
DEBUG_PAGEALLOC is hibernation, it is updated to make direct map accesses
there more explicit.
[1] https://lore.kernel.org/lkml/2759b4bf-e1e3-d006-7d86-78a40348269d@redhat.com
This patch (of 4):
When CONFIG_DEBUG_PAGEALLOC is enabled, it unmaps pages from the kernel
direct mapping after free_pages(). The pages than need to be mapped back
before they could be used. Theese mapping operations use
__kernel_map_pages() guarded with with debug_pagealloc_enabled().
The only place that calls __kernel_map_pages() without checking whether
DEBUG_PAGEALLOC is enabled is the hibernation code that presumes
availability of this function when ARCH_HAS_SET_DIRECT_MAP is set. Still,
on arm64, __kernel_map_pages() will bail out when DEBUG_PAGEALLOC is not
enabled but set_direct_map_invalid_noflush() may render some pages not
present in the direct map and hibernation code won't be able to save such
pages.
To make page allocation debugging and hibernation interaction more robust,
the dependency on DEBUG_PAGEALLOC or ARCH_HAS_SET_DIRECT_MAP has to be
made more explicit.
Start with combining the guard condition and the call to
__kernel_map_pages() into debug_pagealloc_map_pages() and
debug_pagealloc_unmap_pages() functions to emphasize that
__kernel_map_pages() should not be called without DEBUG_PAGEALLOC and use
these new functions to map/unmap pages when page allocation debugging is
enabled.
Link: https://lkml.kernel.org/r/20201109192128.960-1-rppt@kernel.org
Link: https://lkml.kernel.org/r/20201109192128.960-2-rppt@kernel.org
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: "Edgecombe, Rick P" <rick.p.edgecombe@intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:20 +00:00
|
|
|
debug_pagealloc_map_pages(page, 1 << order);
|
2016-07-26 22:23:58 +00:00
|
|
|
kasan_alloc_pages(page, order);
|
2020-12-15 03:13:34 +00:00
|
|
|
kernel_unpoison_pages(page, 1 << order);
|
2016-07-26 22:23:58 +00:00
|
|
|
set_page_owner(page, order, gfp_flags);
|
2020-12-15 03:11:15 +00:00
|
|
|
|
2020-12-15 03:13:45 +00:00
|
|
|
if (!want_init_on_free() && want_init_on_alloc(gfp_flags))
|
2020-12-15 03:11:15 +00:00
|
|
|
kernel_init_free_pages(page, 1 << order);
|
2016-07-26 22:23:58 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:35 +00:00
|
|
|
static void prep_new_page(struct page *page, unsigned int order, gfp_t gfp_flags,
|
2016-05-20 00:13:38 +00:00
|
|
|
unsigned int alloc_flags)
|
2009-09-16 09:50:12 +00:00
|
|
|
{
|
2016-07-26 22:23:58 +00:00
|
|
|
post_alloc_hook(page, order, gfp_flags);
|
2006-03-22 08:08:41 +00:00
|
|
|
|
|
|
|
if (order && (gfp_flags & __GFP_COMP))
|
|
|
|
prep_compound_page(page, order);
|
|
|
|
|
mm: set page->pfmemalloc in prep_new_page()
The possibility of replacing the numerous parameters of alloc_pages*
functions with a single structure has been discussed when Minchan proposed
to expand the x86 kernel stack [1]. This series implements the change,
along with few more cleanups/microoptimizations.
The series is based on next-20150108 and I used gcc 4.8.3 20140627 on
openSUSE 13.2 for compiling. Config includess NUMA and COMPACTION.
The core change is the introduction of a new struct alloc_context, which looks
like this:
struct alloc_context {
struct zonelist *zonelist;
nodemask_t *nodemask;
struct zone *preferred_zone;
int classzone_idx;
int migratetype;
enum zone_type high_zoneidx;
};
All the contents is mostly constant, except that __alloc_pages_slowpath()
changes preferred_zone, classzone_idx and potentially zonelist. But
that's not a problem in case control returns to retry_cpuset: in
__alloc_pages_nodemask(), those will be reset to initial values again
(although it's a bit subtle). On the other hand, gfp_flags and alloc_info
mutate so much that it doesn't make sense to put them into alloc_context.
Still, the result is one parameter instead of up to 7. This is all in
Patch 2.
Patch 3 is a step to expand alloc_context usage out of page_alloc.c
itself. The function try_to_compact_pages() can also much benefit from
the parameter reduction, but it means the struct definition has to be
moved to a shared header.
Patch 1 should IMHO be included even if the rest is deemed not useful
enough. It improves maintainability and also has some code/stack
reduction. Patch 4 is OTOH a tiny optimization.
Overall bloat-o-meter results:
add/remove: 0/0 grow/shrink: 0/4 up/down: 0/-460 (-460)
function old new delta
nr_free_zone_pages 129 115 -14
__alloc_pages_direct_compact 329 256 -73
get_page_from_freelist 2670 2576 -94
__alloc_pages_nodemask 2564 2285 -279
try_to_compact_pages 582 579 -3
Overall stack sizes per ./scripts/checkstack.pl:
old new delta
get_page_from_freelist: 184 184 0
__alloc_pages_nodemask 248 200 -48
__alloc_pages_direct_c 40 - -40
try_to_compact_pages 72 72 0
-88
[1] http://marc.info/?l=linux-mm&m=140142462528257&w=2
This patch (of 4):
prep_new_page() sets almost everything in the struct page of the page
being allocated, except page->pfmemalloc. This is not obvious and has at
least once led to a bug where page->pfmemalloc was forgotten to be set
correctly, see commit 8fb74b9fb2b1 ("mm: compaction: partially revert
capture of suitable high-order page").
This patch moves the pfmemalloc setting to prep_new_page(), which means it
needs to gain alloc_flags parameter. The call to prep_new_page is moved
from buffered_rmqueue() to get_page_from_freelist(), which also leads to
simpler code. An obsolete comment for buffered_rmqueue() is replaced.
In addition to better maintainability there is a small reduction of code
and stack usage for get_page_from_freelist(), which inlines the other
functions involved.
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-145 (-145)
function old new delta
get_page_from_freelist 2670 2525 -145
Stack usage is reduced from 184 to 168 bytes.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:25:38 +00:00
|
|
|
/*
|
2015-08-21 21:11:51 +00:00
|
|
|
* page is set pfmemalloc when ALLOC_NO_WATERMARKS was necessary to
|
mm: set page->pfmemalloc in prep_new_page()
The possibility of replacing the numerous parameters of alloc_pages*
functions with a single structure has been discussed when Minchan proposed
to expand the x86 kernel stack [1]. This series implements the change,
along with few more cleanups/microoptimizations.
The series is based on next-20150108 and I used gcc 4.8.3 20140627 on
openSUSE 13.2 for compiling. Config includess NUMA and COMPACTION.
The core change is the introduction of a new struct alloc_context, which looks
like this:
struct alloc_context {
struct zonelist *zonelist;
nodemask_t *nodemask;
struct zone *preferred_zone;
int classzone_idx;
int migratetype;
enum zone_type high_zoneidx;
};
All the contents is mostly constant, except that __alloc_pages_slowpath()
changes preferred_zone, classzone_idx and potentially zonelist. But
that's not a problem in case control returns to retry_cpuset: in
__alloc_pages_nodemask(), those will be reset to initial values again
(although it's a bit subtle). On the other hand, gfp_flags and alloc_info
mutate so much that it doesn't make sense to put them into alloc_context.
Still, the result is one parameter instead of up to 7. This is all in
Patch 2.
Patch 3 is a step to expand alloc_context usage out of page_alloc.c
itself. The function try_to_compact_pages() can also much benefit from
the parameter reduction, but it means the struct definition has to be
moved to a shared header.
Patch 1 should IMHO be included even if the rest is deemed not useful
enough. It improves maintainability and also has some code/stack
reduction. Patch 4 is OTOH a tiny optimization.
Overall bloat-o-meter results:
add/remove: 0/0 grow/shrink: 0/4 up/down: 0/-460 (-460)
function old new delta
nr_free_zone_pages 129 115 -14
__alloc_pages_direct_compact 329 256 -73
get_page_from_freelist 2670 2576 -94
__alloc_pages_nodemask 2564 2285 -279
try_to_compact_pages 582 579 -3
Overall stack sizes per ./scripts/checkstack.pl:
old new delta
get_page_from_freelist: 184 184 0
__alloc_pages_nodemask 248 200 -48
__alloc_pages_direct_c 40 - -40
try_to_compact_pages 72 72 0
-88
[1] http://marc.info/?l=linux-mm&m=140142462528257&w=2
This patch (of 4):
prep_new_page() sets almost everything in the struct page of the page
being allocated, except page->pfmemalloc. This is not obvious and has at
least once led to a bug where page->pfmemalloc was forgotten to be set
correctly, see commit 8fb74b9fb2b1 ("mm: compaction: partially revert
capture of suitable high-order page").
This patch moves the pfmemalloc setting to prep_new_page(), which means it
needs to gain alloc_flags parameter. The call to prep_new_page is moved
from buffered_rmqueue() to get_page_from_freelist(), which also leads to
simpler code. An obsolete comment for buffered_rmqueue() is replaced.
In addition to better maintainability there is a small reduction of code
and stack usage for get_page_from_freelist(), which inlines the other
functions involved.
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-145 (-145)
function old new delta
get_page_from_freelist 2670 2525 -145
Stack usage is reduced from 184 to 168 bytes.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:25:38 +00:00
|
|
|
* allocate the page. The expectation is that the caller is taking
|
|
|
|
* steps that will free more memory. The caller should avoid the page
|
|
|
|
* being used for !PFMEMALLOC purposes.
|
|
|
|
*/
|
2015-08-21 21:11:51 +00:00
|
|
|
if (alloc_flags & ALLOC_NO_WATERMARKS)
|
|
|
|
set_page_pfmemalloc(page);
|
|
|
|
else
|
|
|
|
clear_page_pfmemalloc(page);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
Bias the location of pages freed for min_free_kbytes in the same MAX_ORDER_NR_PAGES blocks
The standard buddy allocator always favours the smallest block of pages.
The effect of this is that the pages free to satisfy min_free_kbytes tends
to be preserved since boot time at the same location of memory ffor a very
long time and as a contiguous block. When an administrator sets the
reserve at 16384 at boot time, it tends to be the same MAX_ORDER blocks
that remain free. This allows the occasional high atomic allocation to
succeed up until the point the blocks are split. In practice, it is
difficult to split these blocks but when they do split, the benefit of
having min_free_kbytes for contiguous blocks disappears. Additionally,
increasing min_free_kbytes once the system has been running for some time
has no guarantee of creating contiguous blocks.
On the other hand, CONFIG_PAGE_GROUP_BY_MOBILITY favours splitting large
blocks when there are no free pages of the appropriate type available. A
side-effect of this is that all blocks in memory tends to be used up and
the contiguous free blocks from boot time are not preserved like in the
vanilla allocator. This can cause a problem if a new caller is unwilling
to reclaim or does not reclaim for long enough.
A failure scenario was found for a wireless network device allocating
order-1 atomic allocations but the allocations were not intense or frequent
enough for a whole block of pages to be preserved for MIGRATE_HIGHALLOC.
This was reproduced on a desktop by booting with mem=256mb, forcing the
driver to allocate at order-1, running a bittorrent client (downloading a
debian ISO) and building a kernel with -j2.
This patch addresses the problem on the desktop machine booted with
mem=256mb. It works by setting aside a reserve of MAX_ORDER_NR_PAGES
blocks, the number of which depends on the value of min_free_kbytes. These
blocks are only fallen back to when there is no other free pages. Then the
smallest possible page is used just like the normal buddy allocator instead
of the largest possible page to preserve contiguous pages The pages in free
lists in the reserve blocks are never taken for another migrate type. The
results is that even if min_free_kbytes is set to a low value, contiguous
blocks will be preserved in the MIGRATE_RESERVE blocks.
This works better than the vanilla allocator because if min_free_kbytes is
increased, a new reserve block will be chosen based on the location of
reclaimable pages and the block will free up as contiguous pages. In the
vanilla allocator, no effort is made to target a block of pages to free as
contiguous pages and min_free_kbytes pages are scattered randomly.
This effect has been observed on the test machine. min_free_kbytes was set
initially low but it was kept as a contiguous free block within
MIGRATE_RESERVE. min_free_kbytes was then set to a higher value and over a
period of time, the free blocks were within the reserve and coalescing.
How long it takes to free up depends on how quickly LRU is rotating.
Amusingly, this means that more activity will free the blocks faster.
This mechanism potentially replaces MIGRATE_HIGHALLOC as it may be more
effective than grouping contiguous free pages together. It all depends on
whether the number of active atomic high allocations exceeds
min_free_kbytes or not. If the number of active allocations exceeds
min_free_kbytes, it's worth it but maybe in that situation, min_free_kbytes
should be set higher. Once there are no more reports of allocation
failures, a patch will be submitted that backs out MIGRATE_HIGHALLOC and
see if the reports stay missing.
Credit to Mariusz Kozlowski for discovering the problem, describing the
failure scenario and testing patches and scenarios.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:58 +00:00
|
|
|
/*
|
|
|
|
* Go through the free lists for the given migratetype and remove
|
|
|
|
* the smallest available page from the freelists
|
|
|
|
*/
|
2017-11-16 01:36:53 +00:00
|
|
|
static __always_inline
|
2009-06-16 22:32:04 +00:00
|
|
|
struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
|
Bias the location of pages freed for min_free_kbytes in the same MAX_ORDER_NR_PAGES blocks
The standard buddy allocator always favours the smallest block of pages.
The effect of this is that the pages free to satisfy min_free_kbytes tends
to be preserved since boot time at the same location of memory ffor a very
long time and as a contiguous block. When an administrator sets the
reserve at 16384 at boot time, it tends to be the same MAX_ORDER blocks
that remain free. This allows the occasional high atomic allocation to
succeed up until the point the blocks are split. In practice, it is
difficult to split these blocks but when they do split, the benefit of
having min_free_kbytes for contiguous blocks disappears. Additionally,
increasing min_free_kbytes once the system has been running for some time
has no guarantee of creating contiguous blocks.
On the other hand, CONFIG_PAGE_GROUP_BY_MOBILITY favours splitting large
blocks when there are no free pages of the appropriate type available. A
side-effect of this is that all blocks in memory tends to be used up and
the contiguous free blocks from boot time are not preserved like in the
vanilla allocator. This can cause a problem if a new caller is unwilling
to reclaim or does not reclaim for long enough.
A failure scenario was found for a wireless network device allocating
order-1 atomic allocations but the allocations were not intense or frequent
enough for a whole block of pages to be preserved for MIGRATE_HIGHALLOC.
This was reproduced on a desktop by booting with mem=256mb, forcing the
driver to allocate at order-1, running a bittorrent client (downloading a
debian ISO) and building a kernel with -j2.
This patch addresses the problem on the desktop machine booted with
mem=256mb. It works by setting aside a reserve of MAX_ORDER_NR_PAGES
blocks, the number of which depends on the value of min_free_kbytes. These
blocks are only fallen back to when there is no other free pages. Then the
smallest possible page is used just like the normal buddy allocator instead
of the largest possible page to preserve contiguous pages The pages in free
lists in the reserve blocks are never taken for another migrate type. The
results is that even if min_free_kbytes is set to a low value, contiguous
blocks will be preserved in the MIGRATE_RESERVE blocks.
This works better than the vanilla allocator because if min_free_kbytes is
increased, a new reserve block will be chosen based on the location of
reclaimable pages and the block will free up as contiguous pages. In the
vanilla allocator, no effort is made to target a block of pages to free as
contiguous pages and min_free_kbytes pages are scattered randomly.
This effect has been observed on the test machine. min_free_kbytes was set
initially low but it was kept as a contiguous free block within
MIGRATE_RESERVE. min_free_kbytes was then set to a higher value and over a
period of time, the free blocks were within the reserve and coalescing.
How long it takes to free up depends on how quickly LRU is rotating.
Amusingly, this means that more activity will free the blocks faster.
This mechanism potentially replaces MIGRATE_HIGHALLOC as it may be more
effective than grouping contiguous free pages together. It all depends on
whether the number of active atomic high allocations exceeds
min_free_kbytes or not. If the number of active allocations exceeds
min_free_kbytes, it's worth it but maybe in that situation, min_free_kbytes
should be set higher. Once there are no more reports of allocation
failures, a patch will be submitted that backs out MIGRATE_HIGHALLOC and
see if the reports stay missing.
Credit to Mariusz Kozlowski for discovering the problem, describing the
failure scenario and testing patches and scenarios.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:58 +00:00
|
|
|
int migratetype)
|
|
|
|
{
|
|
|
|
unsigned int current_order;
|
2013-09-11 21:20:34 +00:00
|
|
|
struct free_area *area;
|
Bias the location of pages freed for min_free_kbytes in the same MAX_ORDER_NR_PAGES blocks
The standard buddy allocator always favours the smallest block of pages.
The effect of this is that the pages free to satisfy min_free_kbytes tends
to be preserved since boot time at the same location of memory ffor a very
long time and as a contiguous block. When an administrator sets the
reserve at 16384 at boot time, it tends to be the same MAX_ORDER blocks
that remain free. This allows the occasional high atomic allocation to
succeed up until the point the blocks are split. In practice, it is
difficult to split these blocks but when they do split, the benefit of
having min_free_kbytes for contiguous blocks disappears. Additionally,
increasing min_free_kbytes once the system has been running for some time
has no guarantee of creating contiguous blocks.
On the other hand, CONFIG_PAGE_GROUP_BY_MOBILITY favours splitting large
blocks when there are no free pages of the appropriate type available. A
side-effect of this is that all blocks in memory tends to be used up and
the contiguous free blocks from boot time are not preserved like in the
vanilla allocator. This can cause a problem if a new caller is unwilling
to reclaim or does not reclaim for long enough.
A failure scenario was found for a wireless network device allocating
order-1 atomic allocations but the allocations were not intense or frequent
enough for a whole block of pages to be preserved for MIGRATE_HIGHALLOC.
This was reproduced on a desktop by booting with mem=256mb, forcing the
driver to allocate at order-1, running a bittorrent client (downloading a
debian ISO) and building a kernel with -j2.
This patch addresses the problem on the desktop machine booted with
mem=256mb. It works by setting aside a reserve of MAX_ORDER_NR_PAGES
blocks, the number of which depends on the value of min_free_kbytes. These
blocks are only fallen back to when there is no other free pages. Then the
smallest possible page is used just like the normal buddy allocator instead
of the largest possible page to preserve contiguous pages The pages in free
lists in the reserve blocks are never taken for another migrate type. The
results is that even if min_free_kbytes is set to a low value, contiguous
blocks will be preserved in the MIGRATE_RESERVE blocks.
This works better than the vanilla allocator because if min_free_kbytes is
increased, a new reserve block will be chosen based on the location of
reclaimable pages and the block will free up as contiguous pages. In the
vanilla allocator, no effort is made to target a block of pages to free as
contiguous pages and min_free_kbytes pages are scattered randomly.
This effect has been observed on the test machine. min_free_kbytes was set
initially low but it was kept as a contiguous free block within
MIGRATE_RESERVE. min_free_kbytes was then set to a higher value and over a
period of time, the free blocks were within the reserve and coalescing.
How long it takes to free up depends on how quickly LRU is rotating.
Amusingly, this means that more activity will free the blocks faster.
This mechanism potentially replaces MIGRATE_HIGHALLOC as it may be more
effective than grouping contiguous free pages together. It all depends on
whether the number of active atomic high allocations exceeds
min_free_kbytes or not. If the number of active allocations exceeds
min_free_kbytes, it's worth it but maybe in that situation, min_free_kbytes
should be set higher. Once there are no more reports of allocation
failures, a patch will be submitted that backs out MIGRATE_HIGHALLOC and
see if the reports stay missing.
Credit to Mariusz Kozlowski for discovering the problem, describing the
failure scenario and testing patches and scenarios.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:58 +00:00
|
|
|
struct page *page;
|
|
|
|
|
|
|
|
/* Find a page of the appropriate size in the preferred list */
|
|
|
|
for (current_order = order; current_order < MAX_ORDER; ++current_order) {
|
|
|
|
area = &(zone->free_area[current_order]);
|
2019-05-14 22:41:32 +00:00
|
|
|
page = get_page_from_free_area(area, migratetype);
|
2016-01-14 23:20:30 +00:00
|
|
|
if (!page)
|
|
|
|
continue;
|
2020-04-07 03:04:49 +00:00
|
|
|
del_page_from_free_list(page, zone, current_order);
|
|
|
|
expand(zone, page, order, current_order, migratetype);
|
2015-09-08 22:01:25 +00:00
|
|
|
set_pcppage_migratetype(page, migratetype);
|
Bias the location of pages freed for min_free_kbytes in the same MAX_ORDER_NR_PAGES blocks
The standard buddy allocator always favours the smallest block of pages.
The effect of this is that the pages free to satisfy min_free_kbytes tends
to be preserved since boot time at the same location of memory ffor a very
long time and as a contiguous block. When an administrator sets the
reserve at 16384 at boot time, it tends to be the same MAX_ORDER blocks
that remain free. This allows the occasional high atomic allocation to
succeed up until the point the blocks are split. In practice, it is
difficult to split these blocks but when they do split, the benefit of
having min_free_kbytes for contiguous blocks disappears. Additionally,
increasing min_free_kbytes once the system has been running for some time
has no guarantee of creating contiguous blocks.
On the other hand, CONFIG_PAGE_GROUP_BY_MOBILITY favours splitting large
blocks when there are no free pages of the appropriate type available. A
side-effect of this is that all blocks in memory tends to be used up and
the contiguous free blocks from boot time are not preserved like in the
vanilla allocator. This can cause a problem if a new caller is unwilling
to reclaim or does not reclaim for long enough.
A failure scenario was found for a wireless network device allocating
order-1 atomic allocations but the allocations were not intense or frequent
enough for a whole block of pages to be preserved for MIGRATE_HIGHALLOC.
This was reproduced on a desktop by booting with mem=256mb, forcing the
driver to allocate at order-1, running a bittorrent client (downloading a
debian ISO) and building a kernel with -j2.
This patch addresses the problem on the desktop machine booted with
mem=256mb. It works by setting aside a reserve of MAX_ORDER_NR_PAGES
blocks, the number of which depends on the value of min_free_kbytes. These
blocks are only fallen back to when there is no other free pages. Then the
smallest possible page is used just like the normal buddy allocator instead
of the largest possible page to preserve contiguous pages The pages in free
lists in the reserve blocks are never taken for another migrate type. The
results is that even if min_free_kbytes is set to a low value, contiguous
blocks will be preserved in the MIGRATE_RESERVE blocks.
This works better than the vanilla allocator because if min_free_kbytes is
increased, a new reserve block will be chosen based on the location of
reclaimable pages and the block will free up as contiguous pages. In the
vanilla allocator, no effort is made to target a block of pages to free as
contiguous pages and min_free_kbytes pages are scattered randomly.
This effect has been observed on the test machine. min_free_kbytes was set
initially low but it was kept as a contiguous free block within
MIGRATE_RESERVE. min_free_kbytes was then set to a higher value and over a
period of time, the free blocks were within the reserve and coalescing.
How long it takes to free up depends on how quickly LRU is rotating.
Amusingly, this means that more activity will free the blocks faster.
This mechanism potentially replaces MIGRATE_HIGHALLOC as it may be more
effective than grouping contiguous free pages together. It all depends on
whether the number of active atomic high allocations exceeds
min_free_kbytes or not. If the number of active allocations exceeds
min_free_kbytes, it's worth it but maybe in that situation, min_free_kbytes
should be set higher. Once there are no more reports of allocation
failures, a patch will be submitted that backs out MIGRATE_HIGHALLOC and
see if the reports stay missing.
Credit to Mariusz Kozlowski for discovering the problem, describing the
failure scenario and testing patches and scenarios.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:58 +00:00
|
|
|
return page;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2007-10-16 08:25:48 +00:00
|
|
|
/*
|
|
|
|
* This array describes the order lists are fallen back to when
|
|
|
|
* the free lists for the desirable migrate type are depleted
|
|
|
|
*/
|
2020-08-07 06:25:58 +00:00
|
|
|
static int fallbacks[MIGRATE_TYPES][3] = {
|
2015-11-07 00:28:34 +00:00
|
|
|
[MIGRATE_UNMOVABLE] = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE, MIGRATE_TYPES },
|
|
|
|
[MIGRATE_MOVABLE] = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_TYPES },
|
2018-12-28 08:34:46 +00:00
|
|
|
[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE, MIGRATE_MOVABLE, MIGRATE_TYPES },
|
2011-12-29 12:09:50 +00:00
|
|
|
#ifdef CONFIG_CMA
|
2015-11-07 00:28:34 +00:00
|
|
|
[MIGRATE_CMA] = { MIGRATE_TYPES }, /* Never used */
|
2011-12-29 12:09:50 +00:00
|
|
|
#endif
|
2013-02-23 00:33:58 +00:00
|
|
|
#ifdef CONFIG_MEMORY_ISOLATION
|
2015-11-07 00:28:34 +00:00
|
|
|
[MIGRATE_ISOLATE] = { MIGRATE_TYPES }, /* Never used */
|
2013-02-23 00:33:58 +00:00
|
|
|
#endif
|
2007-10-16 08:25:48 +00:00
|
|
|
};
|
|
|
|
|
2015-04-14 22:45:15 +00:00
|
|
|
#ifdef CONFIG_CMA
|
2017-11-16 01:36:53 +00:00
|
|
|
static __always_inline struct page *__rmqueue_cma_fallback(struct zone *zone,
|
2015-04-14 22:45:15 +00:00
|
|
|
unsigned int order)
|
|
|
|
{
|
|
|
|
return __rmqueue_smallest(zone, order, MIGRATE_CMA);
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static inline struct page *__rmqueue_cma_fallback(struct zone *zone,
|
|
|
|
unsigned int order) { return NULL; }
|
|
|
|
#endif
|
|
|
|
|
2007-10-16 08:25:51 +00:00
|
|
|
/*
|
mm/page_alloc: move pages to tail in move_to_free_list()
Whenever we move pages between freelists via move_to_free_list()/
move_freepages_block(), we don't actually touch the pages:
1. Page isolation doesn't actually touch the pages, it simply isolates
pageblocks and moves all free pages to the MIGRATE_ISOLATE freelist.
When undoing isolation, we move the pages back to the target list.
2. Page stealing (steal_suitable_fallback()) moves free pages directly
between lists without touching them.
3. reserve_highatomic_pageblock()/unreserve_highatomic_pageblock() moves
free pages directly between freelists without touching them.
We already place pages to the tail of the freelists when undoing isolation
via __putback_isolated_page(), let's do it in any case (e.g., if order <=
pageblock_order) and document the behavior. To simplify, let's move the
pages to the tail for all move_to_free_list()/move_freepages_block() users.
In 2., the target list is empty, so there should be no change. In 3., we
might observe a change, however, highatomic is more concerned about
allocations succeeding than cache hotness - if we ever realize this change
degrades a workload, we can special-case this instance and add a proper
comment.
This change results in all pages getting onlined via online_pages() to be
placed to the tail of the freelist.
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Acked-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-4-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:30 +00:00
|
|
|
* Move the free pages in a range to the freelist tail of the requested type.
|
2007-10-16 08:26:01 +00:00
|
|
|
* Note that start_page and end_pages are not aligned on a pageblock
|
2007-10-16 08:25:51 +00:00
|
|
|
* boundary. If alignment is required, use move_freepages_block()
|
|
|
|
*/
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
static int move_freepages(struct zone *zone,
|
2008-07-24 04:28:12 +00:00
|
|
|
struct page *start_page, struct page *end_page,
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
int migratetype, int *num_movable)
|
2007-10-16 08:25:51 +00:00
|
|
|
{
|
|
|
|
struct page *page;
|
2015-11-07 00:29:57 +00:00
|
|
|
unsigned int order;
|
2007-10-16 08:26:00 +00:00
|
|
|
int pages_moved = 0;
|
2007-10-16 08:25:51 +00:00
|
|
|
|
|
|
|
for (page = start_page; page <= end_page;) {
|
|
|
|
if (!pfn_valid_within(page_to_pfn(page))) {
|
|
|
|
page++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!PageBuddy(page)) {
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
/*
|
|
|
|
* We assume that pages that could be isolated for
|
|
|
|
* migration are movable. But we don't actually try
|
|
|
|
* isolating, as that would be expensive.
|
|
|
|
*/
|
|
|
|
if (num_movable &&
|
|
|
|
(PageLRU(page) || __PageMovable(page)))
|
|
|
|
(*num_movable)++;
|
|
|
|
|
2007-10-16 08:25:51 +00:00
|
|
|
page++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
mm, page_alloc: move_freepages should not examine struct page of reserved memory
After commit 907ec5fca3dc ("mm: zero remaining unavailable struct
pages"), struct page of reserved memory is zeroed. This causes
page->flags to be 0 and fixes issues related to reading
/proc/kpageflags, for example, of reserved memory.
The VM_BUG_ON() in move_freepages_block(), however, assumes that
page_zone() is meaningful even for reserved memory. That assumption is
no longer true after the aforementioned commit.
There's no reason why move_freepages_block() should be testing the
legitimacy of page_zone() for reserved memory; its scope is limited only
to pages on the zone's freelist.
Note that pfn_valid() can be true for reserved memory: there is a
backing struct page. The check for page_to_nid(page) is also buggy but
reserved memory normally only appears on node 0 so the zeroing doesn't
affect this.
Move the debug checks to after verifying PageBuddy is true. This
isolates the scope of the checks to only be for buddy pages which are on
the zone's freelist which move_freepages_block() is operating on. In
this case, an incorrect node or zone is a bug worthy of being warned
about (and the examination of struct page is acceptable bcause this
memory is not reserved).
Why does move_freepages_block() gets called on reserved memory? It's
simply math after finding a valid free page from the per-zone free area
to use as fallback. We find the beginning and end of the pageblock of
the valid page and that can bring us into memory that was reserved per
the e820. pfn_valid() is still true (it's backed by a struct page), but
since it's zero'd we shouldn't make any inferences here about comparing
its node or zone. The current node check just happens to succeed most
of the time by luck because reserved memory typically appears on node 0.
The fix here is to validate that we actually have buddy pages before
testing if there's any type of zone or node strangeness going on.
We noticed it almost immediately after bringing 907ec5fca3dc in on
CONFIG_DEBUG_VM builds. It depends on finding specific free pages in
the per-zone free area where the math in move_freepages() will bring the
start or end pfn into reserved memory and wanting to claim that entire
pageblock as a new migratetype. So the path will be rare, require
CONFIG_DEBUG_VM, and require fallback to a different migratetype.
Some struct pages were already zeroed from reserve pages before
907ec5fca3c so it theoretically could trigger before this commit. I
think it's rare enough under a config option that most people don't run
that others may not have noticed. I wouldn't argue against a stable tag
and the backport should be easy enough, but probably wouldn't single out
a commit that this is fixing.
Mel said:
: The overhead of the debugging check is higher with this patch although
: it'll only affect debug builds and the path is not particularly hot.
: If this was a concern, I think it would be reasonable to simply remove
: the debugging check as the zone boundaries are checked in
: move_freepages_block and we never expect a zone/node to be smaller than
: a pageblock and stuck in the middle of another zone.
Link: http://lkml.kernel.org/r/alpine.DEB.2.21.1908122036560.10779@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-25 00:54:40 +00:00
|
|
|
/* Make sure we are not inadvertently changing nodes */
|
|
|
|
VM_BUG_ON_PAGE(page_to_nid(page) != zone_to_nid(zone), page);
|
|
|
|
VM_BUG_ON_PAGE(page_zone(page) != zone, page);
|
|
|
|
|
2020-10-16 03:10:15 +00:00
|
|
|
order = buddy_order(page);
|
2020-04-07 03:04:49 +00:00
|
|
|
move_to_free_list(page, zone, order, migratetype);
|
2007-10-16 08:25:51 +00:00
|
|
|
page += 1 << order;
|
2007-10-16 08:26:00 +00:00
|
|
|
pages_moved += 1 << order;
|
2007-10-16 08:25:51 +00:00
|
|
|
}
|
|
|
|
|
2007-10-16 08:26:00 +00:00
|
|
|
return pages_moved;
|
2007-10-16 08:25:51 +00:00
|
|
|
}
|
|
|
|
|
2012-07-31 23:43:50 +00:00
|
|
|
int move_freepages_block(struct zone *zone, struct page *page,
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
int migratetype, int *num_movable)
|
2007-10-16 08:25:51 +00:00
|
|
|
{
|
|
|
|
unsigned long start_pfn, end_pfn;
|
|
|
|
struct page *start_page, *end_page;
|
|
|
|
|
2018-10-26 22:09:24 +00:00
|
|
|
if (num_movable)
|
|
|
|
*num_movable = 0;
|
|
|
|
|
2007-10-16 08:25:51 +00:00
|
|
|
start_pfn = page_to_pfn(page);
|
2007-10-16 08:26:01 +00:00
|
|
|
start_pfn = start_pfn & ~(pageblock_nr_pages-1);
|
2007-10-16 08:25:51 +00:00
|
|
|
start_page = pfn_to_page(start_pfn);
|
2007-10-16 08:26:01 +00:00
|
|
|
end_page = start_page + pageblock_nr_pages - 1;
|
|
|
|
end_pfn = start_pfn + pageblock_nr_pages - 1;
|
2007-10-16 08:25:51 +00:00
|
|
|
|
|
|
|
/* Do not cross zone boundaries */
|
2013-02-23 00:35:23 +00:00
|
|
|
if (!zone_spans_pfn(zone, start_pfn))
|
2007-10-16 08:25:51 +00:00
|
|
|
start_page = page;
|
2013-02-23 00:35:23 +00:00
|
|
|
if (!zone_spans_pfn(zone, end_pfn))
|
2007-10-16 08:25:51 +00:00
|
|
|
return 0;
|
|
|
|
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
return move_freepages(zone, start_page, end_page, migratetype,
|
|
|
|
num_movable);
|
2007-10-16 08:25:51 +00:00
|
|
|
}
|
|
|
|
|
2009-09-22 00:02:31 +00:00
|
|
|
static void change_pageblock_range(struct page *pageblock_page,
|
|
|
|
int start_order, int migratetype)
|
|
|
|
{
|
|
|
|
int nr_pageblocks = 1 << (start_order - pageblock_order);
|
|
|
|
|
|
|
|
while (nr_pageblocks--) {
|
|
|
|
set_pageblock_migratetype(pageblock_page, migratetype);
|
|
|
|
pageblock_page += pageblock_nr_pages;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
/*
|
mm: more aggressive page stealing for UNMOVABLE allocations
When allocation falls back to stealing free pages of another migratetype,
it can decide to steal extra pages, or even the whole pageblock in order
to reduce fragmentation, which could happen if further allocation
fallbacks pick a different pageblock. In try_to_steal_freepages(), one of
the situations where extra pages are stolen happens when we are trying to
allocate a MIGRATE_RECLAIMABLE page.
However, MIGRATE_UNMOVABLE allocations are not treated the same way,
although spreading such allocation over multiple fallback pageblocks is
arguably even worse than it is for RECLAIMABLE allocations. To minimize
fragmentation, we should minimize the number of such fallbacks, and thus
steal as much as is possible from each fallback pageblock.
Note that in theory this might put more pressure on movable pageblocks and
cause movable allocations to steal back from unmovable pageblocks.
However, movable allocations are not as aggressive with stealing, and do
not cause permanent fragmentation, so the tradeoff is reasonable, and
evaluation seems to support the change.
This patch thus adds a check for MIGRATE_UNMOVABLE to the decision to
steal extra free pages. When evaluating with stress-highalloc from
mmtests, this has reduced the number of MIGRATE_UNMOVABLE fallbacks to
roughly 1/6. The number of these fallbacks stealing from MIGRATE_MOVABLE
block is reduced to 1/3. There was no observation of growing number of
unmovable pageblocks over time, and also not of increased movable
allocation fallbacks.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:28:21 +00:00
|
|
|
* When we are falling back to another migratetype during allocation, try to
|
|
|
|
* steal extra free pages from the same pageblocks to satisfy further
|
|
|
|
* allocations, instead of polluting multiple pageblocks.
|
|
|
|
*
|
|
|
|
* If we are stealing a relatively large buddy page, it is likely there will
|
|
|
|
* be more free pages in the pageblock, so try to steal them all. For
|
|
|
|
* reclaimable and unmovable allocations, we steal regardless of page size,
|
|
|
|
* as fragmentation caused by those allocations polluting movable pageblocks
|
|
|
|
* is worse than movable allocations stealing from unmovable and reclaimable
|
|
|
|
* pageblocks.
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
*/
|
2015-04-14 22:45:18 +00:00
|
|
|
static bool can_steal_fallback(unsigned int order, int start_mt)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Leaving this order check is intended, although there is
|
|
|
|
* relaxed order check in next check. The reason is that
|
|
|
|
* we can actually steal whole pageblock if this condition met,
|
|
|
|
* but, below check doesn't guarantee it and that is just heuristic
|
|
|
|
* so could be changed anytime.
|
|
|
|
*/
|
|
|
|
if (order >= pageblock_order)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
if (order >= pageblock_order / 2 ||
|
|
|
|
start_mt == MIGRATE_RECLAIMABLE ||
|
|
|
|
start_mt == MIGRATE_UNMOVABLE ||
|
|
|
|
page_group_by_mobility_disabled)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2020-12-15 03:12:15 +00:00
|
|
|
static inline bool boost_watermark(struct zone *zone)
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
{
|
|
|
|
unsigned long max_boost;
|
|
|
|
|
|
|
|
if (!watermark_boost_factor)
|
2020-12-15 03:12:15 +00:00
|
|
|
return false;
|
mm: limit boost_watermark on small zones
Commit 1c30844d2dfe ("mm: reclaim small amounts of memory when an
external fragmentation event occurs") adds a boost_watermark() function
which increases the min watermark in a zone by at least
pageblock_nr_pages or the number of pages in a page block.
On Arm64, with 64K pages and 512M huge pages, this is 8192 pages or
512M. It does this regardless of the number of managed pages managed in
the zone or the likelihood of success.
This can put the zone immediately under water in terms of allocating
pages from the zone, and can cause a small machine to fail immediately
due to OoM. Unlike set_recommended_min_free_kbytes(), which
substantially increases min_free_kbytes and is tied to THP,
boost_watermark() can be called even if THP is not active.
The problem is most likely to appear on architectures such as Arm64
where pageblock_nr_pages is very large.
It is desirable to run the kdump capture kernel in as small a space as
possible to avoid wasting memory. In some architectures, such as Arm64,
there are restrictions on where the capture kernel can run, and
therefore, the space available. A capture kernel running in 768M can
fail due to OoM immediately after boost_watermark() sets the min in zone
DMA32, where most of the memory is, to 512M. It fails even though there
is over 500M of free memory. With boost_watermark() suppressed, the
capture kernel can run successfully in 448M.
This patch limits boost_watermark() to boosting a zone's min watermark
only when there are enough pages that the boost will produce positive
results. In this case that is estimated to be four times as many pages
as pageblock_nr_pages.
Mel said:
: There is no harm in marking it stable. Clearly it does not happen very
: often but it's not impossible. 32-bit x86 is a lot less common now
: which would previously have been vulnerable to triggering this easily.
: ppc64 has a larger base page size but typically only has one zone.
: arm64 is likely the most vulnerable, particularly when CMA is
: configured with a small movable zone.
Fixes: 1c30844d2dfe ("mm: reclaim small amounts of memory when an external fragmentation event occurs")
Signed-off-by: Henry Willard <henry.willard@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/1588294148-6586-1-git-send-email-henry.willard@oracle.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-05-08 01:36:27 +00:00
|
|
|
/*
|
|
|
|
* Don't bother in zones that are unlikely to produce results.
|
|
|
|
* On small machines, including kdump capture kernels running
|
|
|
|
* in a small area, boosting the watermark can cause an out of
|
|
|
|
* memory situation immediately.
|
|
|
|
*/
|
|
|
|
if ((pageblock_nr_pages * 4) > zone_managed_pages(zone))
|
2020-12-15 03:12:15 +00:00
|
|
|
return false;
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
|
|
|
|
max_boost = mult_frac(zone->_watermark[WMARK_HIGH],
|
|
|
|
watermark_boost_factor, 10000);
|
2019-02-21 06:19:49 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* high watermark may be uninitialised if fragmentation occurs
|
|
|
|
* very early in boot so do not boost. We do not fall
|
|
|
|
* through and boost by pageblock_nr_pages as failing
|
|
|
|
* allocations that early means that reclaim is not going
|
|
|
|
* to help and it may even be impossible to reclaim the
|
|
|
|
* boosted watermark resulting in a hang.
|
|
|
|
*/
|
|
|
|
if (!max_boost)
|
2020-12-15 03:12:15 +00:00
|
|
|
return false;
|
2019-02-21 06:19:49 +00:00
|
|
|
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
max_boost = max(pageblock_nr_pages, max_boost);
|
|
|
|
|
|
|
|
zone->watermark_boost = min(zone->watermark_boost + pageblock_nr_pages,
|
|
|
|
max_boost);
|
2020-12-15 03:12:15 +00:00
|
|
|
|
|
|
|
return true;
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
}
|
|
|
|
|
2015-04-14 22:45:18 +00:00
|
|
|
/*
|
|
|
|
* This function implements actual steal behaviour. If order is large enough,
|
|
|
|
* we can steal whole pageblock. If not, we first move freepages in this
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
* pageblock to our migratetype and determine how many already-allocated pages
|
|
|
|
* are there in the pageblock with a compatible migratetype. If at least half
|
|
|
|
* of pages are free or compatible, we can change migratetype of the pageblock
|
|
|
|
* itself, so pages freed in the future will be put on the correct free list.
|
2015-04-14 22:45:18 +00:00
|
|
|
*/
|
|
|
|
static void steal_suitable_fallback(struct zone *zone, struct page *page,
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
unsigned int alloc_flags, int start_type, bool whole_block)
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
{
|
2020-10-16 03:10:15 +00:00
|
|
|
unsigned int current_order = buddy_order(page);
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
int free_pages, movable_pages, alike_pages;
|
|
|
|
int old_block_type;
|
|
|
|
|
|
|
|
old_block_type = get_pageblock_migratetype(page);
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
|
2017-05-08 22:54:37 +00:00
|
|
|
/*
|
|
|
|
* This can happen due to races and we want to prevent broken
|
|
|
|
* highatomic accounting.
|
|
|
|
*/
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
if (is_migrate_highatomic(old_block_type))
|
2017-05-08 22:54:37 +00:00
|
|
|
goto single_page;
|
|
|
|
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
/* Take ownership for orders >= pageblock_order */
|
|
|
|
if (current_order >= pageblock_order) {
|
|
|
|
change_pageblock_range(page, current_order, start_type);
|
2017-05-08 22:54:37 +00:00
|
|
|
goto single_page;
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
}
|
|
|
|
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
/*
|
|
|
|
* Boost watermarks to increase reclaim pressure to reduce the
|
|
|
|
* likelihood of future fallbacks. Wake kswapd now as the node
|
|
|
|
* may be balanced overall and kswapd will not wake naturally.
|
|
|
|
*/
|
2020-12-15 03:12:15 +00:00
|
|
|
if (boost_watermark(zone) && (alloc_flags & ALLOC_KSWAPD))
|
mm, page_alloc: do not wake kswapd with zone lock held
syzbot reported the following regression in the latest merge window and
it was confirmed by Qian Cai that a similar bug was visible from a
different context.
======================================================
WARNING: possible circular locking dependency detected
4.20.0+ #297 Not tainted
------------------------------------------------------
syz-executor0/8529 is trying to acquire lock:
000000005e7fb829 (&pgdat->kswapd_wait){....}, at:
__wake_up_common_lock+0x19e/0x330 kernel/sched/wait.c:120
but task is already holding lock:
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: spin_lock
include/linux/spinlock.h:329 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: rmqueue_bulk
mm/page_alloc.c:2548 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: __rmqueue_pcplist
mm/page_alloc.c:3021 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: rmqueue_pcplist
mm/page_alloc.c:3050 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: rmqueue
mm/page_alloc.c:3072 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at:
get_page_from_freelist+0x1bae/0x52a0 mm/page_alloc.c:3491
It appears to be a false positive in that the only way the lock ordering
should be inverted is if kswapd is waking itself and the wakeup
allocates debugging objects which should already be allocated if it's
kswapd doing the waking. Nevertheless, the possibility exists and so
it's best to avoid the problem.
This patch flags a zone as needing a kswapd using the, surprisingly,
unused zone flag field. The flag is read without the lock held to do
the wakeup. It's possible that the flag setting context is not the same
as the flag clearing context or for small races to occur. However, each
race possibility is harmless and there is no visible degredation in
fragmentation treatment.
While zone->flag could have continued to be unused, there is potential
for moving some existing fields into the flags field instead.
Particularly read-mostly ones like zone->initialized and
zone->contiguous.
Link: http://lkml.kernel.org/r/20190103225712.GJ31517@techsingularity.net
Fixes: 1c30844d2dfe ("mm: reclaim small amounts of memory when an external fragmentation event occurs")
Reported-by: syzbot+93d94a001cfbce9e60e1@syzkaller.appspotmail.com
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Qian Cai <cai@lca.pw>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-08 23:23:39 +00:00
|
|
|
set_bit(ZONE_BOOSTED_WATERMARK, &zone->flags);
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
|
2017-05-08 22:54:37 +00:00
|
|
|
/* We are not allowed to try stealing from the whole block */
|
|
|
|
if (!whole_block)
|
|
|
|
goto single_page;
|
|
|
|
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
free_pages = move_freepages_block(zone, page, start_type,
|
|
|
|
&movable_pages);
|
|
|
|
/*
|
|
|
|
* Determine how many pages are compatible with our allocation.
|
|
|
|
* For movable allocation, it's the number of movable pages which
|
|
|
|
* we just obtained. For other types it's a bit more tricky.
|
|
|
|
*/
|
|
|
|
if (start_type == MIGRATE_MOVABLE) {
|
|
|
|
alike_pages = movable_pages;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* If we are falling back a RECLAIMABLE or UNMOVABLE allocation
|
|
|
|
* to MOVABLE pageblock, consider all non-movable pages as
|
|
|
|
* compatible. If it's UNMOVABLE falling back to RECLAIMABLE or
|
|
|
|
* vice versa, be conservative since we can't distinguish the
|
|
|
|
* exact migratetype of non-movable pages.
|
|
|
|
*/
|
|
|
|
if (old_block_type == MIGRATE_MOVABLE)
|
|
|
|
alike_pages = pageblock_nr_pages
|
|
|
|
- (free_pages + movable_pages);
|
|
|
|
else
|
|
|
|
alike_pages = 0;
|
|
|
|
}
|
|
|
|
|
2017-05-08 22:54:37 +00:00
|
|
|
/* moving whole block can fail due to zone boundary conditions */
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
if (!free_pages)
|
2017-05-08 22:54:37 +00:00
|
|
|
goto single_page;
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
/*
|
|
|
|
* If a sufficient number of pages in the block are either free or of
|
|
|
|
* comparable migratability as our allocation, claim the whole block.
|
|
|
|
*/
|
|
|
|
if (free_pages + alike_pages >= (1 << (pageblock_order-1)) ||
|
2015-04-14 22:45:18 +00:00
|
|
|
page_group_by_mobility_disabled)
|
|
|
|
set_pageblock_migratetype(page, start_type);
|
2017-05-08 22:54:37 +00:00
|
|
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
single_page:
|
2020-04-07 03:04:49 +00:00
|
|
|
move_to_free_list(page, zone, current_order, start_type);
|
2015-04-14 22:45:18 +00:00
|
|
|
}
|
|
|
|
|
mm/compaction: enhance compaction finish condition
Compaction has anti fragmentation algorithm. It is that freepage should
be more than pageblock order to finish the compaction if we don't find any
freepage in requested migratetype buddy list. This is for mitigating
fragmentation, but, there is a lack of migratetype consideration and it is
too excessive compared to page allocator's anti fragmentation algorithm.
Not considering migratetype would cause premature finish of compaction.
For example, if allocation request is for unmovable migratetype, freepage
with CMA migratetype doesn't help that allocation and compaction should
not be stopped. But, current logic regards this situation as compaction
is no longer needed, so finish the compaction.
Secondly, condition is too excessive compared to page allocator's logic.
We can steal freepage from other migratetype and change pageblock
migratetype on more relaxed conditions in page allocator. This is
designed to prevent fragmentation and we can use it here. Imposing hard
constraint only to the compaction doesn't help much in this case since
page allocator would cause fragmentation again.
To solve these problems, this patch borrows anti fragmentation logic from
page allocator. It will reduce premature compaction finish in some cases
and reduce excessive compaction work.
stress-highalloc test in mmtests with non movable order 7 allocation shows
considerable increase of compaction success rate.
Compaction success rate (Compaction success * 100 / Compaction stalls, %)
31.82 : 42.20
I tested it on non-reboot 5 runs stress-highalloc benchmark and found that
there is no more degradation on allocation success rate than before. That
roughly means that this patch doesn't result in more fragmentations.
Vlastimil suggests additional idea that we only test for fallbacks when
migration scanner has scanned a whole pageblock. It looked good for
fragmentation because chance of stealing increase due to making more free
pages in certain pageblock. So, I tested it, but, it results in decreased
compaction success rate, roughly 38.00. I guess the reason that if system
is low memory condition, watermark check could be failed due to not enough
order 0 free page and so, sometimes, we can't reach a fallback check
although migrate_pfn is aligned to pageblock_nr_pages. I can insert code
to cope with this situation but it makes code more complicated so I don't
include his idea at this patch.
[akpm@linux-foundation.org: fix CONFIG_CMA=n build]
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:45:21 +00:00
|
|
|
/*
|
|
|
|
* Check whether there is a suitable fallback freepage with requested order.
|
|
|
|
* If only_stealable is true, this function returns fallback_mt only if
|
|
|
|
* we can steal other freepages all together. This would help to reduce
|
|
|
|
* fragmentation due to mixed migratetype pages in one pageblock.
|
|
|
|
*/
|
|
|
|
int find_suitable_fallback(struct free_area *area, unsigned int order,
|
|
|
|
int migratetype, bool only_stealable, bool *can_steal)
|
2015-04-14 22:45:18 +00:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
int fallback_mt;
|
|
|
|
|
|
|
|
if (area->nr_free == 0)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
*can_steal = false;
|
|
|
|
for (i = 0;; i++) {
|
|
|
|
fallback_mt = fallbacks[migratetype][i];
|
2015-11-07 00:28:34 +00:00
|
|
|
if (fallback_mt == MIGRATE_TYPES)
|
2015-04-14 22:45:18 +00:00
|
|
|
break;
|
|
|
|
|
2019-05-14 22:41:32 +00:00
|
|
|
if (free_area_empty(area, fallback_mt))
|
2015-04-14 22:45:18 +00:00
|
|
|
continue;
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
|
2015-04-14 22:45:18 +00:00
|
|
|
if (can_steal_fallback(order, migratetype))
|
|
|
|
*can_steal = true;
|
|
|
|
|
mm/compaction: enhance compaction finish condition
Compaction has anti fragmentation algorithm. It is that freepage should
be more than pageblock order to finish the compaction if we don't find any
freepage in requested migratetype buddy list. This is for mitigating
fragmentation, but, there is a lack of migratetype consideration and it is
too excessive compared to page allocator's anti fragmentation algorithm.
Not considering migratetype would cause premature finish of compaction.
For example, if allocation request is for unmovable migratetype, freepage
with CMA migratetype doesn't help that allocation and compaction should
not be stopped. But, current logic regards this situation as compaction
is no longer needed, so finish the compaction.
Secondly, condition is too excessive compared to page allocator's logic.
We can steal freepage from other migratetype and change pageblock
migratetype on more relaxed conditions in page allocator. This is
designed to prevent fragmentation and we can use it here. Imposing hard
constraint only to the compaction doesn't help much in this case since
page allocator would cause fragmentation again.
To solve these problems, this patch borrows anti fragmentation logic from
page allocator. It will reduce premature compaction finish in some cases
and reduce excessive compaction work.
stress-highalloc test in mmtests with non movable order 7 allocation shows
considerable increase of compaction success rate.
Compaction success rate (Compaction success * 100 / Compaction stalls, %)
31.82 : 42.20
I tested it on non-reboot 5 runs stress-highalloc benchmark and found that
there is no more degradation on allocation success rate than before. That
roughly means that this patch doesn't result in more fragmentations.
Vlastimil suggests additional idea that we only test for fallbacks when
migration scanner has scanned a whole pageblock. It looked good for
fragmentation because chance of stealing increase due to making more free
pages in certain pageblock. So, I tested it, but, it results in decreased
compaction success rate, roughly 38.00. I guess the reason that if system
is low memory condition, watermark check could be failed due to not enough
order 0 free page and so, sometimes, we can't reach a fallback check
although migrate_pfn is aligned to pageblock_nr_pages. I can insert code
to cope with this situation but it makes code more complicated so I don't
include his idea at this patch.
[akpm@linux-foundation.org: fix CONFIG_CMA=n build]
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:45:21 +00:00
|
|
|
if (!only_stealable)
|
|
|
|
return fallback_mt;
|
|
|
|
|
|
|
|
if (*can_steal)
|
|
|
|
return fallback_mt;
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
}
|
2015-04-14 22:45:18 +00:00
|
|
|
|
|
|
|
return -1;
|
mm/page_allo.c: restructure free-page stealing code and fix a bug
The free-page stealing code in __rmqueue_fallback() is somewhat hard to
follow, and has an incredible amount of subtlety hidden inside!
First off, there is a minor bug in the reporting of change-of-ownership of
pageblocks. Under some conditions, we try to move upto
'pageblock_nr_pages' no. of pages to the preferred allocation list. But
we change the ownership of that pageblock to the preferred type only if we
manage to successfully move atleast half of that pageblock (or if
page_group_by_mobility_disabled is set).
However, the current code ignores the latter part and sets the
'migratetype' variable to the preferred type, irrespective of whether we
actually changed the pageblock migratetype of that block or not. So, the
page_alloc_extfrag tracepoint can end up printing incorrect info (i.e.,
'change_ownership' might be shown as 1 when it must have been 0).
So fixing this involves moving the update of the 'migratetype' variable to
the right place. But looking closer, we observe that the 'migratetype'
variable is used subsequently for checks such as "is_migrate_cma()".
Obviously the intent there is to check if the *fallback* type is
MIGRATE_CMA, but since we already set the 'migratetype' variable to
start_migratetype, we end up checking if the *preferred* type is
MIGRATE_CMA!!
To make things more interesting, this actually doesn't cause a bug in
practice, because we never change *anything* if the fallback type is CMA.
So, restructure the code in such a way that it is trivial to understand
what is going on, and also fix the above mentioned bug. And while at it,
also add a comment explaining the subtlety behind the migratetype used in
the call to expand().
[akpm@linux-foundation.org: remove unneeded `inline', small coding-style fix]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Cody P Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:35 +00:00
|
|
|
}
|
|
|
|
|
2015-11-07 00:28:37 +00:00
|
|
|
/*
|
|
|
|
* Reserve a pageblock for exclusive use of high-order atomic allocations if
|
|
|
|
* there are no empty page blocks that contain a page with a suitable order
|
|
|
|
*/
|
|
|
|
static void reserve_highatomic_pageblock(struct page *page, struct zone *zone,
|
|
|
|
unsigned int alloc_order)
|
|
|
|
{
|
|
|
|
int mt;
|
|
|
|
unsigned long max_managed, flags;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Limit the number reserved to 1 pageblock or roughly 1% of a zone.
|
|
|
|
* Check is race-prone but harmless.
|
|
|
|
*/
|
2018-12-28 08:34:24 +00:00
|
|
|
max_managed = (zone_managed_pages(zone) / 100) + pageblock_nr_pages;
|
2015-11-07 00:28:37 +00:00
|
|
|
if (zone->nr_reserved_highatomic >= max_managed)
|
|
|
|
return;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
|
|
|
|
|
|
|
/* Recheck the nr_reserved_highatomic limit under the lock */
|
|
|
|
if (zone->nr_reserved_highatomic >= max_managed)
|
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
/* Yoink! */
|
|
|
|
mt = get_pageblock_migratetype(page);
|
2017-05-03 21:52:52 +00:00
|
|
|
if (!is_migrate_highatomic(mt) && !is_migrate_isolate(mt)
|
|
|
|
&& !is_migrate_cma(mt)) {
|
2015-11-07 00:28:37 +00:00
|
|
|
zone->nr_reserved_highatomic += pageblock_nr_pages;
|
|
|
|
set_pageblock_migratetype(page, MIGRATE_HIGHATOMIC);
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
move_freepages_block(zone, page, MIGRATE_HIGHATOMIC, NULL);
|
2015-11-07 00:28:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Used when an allocation is about to fail under memory pressure. This
|
|
|
|
* potentially hurts the reliability of high-order allocations when under
|
|
|
|
* intense memory pressure but failed atomic allocations should be easier
|
|
|
|
* to recover from than an OOM.
|
2016-12-13 00:42:14 +00:00
|
|
|
*
|
|
|
|
* If @force is true, try to unreserve a pageblock even though highatomic
|
|
|
|
* pageblock is exhausted.
|
2015-11-07 00:28:37 +00:00
|
|
|
*/
|
2016-12-13 00:42:14 +00:00
|
|
|
static bool unreserve_highatomic_pageblock(const struct alloc_context *ac,
|
|
|
|
bool force)
|
2015-11-07 00:28:37 +00:00
|
|
|
{
|
|
|
|
struct zonelist *zonelist = ac->zonelist;
|
|
|
|
unsigned long flags;
|
|
|
|
struct zoneref *z;
|
|
|
|
struct zone *zone;
|
|
|
|
struct page *page;
|
|
|
|
int order;
|
mm: try to exhaust highatomic reserve before the OOM
I got OOM report from production team with v4.4 kernel. It had enough
free memory but failed to allocate GFP_KERNEL order-0 page and finally
encountered OOM kill. It occured during QA process which launches
several apps, switching and so on. It happned rarely. IOW, In normal
situation, it was not a problem but if we are unluck so that several
apps uses peak memory at the same time, it can happen. If we manage to
pass the phase, the system can go working well.
I could reproduce it with my test(memory spike easily. Look at below.
The reason is free pages(19M) of DMA32 zone are reserved for
HIGHORDERATOMIC and doesn't unreserved before the OOM.
balloon invoked oom-killer: gfp_mask=0x24280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), order=0, oom_score_adj=0
balloon cpuset=/ mems_allowed=0
CPU: 1 PID: 8473 Comm: balloon Tainted: G W OE 4.8.0-rc7-00219-g3f74c9559583-dirty #3161
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
dump_header+0x5c/0x1ce
oom_kill_process+0x22e/0x400
out_of_memory+0x1ac/0x210
__alloc_pages_nodemask+0x101e/0x1040
handle_mm_fault+0xa0a/0xbf0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:383949 inactive_anon:106724 isolated_anon:0
active_file:15 inactive_file:44 isolated_file:0
unevictable:0 dirty:0 writeback:24 unstable:0
slab_reclaimable:2483 slab_unreclaimable:3326
mapped:0 shmem:0 pagetables:1906 bounce:0
free:6898 free_pcp:291 free_cma:0
Node 0 active_anon:1535796kB inactive_anon:426896kB active_file:60kB inactive_file:176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:96kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1418 all_unreclaimable? no
DMA free:8188kB min:44kB low:56kB high:68kB active_anon:7648kB inactive_anon:0kB active_file:0kB inactive_file:4kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:20kB kernel_stack:0kB pagetables:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:19404kB min:5628kB low:7624kB high:9620kB active_anon:1528148kB inactive_anon:426896kB active_file:60kB inactive_file:420kB unevictable:0kB writepending:96kB present:2080640kB managed:2030092kB mlocked:0kB slab_reclaimable:9932kB slab_unreclaimable:13284kB kernel_stack:2496kB pagetables:7624kB bounce:0kB free_pcp:900kB local_pcp:112kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 0*4kB 0*8kB 0*16kB 0*32kB 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 2*4096kB (H) = 8192kB
DMA32: 7*4kB (H) 8*8kB (H) 30*16kB (H) 31*32kB (H) 14*64kB (H) 9*128kB (H) 2*256kB (H) 2*512kB (H) 4*1024kB (H) 5*2048kB (H) 0*4096kB = 19484kB
51131 total pagecache pages
50795 pages in swap cache
Swap cache stats: add 3532405601, delete 3532354806, find 124289150/1822712228
Free swap = 8kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12658 pages reserved
0 pages cma reserved
0 pages hwpoisoned
Another example exceeded the limit by the race is
in:imklog: page allocation failure: order:0, mode:0x2280020(GFP_ATOMIC|__GFP_NOTRACK)
CPU: 0 PID: 476 Comm: in:imklog Tainted: G E 4.8.0-rc7-00217-g266ef83c51e5-dirty #3135
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
warn_alloc_failed+0xdb/0x130
__alloc_pages_nodemask+0x4d6/0xdb0
new_slab+0x339/0x490
___slab_alloc.constprop.74+0x367/0x480
__slab_alloc.constprop.73+0x20/0x40
__kmalloc+0x1a4/0x1e0
alloc_indirect.isra.14+0x1d/0x50
virtqueue_add_sgs+0x1c4/0x470
__virtblk_add_req+0xae/0x1f0
virtio_queue_rq+0x12d/0x290
__blk_mq_run_hw_queue+0x239/0x370
blk_mq_run_hw_queue+0x8f/0xb0
blk_mq_insert_requests+0x18c/0x1a0
blk_mq_flush_plug_list+0x125/0x140
blk_flush_plug_list+0xc7/0x220
blk_finish_plug+0x2c/0x40
__do_page_cache_readahead+0x196/0x230
filemap_fault+0x448/0x4f0
ext4_filemap_fault+0x36/0x50
__do_fault+0x75/0x140
handle_mm_fault+0x84d/0xbe0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:363826 inactive_anon:121283 isolated_anon:32
active_file:65 inactive_file:152 isolated_file:0
unevictable:0 dirty:0 writeback:46 unstable:0
slab_reclaimable:2778 slab_unreclaimable:3070
mapped:112 shmem:0 pagetables:1822 bounce:0
free:9469 free_pcp:231 free_cma:0
Node 0 active_anon:1455304kB inactive_anon:485132kB active_file:260kB inactive_file:608kB unevictable:0kB isolated(anon):128kB isolated(file):0kB mapped:448kB dirty:0kB writeback:184kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:13641 all_unreclaimable? no
DMA free:7748kB min:44kB low:56kB high:68kB active_anon:7944kB inactive_anon:104kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:108kB kernel_stack:0kB pagetables:4kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:30128kB min:5628kB low:7624kB high:9620kB active_anon:1447360kB inactive_anon:485028kB active_file:260kB inactive_file:608kB unevictable:0kB writepending:184kB present:2080640kB managed:2030132kB mlocked:0kB slab_reclaimable:11112kB slab_unreclaimable:12172kB kernel_stack:2400kB pagetables:7284kB bounce:0kB free_pcp:924kB local_pcp:72kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 7*4kB (UE) 3*8kB (UH) 1*16kB (M) 0*32kB 2*64kB (U) 1*128kB (M) 1*256kB (U) 0*512kB 1*1024kB (U) 1*2048kB (U) 1*4096kB (H) = 7748kB
DMA32: 10*4kB (H) 3*8kB (H) 47*16kB (H) 38*32kB (H) 5*64kB (H) 1*128kB (H) 2*256kB (H) 3*512kB (H) 3*1024kB (H) 3*2048kB (H) 4*4096kB (H) = 30128kB
2775 total pagecache pages
2536 pages in swap cache
Swap cache stats: add 206786828, delete 206784292, find 7323106/106686077
Free swap = 108744kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12648 pages reserved
0 pages cma reserved
0 pages hwpoisoned
It's weird to show that zone has enough free memory above min watermark
but OOMed with 4K GFP_KERNEL allocation due to reserved highatomic
pages. As last resort, try to unreserve highatomic pages again and if
it has moved pages to non-highatmoc free list, retry reclaim once more.
Link: http://lkml.kernel.org/r/1476259429-18279-4-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Sangseok Lee <sangseok.lee@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-12-13 00:42:11 +00:00
|
|
|
bool ret;
|
2015-11-07 00:28:37 +00:00
|
|
|
|
2020-06-03 22:59:01 +00:00
|
|
|
for_each_zone_zonelist_nodemask(zone, z, zonelist, ac->highest_zoneidx,
|
2015-11-07 00:28:37 +00:00
|
|
|
ac->nodemask) {
|
2016-12-13 00:42:14 +00:00
|
|
|
/*
|
|
|
|
* Preserve at least one pageblock unless memory pressure
|
|
|
|
* is really high.
|
|
|
|
*/
|
|
|
|
if (!force && zone->nr_reserved_highatomic <=
|
|
|
|
pageblock_nr_pages)
|
2015-11-07 00:28:37 +00:00
|
|
|
continue;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
|
|
|
for (order = 0; order < MAX_ORDER; order++) {
|
|
|
|
struct free_area *area = &(zone->free_area[order]);
|
|
|
|
|
2019-05-14 22:41:32 +00:00
|
|
|
page = get_page_from_free_area(area, MIGRATE_HIGHATOMIC);
|
2016-01-14 23:20:30 +00:00
|
|
|
if (!page)
|
2015-11-07 00:28:37 +00:00
|
|
|
continue;
|
|
|
|
|
|
|
|
/*
|
2016-12-13 00:42:08 +00:00
|
|
|
* In page freeing path, migratetype change is racy so
|
|
|
|
* we can counter several free pages in a pageblock
|
|
|
|
* in this loop althoug we changed the pageblock type
|
|
|
|
* from highatomic to ac->migratetype. So we should
|
|
|
|
* adjust the count once.
|
2015-11-07 00:28:37 +00:00
|
|
|
*/
|
2017-05-03 21:52:52 +00:00
|
|
|
if (is_migrate_highatomic_page(page)) {
|
2016-12-13 00:42:08 +00:00
|
|
|
/*
|
|
|
|
* It should never happen but changes to
|
|
|
|
* locking could inadvertently allow a per-cpu
|
|
|
|
* drain to add pages to MIGRATE_HIGHATOMIC
|
|
|
|
* while unreserving so be safe and watch for
|
|
|
|
* underflows.
|
|
|
|
*/
|
|
|
|
zone->nr_reserved_highatomic -= min(
|
|
|
|
pageblock_nr_pages,
|
|
|
|
zone->nr_reserved_highatomic);
|
|
|
|
}
|
2015-11-07 00:28:37 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Convert to ac->migratetype and avoid the normal
|
|
|
|
* pageblock stealing heuristics. Minimally, the caller
|
|
|
|
* is doing the work and needs the pages. More
|
|
|
|
* importantly, if the block was always converted to
|
|
|
|
* MIGRATE_UNMOVABLE or another type then the number
|
|
|
|
* of pageblocks that cannot be completely freed
|
|
|
|
* may increase.
|
|
|
|
*/
|
|
|
|
set_pageblock_migratetype(page, ac->migratetype);
|
mm, page_alloc: count movable pages when stealing from pageblock
When stealing pages from pageblock of a different migratetype, we count
how many free pages were stolen, and change the pageblock's migratetype
if more than half of the pageblock was free. This might be too
conservative, as there might be other pages that are not free, but were
allocated with the same migratetype as our allocation requested.
While we cannot determine the migratetype of allocated pages precisely
(at least without the page_owner functionality enabled), we can count
pages that compaction would try to isolate for migration - those are
either on LRU or __PageMovable(). The rest can be assumed to be
MIGRATE_RECLAIMABLE or MIGRATE_UNMOVABLE, which we cannot easily
distinguish. This counting can be done as part of free page stealing
with little additional overhead.
The page stealing code is changed so that it considers free pages plus
pages of the "good" migratetype for the decision whether to change
pageblock's migratetype.
The result should be more accurate migratetype of pageblocks wrt the
actual pages in the pageblocks, when stealing from semi-occupied
pageblocks. This should help the efficiency of page grouping by
mobility.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 47%. The number of movable allocations falling back to other
pageblocks are increased by 55%, but these events don't cause permanent
fragmentation, so the tradeoff should be positive. Later patches also
offset the movable fallback increase to some extent.
[akpm@linux-foundation.org: merge fix]
Link: http://lkml.kernel.org/r/20170307131545.28577-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:40 +00:00
|
|
|
ret = move_freepages_block(zone, page, ac->migratetype,
|
|
|
|
NULL);
|
2016-12-13 00:42:14 +00:00
|
|
|
if (ret) {
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
|
|
|
return ret;
|
|
|
|
}
|
2015-11-07 00:28:37 +00:00
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
|
|
|
}
|
mm: try to exhaust highatomic reserve before the OOM
I got OOM report from production team with v4.4 kernel. It had enough
free memory but failed to allocate GFP_KERNEL order-0 page and finally
encountered OOM kill. It occured during QA process which launches
several apps, switching and so on. It happned rarely. IOW, In normal
situation, it was not a problem but if we are unluck so that several
apps uses peak memory at the same time, it can happen. If we manage to
pass the phase, the system can go working well.
I could reproduce it with my test(memory spike easily. Look at below.
The reason is free pages(19M) of DMA32 zone are reserved for
HIGHORDERATOMIC and doesn't unreserved before the OOM.
balloon invoked oom-killer: gfp_mask=0x24280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), order=0, oom_score_adj=0
balloon cpuset=/ mems_allowed=0
CPU: 1 PID: 8473 Comm: balloon Tainted: G W OE 4.8.0-rc7-00219-g3f74c9559583-dirty #3161
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
dump_header+0x5c/0x1ce
oom_kill_process+0x22e/0x400
out_of_memory+0x1ac/0x210
__alloc_pages_nodemask+0x101e/0x1040
handle_mm_fault+0xa0a/0xbf0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:383949 inactive_anon:106724 isolated_anon:0
active_file:15 inactive_file:44 isolated_file:0
unevictable:0 dirty:0 writeback:24 unstable:0
slab_reclaimable:2483 slab_unreclaimable:3326
mapped:0 shmem:0 pagetables:1906 bounce:0
free:6898 free_pcp:291 free_cma:0
Node 0 active_anon:1535796kB inactive_anon:426896kB active_file:60kB inactive_file:176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:96kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1418 all_unreclaimable? no
DMA free:8188kB min:44kB low:56kB high:68kB active_anon:7648kB inactive_anon:0kB active_file:0kB inactive_file:4kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:20kB kernel_stack:0kB pagetables:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:19404kB min:5628kB low:7624kB high:9620kB active_anon:1528148kB inactive_anon:426896kB active_file:60kB inactive_file:420kB unevictable:0kB writepending:96kB present:2080640kB managed:2030092kB mlocked:0kB slab_reclaimable:9932kB slab_unreclaimable:13284kB kernel_stack:2496kB pagetables:7624kB bounce:0kB free_pcp:900kB local_pcp:112kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 0*4kB 0*8kB 0*16kB 0*32kB 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 2*4096kB (H) = 8192kB
DMA32: 7*4kB (H) 8*8kB (H) 30*16kB (H) 31*32kB (H) 14*64kB (H) 9*128kB (H) 2*256kB (H) 2*512kB (H) 4*1024kB (H) 5*2048kB (H) 0*4096kB = 19484kB
51131 total pagecache pages
50795 pages in swap cache
Swap cache stats: add 3532405601, delete 3532354806, find 124289150/1822712228
Free swap = 8kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12658 pages reserved
0 pages cma reserved
0 pages hwpoisoned
Another example exceeded the limit by the race is
in:imklog: page allocation failure: order:0, mode:0x2280020(GFP_ATOMIC|__GFP_NOTRACK)
CPU: 0 PID: 476 Comm: in:imklog Tainted: G E 4.8.0-rc7-00217-g266ef83c51e5-dirty #3135
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
warn_alloc_failed+0xdb/0x130
__alloc_pages_nodemask+0x4d6/0xdb0
new_slab+0x339/0x490
___slab_alloc.constprop.74+0x367/0x480
__slab_alloc.constprop.73+0x20/0x40
__kmalloc+0x1a4/0x1e0
alloc_indirect.isra.14+0x1d/0x50
virtqueue_add_sgs+0x1c4/0x470
__virtblk_add_req+0xae/0x1f0
virtio_queue_rq+0x12d/0x290
__blk_mq_run_hw_queue+0x239/0x370
blk_mq_run_hw_queue+0x8f/0xb0
blk_mq_insert_requests+0x18c/0x1a0
blk_mq_flush_plug_list+0x125/0x140
blk_flush_plug_list+0xc7/0x220
blk_finish_plug+0x2c/0x40
__do_page_cache_readahead+0x196/0x230
filemap_fault+0x448/0x4f0
ext4_filemap_fault+0x36/0x50
__do_fault+0x75/0x140
handle_mm_fault+0x84d/0xbe0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:363826 inactive_anon:121283 isolated_anon:32
active_file:65 inactive_file:152 isolated_file:0
unevictable:0 dirty:0 writeback:46 unstable:0
slab_reclaimable:2778 slab_unreclaimable:3070
mapped:112 shmem:0 pagetables:1822 bounce:0
free:9469 free_pcp:231 free_cma:0
Node 0 active_anon:1455304kB inactive_anon:485132kB active_file:260kB inactive_file:608kB unevictable:0kB isolated(anon):128kB isolated(file):0kB mapped:448kB dirty:0kB writeback:184kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:13641 all_unreclaimable? no
DMA free:7748kB min:44kB low:56kB high:68kB active_anon:7944kB inactive_anon:104kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:108kB kernel_stack:0kB pagetables:4kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:30128kB min:5628kB low:7624kB high:9620kB active_anon:1447360kB inactive_anon:485028kB active_file:260kB inactive_file:608kB unevictable:0kB writepending:184kB present:2080640kB managed:2030132kB mlocked:0kB slab_reclaimable:11112kB slab_unreclaimable:12172kB kernel_stack:2400kB pagetables:7284kB bounce:0kB free_pcp:924kB local_pcp:72kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 7*4kB (UE) 3*8kB (UH) 1*16kB (M) 0*32kB 2*64kB (U) 1*128kB (M) 1*256kB (U) 0*512kB 1*1024kB (U) 1*2048kB (U) 1*4096kB (H) = 7748kB
DMA32: 10*4kB (H) 3*8kB (H) 47*16kB (H) 38*32kB (H) 5*64kB (H) 1*128kB (H) 2*256kB (H) 3*512kB (H) 3*1024kB (H) 3*2048kB (H) 4*4096kB (H) = 30128kB
2775 total pagecache pages
2536 pages in swap cache
Swap cache stats: add 206786828, delete 206784292, find 7323106/106686077
Free swap = 108744kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12648 pages reserved
0 pages cma reserved
0 pages hwpoisoned
It's weird to show that zone has enough free memory above min watermark
but OOMed with 4K GFP_KERNEL allocation due to reserved highatomic
pages. As last resort, try to unreserve highatomic pages again and if
it has moved pages to non-highatmoc free list, retry reclaim once more.
Link: http://lkml.kernel.org/r/1476259429-18279-4-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Sangseok Lee <sangseok.lee@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-12-13 00:42:11 +00:00
|
|
|
|
|
|
|
return false;
|
2015-11-07 00:28:37 +00:00
|
|
|
}
|
|
|
|
|
2017-05-08 22:54:37 +00:00
|
|
|
/*
|
|
|
|
* Try finding a free buddy page on the fallback list and put it on the free
|
|
|
|
* list of requested migratetype, possibly along with other pages from the same
|
|
|
|
* block, depending on fragmentation avoidance heuristics. Returns true if
|
|
|
|
* fallback was found so that __rmqueue_smallest() can grab it.
|
2017-07-10 22:49:26 +00:00
|
|
|
*
|
|
|
|
* The use of signed ints for order and current_order is a deliberate
|
|
|
|
* deviation from the rest of this file, to make the for loop
|
|
|
|
* condition simpler.
|
2017-05-08 22:54:37 +00:00
|
|
|
*/
|
2017-11-16 01:36:53 +00:00
|
|
|
static __always_inline bool
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
__rmqueue_fallback(struct zone *zone, int order, int start_migratetype,
|
|
|
|
unsigned int alloc_flags)
|
2007-10-16 08:25:48 +00:00
|
|
|
{
|
2013-09-11 21:20:34 +00:00
|
|
|
struct free_area *area;
|
2017-07-10 22:49:26 +00:00
|
|
|
int current_order;
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
int min_order = order;
|
2007-10-16 08:25:48 +00:00
|
|
|
struct page *page;
|
2015-04-14 22:45:18 +00:00
|
|
|
int fallback_mt;
|
|
|
|
bool can_steal;
|
2007-10-16 08:25:48 +00:00
|
|
|
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
/*
|
|
|
|
* Do not steal pages from freelists belonging to other pageblocks
|
|
|
|
* i.e. orders < pageblock_order. If there are no local zones free,
|
|
|
|
* the zonelists will be reiterated without ALLOC_NOFRAGMENT.
|
|
|
|
*/
|
|
|
|
if (alloc_flags & ALLOC_NOFRAGMENT)
|
|
|
|
min_order = pageblock_order;
|
|
|
|
|
mm, page_alloc: fallback to smallest page when not stealing whole pageblock
Since commit 3bc48f96cf11 ("mm, page_alloc: split smallest stolen page
in fallback") we pick the smallest (but sufficient) page of all that
have been stolen from a pageblock of different migratetype. However,
there are cases when we decide not to steal the whole pageblock.
Practically in the current implementation it means that we are trying to
fallback for a MIGRATE_MOVABLE allocation of order X, go through the
freelists from MAX_ORDER-1 down to X, and find free page of order Y. If
Y is less than pageblock_order / 2, we decide not to steal all pages
from the pageblock. When Y > X, it means we are potentially splitting a
larger page than we need, as there might be other pages of order Z,
where X <= Z < Y. Since Y is already too small to steal whole
pageblock, picking smallest available Z will result in the same decision
and we avoid splitting a higher-order page in a MIGRATE_UNMOVABLE or
MIGRATE_RECLAIMABLE pageblock.
This patch therefore changes the fallback algorithm so that in the
situation described above, we switch the fallback search strategy to go
from order X upwards to find the smallest suitable fallback. In theory
there shouldn't be a downside of this change wrt fragmentation.
This has been tested with mmtests' stress-highalloc performing
GFP_KERNEL order-4 allocations, here is the relevant extfrag tracepoint
statistics:
4.12.0-rc2 4.12.0-rc2
1-kernel4 2-kernel4
Page alloc extfrag event 25640976 69680977
Extfrag fragmenting 25621086 69661364
Extfrag fragmenting for unmovable 74409 73204
Extfrag fragmenting unmovable placed with movable 69003 67684
Extfrag fragmenting unmovable placed with reclaim. 5406 5520
Extfrag fragmenting for reclaimable 6398 8467
Extfrag fragmenting reclaimable placed with movable 869 884
Extfrag fragmenting reclaimable placed with unmov. 5529 7583
Extfrag fragmenting for movable 25540279 69579693
Since we force movable allocations to steal the smallest available page
(which we then practially always split), we steal less per fallback, so
the number of fallbacks increases and steals potentially happen from
different pageblocks. This is however not an issue for movable pages
that can be compacted.
Importantly, the "unmovable placed with movable" statistics is lower,
which is the result of less fragmentation in the unmovable pageblocks.
The effect on reclaimable allocation is a bit unclear.
Link: http://lkml.kernel.org/r/20170529093947.22618-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-10 22:47:14 +00:00
|
|
|
/*
|
|
|
|
* Find the largest available free page in the other list. This roughly
|
|
|
|
* approximates finding the pageblock with the most free pages, which
|
|
|
|
* would be too costly to do exactly.
|
|
|
|
*/
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
for (current_order = MAX_ORDER - 1; current_order >= min_order;
|
2014-06-04 23:10:21 +00:00
|
|
|
--current_order) {
|
2015-04-14 22:45:18 +00:00
|
|
|
area = &(zone->free_area[current_order]);
|
|
|
|
fallback_mt = find_suitable_fallback(area, current_order,
|
mm/compaction: enhance compaction finish condition
Compaction has anti fragmentation algorithm. It is that freepage should
be more than pageblock order to finish the compaction if we don't find any
freepage in requested migratetype buddy list. This is for mitigating
fragmentation, but, there is a lack of migratetype consideration and it is
too excessive compared to page allocator's anti fragmentation algorithm.
Not considering migratetype would cause premature finish of compaction.
For example, if allocation request is for unmovable migratetype, freepage
with CMA migratetype doesn't help that allocation and compaction should
not be stopped. But, current logic regards this situation as compaction
is no longer needed, so finish the compaction.
Secondly, condition is too excessive compared to page allocator's logic.
We can steal freepage from other migratetype and change pageblock
migratetype on more relaxed conditions in page allocator. This is
designed to prevent fragmentation and we can use it here. Imposing hard
constraint only to the compaction doesn't help much in this case since
page allocator would cause fragmentation again.
To solve these problems, this patch borrows anti fragmentation logic from
page allocator. It will reduce premature compaction finish in some cases
and reduce excessive compaction work.
stress-highalloc test in mmtests with non movable order 7 allocation shows
considerable increase of compaction success rate.
Compaction success rate (Compaction success * 100 / Compaction stalls, %)
31.82 : 42.20
I tested it on non-reboot 5 runs stress-highalloc benchmark and found that
there is no more degradation on allocation success rate than before. That
roughly means that this patch doesn't result in more fragmentations.
Vlastimil suggests additional idea that we only test for fallbacks when
migration scanner has scanned a whole pageblock. It looked good for
fragmentation because chance of stealing increase due to making more free
pages in certain pageblock. So, I tested it, but, it results in decreased
compaction success rate, roughly 38.00. I guess the reason that if system
is low memory condition, watermark check could be failed due to not enough
order 0 free page and so, sometimes, we can't reach a fallback check
although migrate_pfn is aligned to pageblock_nr_pages. I can insert code
to cope with this situation but it makes code more complicated so I don't
include his idea at this patch.
[akpm@linux-foundation.org: fix CONFIG_CMA=n build]
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:45:21 +00:00
|
|
|
start_migratetype, false, &can_steal);
|
2015-04-14 22:45:18 +00:00
|
|
|
if (fallback_mt == -1)
|
|
|
|
continue;
|
2007-10-16 08:25:48 +00:00
|
|
|
|
mm, page_alloc: fallback to smallest page when not stealing whole pageblock
Since commit 3bc48f96cf11 ("mm, page_alloc: split smallest stolen page
in fallback") we pick the smallest (but sufficient) page of all that
have been stolen from a pageblock of different migratetype. However,
there are cases when we decide not to steal the whole pageblock.
Practically in the current implementation it means that we are trying to
fallback for a MIGRATE_MOVABLE allocation of order X, go through the
freelists from MAX_ORDER-1 down to X, and find free page of order Y. If
Y is less than pageblock_order / 2, we decide not to steal all pages
from the pageblock. When Y > X, it means we are potentially splitting a
larger page than we need, as there might be other pages of order Z,
where X <= Z < Y. Since Y is already too small to steal whole
pageblock, picking smallest available Z will result in the same decision
and we avoid splitting a higher-order page in a MIGRATE_UNMOVABLE or
MIGRATE_RECLAIMABLE pageblock.
This patch therefore changes the fallback algorithm so that in the
situation described above, we switch the fallback search strategy to go
from order X upwards to find the smallest suitable fallback. In theory
there shouldn't be a downside of this change wrt fragmentation.
This has been tested with mmtests' stress-highalloc performing
GFP_KERNEL order-4 allocations, here is the relevant extfrag tracepoint
statistics:
4.12.0-rc2 4.12.0-rc2
1-kernel4 2-kernel4
Page alloc extfrag event 25640976 69680977
Extfrag fragmenting 25621086 69661364
Extfrag fragmenting for unmovable 74409 73204
Extfrag fragmenting unmovable placed with movable 69003 67684
Extfrag fragmenting unmovable placed with reclaim. 5406 5520
Extfrag fragmenting for reclaimable 6398 8467
Extfrag fragmenting reclaimable placed with movable 869 884
Extfrag fragmenting reclaimable placed with unmov. 5529 7583
Extfrag fragmenting for movable 25540279 69579693
Since we force movable allocations to steal the smallest available page
(which we then practially always split), we steal less per fallback, so
the number of fallbacks increases and steals potentially happen from
different pageblocks. This is however not an issue for movable pages
that can be compacted.
Importantly, the "unmovable placed with movable" statistics is lower,
which is the result of less fragmentation in the unmovable pageblocks.
The effect on reclaimable allocation is a bit unclear.
Link: http://lkml.kernel.org/r/20170529093947.22618-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-10 22:47:14 +00:00
|
|
|
/*
|
|
|
|
* We cannot steal all free pages from the pageblock and the
|
|
|
|
* requested migratetype is movable. In that case it's better to
|
|
|
|
* steal and split the smallest available page instead of the
|
|
|
|
* largest available page, because even if the next movable
|
|
|
|
* allocation falls back into a different pageblock than this
|
|
|
|
* one, it won't cause permanent fragmentation.
|
|
|
|
*/
|
|
|
|
if (!can_steal && start_migratetype == MIGRATE_MOVABLE
|
|
|
|
&& current_order > order)
|
|
|
|
goto find_smallest;
|
2007-10-16 08:25:48 +00:00
|
|
|
|
mm, page_alloc: fallback to smallest page when not stealing whole pageblock
Since commit 3bc48f96cf11 ("mm, page_alloc: split smallest stolen page
in fallback") we pick the smallest (but sufficient) page of all that
have been stolen from a pageblock of different migratetype. However,
there are cases when we decide not to steal the whole pageblock.
Practically in the current implementation it means that we are trying to
fallback for a MIGRATE_MOVABLE allocation of order X, go through the
freelists from MAX_ORDER-1 down to X, and find free page of order Y. If
Y is less than pageblock_order / 2, we decide not to steal all pages
from the pageblock. When Y > X, it means we are potentially splitting a
larger page than we need, as there might be other pages of order Z,
where X <= Z < Y. Since Y is already too small to steal whole
pageblock, picking smallest available Z will result in the same decision
and we avoid splitting a higher-order page in a MIGRATE_UNMOVABLE or
MIGRATE_RECLAIMABLE pageblock.
This patch therefore changes the fallback algorithm so that in the
situation described above, we switch the fallback search strategy to go
from order X upwards to find the smallest suitable fallback. In theory
there shouldn't be a downside of this change wrt fragmentation.
This has been tested with mmtests' stress-highalloc performing
GFP_KERNEL order-4 allocations, here is the relevant extfrag tracepoint
statistics:
4.12.0-rc2 4.12.0-rc2
1-kernel4 2-kernel4
Page alloc extfrag event 25640976 69680977
Extfrag fragmenting 25621086 69661364
Extfrag fragmenting for unmovable 74409 73204
Extfrag fragmenting unmovable placed with movable 69003 67684
Extfrag fragmenting unmovable placed with reclaim. 5406 5520
Extfrag fragmenting for reclaimable 6398 8467
Extfrag fragmenting reclaimable placed with movable 869 884
Extfrag fragmenting reclaimable placed with unmov. 5529 7583
Extfrag fragmenting for movable 25540279 69579693
Since we force movable allocations to steal the smallest available page
(which we then practially always split), we steal less per fallback, so
the number of fallbacks increases and steals potentially happen from
different pageblocks. This is however not an issue for movable pages
that can be compacted.
Importantly, the "unmovable placed with movable" statistics is lower,
which is the result of less fragmentation in the unmovable pageblocks.
The effect on reclaimable allocation is a bit unclear.
Link: http://lkml.kernel.org/r/20170529093947.22618-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-10 22:47:14 +00:00
|
|
|
goto do_steal;
|
|
|
|
}
|
2009-09-22 00:02:42 +00:00
|
|
|
|
mm, page_alloc: fallback to smallest page when not stealing whole pageblock
Since commit 3bc48f96cf11 ("mm, page_alloc: split smallest stolen page
in fallback") we pick the smallest (but sufficient) page of all that
have been stolen from a pageblock of different migratetype. However,
there are cases when we decide not to steal the whole pageblock.
Practically in the current implementation it means that we are trying to
fallback for a MIGRATE_MOVABLE allocation of order X, go through the
freelists from MAX_ORDER-1 down to X, and find free page of order Y. If
Y is less than pageblock_order / 2, we decide not to steal all pages
from the pageblock. When Y > X, it means we are potentially splitting a
larger page than we need, as there might be other pages of order Z,
where X <= Z < Y. Since Y is already too small to steal whole
pageblock, picking smallest available Z will result in the same decision
and we avoid splitting a higher-order page in a MIGRATE_UNMOVABLE or
MIGRATE_RECLAIMABLE pageblock.
This patch therefore changes the fallback algorithm so that in the
situation described above, we switch the fallback search strategy to go
from order X upwards to find the smallest suitable fallback. In theory
there shouldn't be a downside of this change wrt fragmentation.
This has been tested with mmtests' stress-highalloc performing
GFP_KERNEL order-4 allocations, here is the relevant extfrag tracepoint
statistics:
4.12.0-rc2 4.12.0-rc2
1-kernel4 2-kernel4
Page alloc extfrag event 25640976 69680977
Extfrag fragmenting 25621086 69661364
Extfrag fragmenting for unmovable 74409 73204
Extfrag fragmenting unmovable placed with movable 69003 67684
Extfrag fragmenting unmovable placed with reclaim. 5406 5520
Extfrag fragmenting for reclaimable 6398 8467
Extfrag fragmenting reclaimable placed with movable 869 884
Extfrag fragmenting reclaimable placed with unmov. 5529 7583
Extfrag fragmenting for movable 25540279 69579693
Since we force movable allocations to steal the smallest available page
(which we then practially always split), we steal less per fallback, so
the number of fallbacks increases and steals potentially happen from
different pageblocks. This is however not an issue for movable pages
that can be compacted.
Importantly, the "unmovable placed with movable" statistics is lower,
which is the result of less fragmentation in the unmovable pageblocks.
The effect on reclaimable allocation is a bit unclear.
Link: http://lkml.kernel.org/r/20170529093947.22618-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-10 22:47:14 +00:00
|
|
|
return false;
|
2009-09-22 00:02:42 +00:00
|
|
|
|
mm, page_alloc: fallback to smallest page when not stealing whole pageblock
Since commit 3bc48f96cf11 ("mm, page_alloc: split smallest stolen page
in fallback") we pick the smallest (but sufficient) page of all that
have been stolen from a pageblock of different migratetype. However,
there are cases when we decide not to steal the whole pageblock.
Practically in the current implementation it means that we are trying to
fallback for a MIGRATE_MOVABLE allocation of order X, go through the
freelists from MAX_ORDER-1 down to X, and find free page of order Y. If
Y is less than pageblock_order / 2, we decide not to steal all pages
from the pageblock. When Y > X, it means we are potentially splitting a
larger page than we need, as there might be other pages of order Z,
where X <= Z < Y. Since Y is already too small to steal whole
pageblock, picking smallest available Z will result in the same decision
and we avoid splitting a higher-order page in a MIGRATE_UNMOVABLE or
MIGRATE_RECLAIMABLE pageblock.
This patch therefore changes the fallback algorithm so that in the
situation described above, we switch the fallback search strategy to go
from order X upwards to find the smallest suitable fallback. In theory
there shouldn't be a downside of this change wrt fragmentation.
This has been tested with mmtests' stress-highalloc performing
GFP_KERNEL order-4 allocations, here is the relevant extfrag tracepoint
statistics:
4.12.0-rc2 4.12.0-rc2
1-kernel4 2-kernel4
Page alloc extfrag event 25640976 69680977
Extfrag fragmenting 25621086 69661364
Extfrag fragmenting for unmovable 74409 73204
Extfrag fragmenting unmovable placed with movable 69003 67684
Extfrag fragmenting unmovable placed with reclaim. 5406 5520
Extfrag fragmenting for reclaimable 6398 8467
Extfrag fragmenting reclaimable placed with movable 869 884
Extfrag fragmenting reclaimable placed with unmov. 5529 7583
Extfrag fragmenting for movable 25540279 69579693
Since we force movable allocations to steal the smallest available page
(which we then practially always split), we steal less per fallback, so
the number of fallbacks increases and steals potentially happen from
different pageblocks. This is however not an issue for movable pages
that can be compacted.
Importantly, the "unmovable placed with movable" statistics is lower,
which is the result of less fragmentation in the unmovable pageblocks.
The effect on reclaimable allocation is a bit unclear.
Link: http://lkml.kernel.org/r/20170529093947.22618-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-10 22:47:14 +00:00
|
|
|
find_smallest:
|
|
|
|
for (current_order = order; current_order < MAX_ORDER;
|
|
|
|
current_order++) {
|
|
|
|
area = &(zone->free_area[current_order]);
|
|
|
|
fallback_mt = find_suitable_fallback(area, current_order,
|
|
|
|
start_migratetype, false, &can_steal);
|
|
|
|
if (fallback_mt != -1)
|
|
|
|
break;
|
2007-10-16 08:25:48 +00:00
|
|
|
}
|
|
|
|
|
mm, page_alloc: fallback to smallest page when not stealing whole pageblock
Since commit 3bc48f96cf11 ("mm, page_alloc: split smallest stolen page
in fallback") we pick the smallest (but sufficient) page of all that
have been stolen from a pageblock of different migratetype. However,
there are cases when we decide not to steal the whole pageblock.
Practically in the current implementation it means that we are trying to
fallback for a MIGRATE_MOVABLE allocation of order X, go through the
freelists from MAX_ORDER-1 down to X, and find free page of order Y. If
Y is less than pageblock_order / 2, we decide not to steal all pages
from the pageblock. When Y > X, it means we are potentially splitting a
larger page than we need, as there might be other pages of order Z,
where X <= Z < Y. Since Y is already too small to steal whole
pageblock, picking smallest available Z will result in the same decision
and we avoid splitting a higher-order page in a MIGRATE_UNMOVABLE or
MIGRATE_RECLAIMABLE pageblock.
This patch therefore changes the fallback algorithm so that in the
situation described above, we switch the fallback search strategy to go
from order X upwards to find the smallest suitable fallback. In theory
there shouldn't be a downside of this change wrt fragmentation.
This has been tested with mmtests' stress-highalloc performing
GFP_KERNEL order-4 allocations, here is the relevant extfrag tracepoint
statistics:
4.12.0-rc2 4.12.0-rc2
1-kernel4 2-kernel4
Page alloc extfrag event 25640976 69680977
Extfrag fragmenting 25621086 69661364
Extfrag fragmenting for unmovable 74409 73204
Extfrag fragmenting unmovable placed with movable 69003 67684
Extfrag fragmenting unmovable placed with reclaim. 5406 5520
Extfrag fragmenting for reclaimable 6398 8467
Extfrag fragmenting reclaimable placed with movable 869 884
Extfrag fragmenting reclaimable placed with unmov. 5529 7583
Extfrag fragmenting for movable 25540279 69579693
Since we force movable allocations to steal the smallest available page
(which we then practially always split), we steal less per fallback, so
the number of fallbacks increases and steals potentially happen from
different pageblocks. This is however not an issue for movable pages
that can be compacted.
Importantly, the "unmovable placed with movable" statistics is lower,
which is the result of less fragmentation in the unmovable pageblocks.
The effect on reclaimable allocation is a bit unclear.
Link: http://lkml.kernel.org/r/20170529093947.22618-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-10 22:47:14 +00:00
|
|
|
/*
|
|
|
|
* This should not happen - we already found a suitable fallback
|
|
|
|
* when looking for the largest page.
|
|
|
|
*/
|
|
|
|
VM_BUG_ON(current_order == MAX_ORDER);
|
|
|
|
|
|
|
|
do_steal:
|
2019-05-14 22:41:32 +00:00
|
|
|
page = get_page_from_free_area(area, fallback_mt);
|
mm, page_alloc: fallback to smallest page when not stealing whole pageblock
Since commit 3bc48f96cf11 ("mm, page_alloc: split smallest stolen page
in fallback") we pick the smallest (but sufficient) page of all that
have been stolen from a pageblock of different migratetype. However,
there are cases when we decide not to steal the whole pageblock.
Practically in the current implementation it means that we are trying to
fallback for a MIGRATE_MOVABLE allocation of order X, go through the
freelists from MAX_ORDER-1 down to X, and find free page of order Y. If
Y is less than pageblock_order / 2, we decide not to steal all pages
from the pageblock. When Y > X, it means we are potentially splitting a
larger page than we need, as there might be other pages of order Z,
where X <= Z < Y. Since Y is already too small to steal whole
pageblock, picking smallest available Z will result in the same decision
and we avoid splitting a higher-order page in a MIGRATE_UNMOVABLE or
MIGRATE_RECLAIMABLE pageblock.
This patch therefore changes the fallback algorithm so that in the
situation described above, we switch the fallback search strategy to go
from order X upwards to find the smallest suitable fallback. In theory
there shouldn't be a downside of this change wrt fragmentation.
This has been tested with mmtests' stress-highalloc performing
GFP_KERNEL order-4 allocations, here is the relevant extfrag tracepoint
statistics:
4.12.0-rc2 4.12.0-rc2
1-kernel4 2-kernel4
Page alloc extfrag event 25640976 69680977
Extfrag fragmenting 25621086 69661364
Extfrag fragmenting for unmovable 74409 73204
Extfrag fragmenting unmovable placed with movable 69003 67684
Extfrag fragmenting unmovable placed with reclaim. 5406 5520
Extfrag fragmenting for reclaimable 6398 8467
Extfrag fragmenting reclaimable placed with movable 869 884
Extfrag fragmenting reclaimable placed with unmov. 5529 7583
Extfrag fragmenting for movable 25540279 69579693
Since we force movable allocations to steal the smallest available page
(which we then practially always split), we steal less per fallback, so
the number of fallbacks increases and steals potentially happen from
different pageblocks. This is however not an issue for movable pages
that can be compacted.
Importantly, the "unmovable placed with movable" statistics is lower,
which is the result of less fragmentation in the unmovable pageblocks.
The effect on reclaimable allocation is a bit unclear.
Link: http://lkml.kernel.org/r/20170529093947.22618-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-10 22:47:14 +00:00
|
|
|
|
mm: reclaim small amounts of memory when an external fragmentation event occurs
An external fragmentation event was previously described as
When the page allocator fragments memory, it records the event using
the mm_page_alloc_extfrag event. If the fallback_order is smaller
than a pageblock order (order-9 on 64-bit x86) then it's considered
an event that will cause external fragmentation issues in the future.
The kernel reduces the probability of such events by increasing the
watermark sizes by calling set_recommended_min_free_kbytes early in the
lifetime of the system. This works reasonably well in general but if
there are enough sparsely populated pageblocks then the problem can still
occur as enough memory is free overall and kswapd stays asleep.
This patch introduces a watermark_boost_factor sysctl that allows a zone
watermark to be temporarily boosted when an external fragmentation causing
events occurs. The boosting will stall allocations that would decrease
free memory below the boosted low watermark and kswapd is woken if the
calling context allows to reclaim an amount of memory relative to the size
of the high watermark and the watermark_boost_factor until the boost is
cleared. When kswapd finishes, it wakes kcompactd at the pageblock order
to clean some of the pageblocks that may have been affected by the
fragmentation event. kswapd avoids any writeback, slab shrinkage and swap
from reclaim context during this operation to avoid excessive system
disruption in the name of fragmentation avoidance. Care is taken so that
kswapd will do normal reclaim work if the system is really low on memory.
This was evaluated using the same workloads as "mm, page_alloc: Spread
allocations across zones before introducing fragmentation".
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
4.20-rc3+patch1-4: 18421 (98% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-1 653.58 ( 0.00%) 652.71 ( 0.13%)
Amean fault-huge-1 0.00 ( 0.00%) 178.93 * -99.00%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 0.00 ( 0.00%) 5.12 ( 100.00%)
Note that external fragmentation causing events are massively reduced by
this path whether in comparison to the previous kernel or the vanilla
kernel. The fault latency for huge pages appears to be increased but that
is only because THP allocations were successful with the patch applied.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
4.20-rc3+patch1-4: 13464 (95% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Min fault-base-1 912.00 ( 0.00%) 905.00 ( 0.77%)
Min fault-huge-1 127.00 ( 0.00%) 135.00 ( -6.30%)
Amean fault-base-1 1467.55 ( 0.00%) 1481.67 ( -0.96%)
Amean fault-huge-1 1127.11 ( 0.00%) 1063.88 * 5.61%*
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-1 77.64 ( 0.00%) 83.46 ( 7.49%)
As before, massive reduction in external fragmentation events, some jitter
on latencies and an increase in THP allocation success rates.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
4.20-rc3+patch1-4: 14263 (93% reduction)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 1346.45 ( 0.00%) 1306.87 ( 2.94%)
Amean fault-huge-5 3418.60 ( 0.00%) 1348.94 ( 60.54%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 0.78 ( 0.00%) 7.91 ( 910.64%)
There is a 93% reduction in fragmentation causing events, there is a big
reduction in the huge page fault latency and allocation success rate is
higher.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
4.20-rc3+patch1-4: 11095 (93% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Amean fault-base-5 6217.43 ( 0.00%) 7419.67 * -19.34%*
Amean fault-huge-5 3163.33 ( 0.00%) 3263.80 ( -3.18%)
4.20.0-rc3 4.20.0-rc3
lowzone-v5r8 boost-v5r8
Percentage huge-5 95.14 ( 0.00%) 87.98 ( -7.53%)
There is a large reduction in fragmentation events with some jitter around
the latencies and success rates. As before, the high THP allocation
success rate does mean the system is under a lot of pressure. However, as
the fragmentation events are reduced, it would be expected that the
long-term allocation success rate would be higher.
Link: http://lkml.kernel.org/r/20181123114528.28802-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:52 +00:00
|
|
|
steal_suitable_fallback(zone, page, alloc_flags, start_migratetype,
|
|
|
|
can_steal);
|
mm, page_alloc: fallback to smallest page when not stealing whole pageblock
Since commit 3bc48f96cf11 ("mm, page_alloc: split smallest stolen page
in fallback") we pick the smallest (but sufficient) page of all that
have been stolen from a pageblock of different migratetype. However,
there are cases when we decide not to steal the whole pageblock.
Practically in the current implementation it means that we are trying to
fallback for a MIGRATE_MOVABLE allocation of order X, go through the
freelists from MAX_ORDER-1 down to X, and find free page of order Y. If
Y is less than pageblock_order / 2, we decide not to steal all pages
from the pageblock. When Y > X, it means we are potentially splitting a
larger page than we need, as there might be other pages of order Z,
where X <= Z < Y. Since Y is already too small to steal whole
pageblock, picking smallest available Z will result in the same decision
and we avoid splitting a higher-order page in a MIGRATE_UNMOVABLE or
MIGRATE_RECLAIMABLE pageblock.
This patch therefore changes the fallback algorithm so that in the
situation described above, we switch the fallback search strategy to go
from order X upwards to find the smallest suitable fallback. In theory
there shouldn't be a downside of this change wrt fragmentation.
This has been tested with mmtests' stress-highalloc performing
GFP_KERNEL order-4 allocations, here is the relevant extfrag tracepoint
statistics:
4.12.0-rc2 4.12.0-rc2
1-kernel4 2-kernel4
Page alloc extfrag event 25640976 69680977
Extfrag fragmenting 25621086 69661364
Extfrag fragmenting for unmovable 74409 73204
Extfrag fragmenting unmovable placed with movable 69003 67684
Extfrag fragmenting unmovable placed with reclaim. 5406 5520
Extfrag fragmenting for reclaimable 6398 8467
Extfrag fragmenting reclaimable placed with movable 869 884
Extfrag fragmenting reclaimable placed with unmov. 5529 7583
Extfrag fragmenting for movable 25540279 69579693
Since we force movable allocations to steal the smallest available page
(which we then practially always split), we steal less per fallback, so
the number of fallbacks increases and steals potentially happen from
different pageblocks. This is however not an issue for movable pages
that can be compacted.
Importantly, the "unmovable placed with movable" statistics is lower,
which is the result of less fragmentation in the unmovable pageblocks.
The effect on reclaimable allocation is a bit unclear.
Link: http://lkml.kernel.org/r/20170529093947.22618-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-10 22:47:14 +00:00
|
|
|
|
|
|
|
trace_mm_page_alloc_extfrag(page, order, current_order,
|
|
|
|
start_migratetype, fallback_mt);
|
|
|
|
|
|
|
|
return true;
|
|
|
|
|
2007-10-16 08:25:48 +00:00
|
|
|
}
|
|
|
|
|
Bias the location of pages freed for min_free_kbytes in the same MAX_ORDER_NR_PAGES blocks
The standard buddy allocator always favours the smallest block of pages.
The effect of this is that the pages free to satisfy min_free_kbytes tends
to be preserved since boot time at the same location of memory ffor a very
long time and as a contiguous block. When an administrator sets the
reserve at 16384 at boot time, it tends to be the same MAX_ORDER blocks
that remain free. This allows the occasional high atomic allocation to
succeed up until the point the blocks are split. In practice, it is
difficult to split these blocks but when they do split, the benefit of
having min_free_kbytes for contiguous blocks disappears. Additionally,
increasing min_free_kbytes once the system has been running for some time
has no guarantee of creating contiguous blocks.
On the other hand, CONFIG_PAGE_GROUP_BY_MOBILITY favours splitting large
blocks when there are no free pages of the appropriate type available. A
side-effect of this is that all blocks in memory tends to be used up and
the contiguous free blocks from boot time are not preserved like in the
vanilla allocator. This can cause a problem if a new caller is unwilling
to reclaim or does not reclaim for long enough.
A failure scenario was found for a wireless network device allocating
order-1 atomic allocations but the allocations were not intense or frequent
enough for a whole block of pages to be preserved for MIGRATE_HIGHALLOC.
This was reproduced on a desktop by booting with mem=256mb, forcing the
driver to allocate at order-1, running a bittorrent client (downloading a
debian ISO) and building a kernel with -j2.
This patch addresses the problem on the desktop machine booted with
mem=256mb. It works by setting aside a reserve of MAX_ORDER_NR_PAGES
blocks, the number of which depends on the value of min_free_kbytes. These
blocks are only fallen back to when there is no other free pages. Then the
smallest possible page is used just like the normal buddy allocator instead
of the largest possible page to preserve contiguous pages The pages in free
lists in the reserve blocks are never taken for another migrate type. The
results is that even if min_free_kbytes is set to a low value, contiguous
blocks will be preserved in the MIGRATE_RESERVE blocks.
This works better than the vanilla allocator because if min_free_kbytes is
increased, a new reserve block will be chosen based on the location of
reclaimable pages and the block will free up as contiguous pages. In the
vanilla allocator, no effort is made to target a block of pages to free as
contiguous pages and min_free_kbytes pages are scattered randomly.
This effect has been observed on the test machine. min_free_kbytes was set
initially low but it was kept as a contiguous free block within
MIGRATE_RESERVE. min_free_kbytes was then set to a higher value and over a
period of time, the free blocks were within the reserve and coalescing.
How long it takes to free up depends on how quickly LRU is rotating.
Amusingly, this means that more activity will free the blocks faster.
This mechanism potentially replaces MIGRATE_HIGHALLOC as it may be more
effective than grouping contiguous free pages together. It all depends on
whether the number of active atomic high allocations exceeds
min_free_kbytes or not. If the number of active allocations exceeds
min_free_kbytes, it's worth it but maybe in that situation, min_free_kbytes
should be set higher. Once there are no more reports of allocation
failures, a patch will be submitted that backs out MIGRATE_HIGHALLOC and
see if the reports stay missing.
Credit to Mariusz Kozlowski for discovering the problem, describing the
failure scenario and testing patches and scenarios.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:58 +00:00
|
|
|
/*
|
2005-04-16 22:20:36 +00:00
|
|
|
* Do the hard work of removing an element from the buddy allocator.
|
|
|
|
* Call me with the zone->lock already held.
|
|
|
|
*/
|
2017-11-16 01:36:53 +00:00
|
|
|
static __always_inline struct page *
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
__rmqueue(struct zone *zone, unsigned int order, int migratetype,
|
|
|
|
unsigned int alloc_flags)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct page *page;
|
|
|
|
|
2021-01-12 23:49:08 +00:00
|
|
|
if (IS_ENABLED(CONFIG_CMA)) {
|
|
|
|
/*
|
|
|
|
* Balance movable allocations between regular and CMA areas by
|
|
|
|
* allocating from CMA when over half of the zone's free memory
|
|
|
|
* is in the CMA area.
|
|
|
|
*/
|
|
|
|
if (alloc_flags & ALLOC_CMA &&
|
|
|
|
zone_page_state(zone, NR_FREE_CMA_PAGES) >
|
|
|
|
zone_page_state(zone, NR_FREE_PAGES) / 2) {
|
|
|
|
page = __rmqueue_cma_fallback(zone, order);
|
|
|
|
if (page)
|
|
|
|
goto out;
|
|
|
|
}
|
2020-06-03 22:58:42 +00:00
|
|
|
}
|
2017-05-08 22:54:37 +00:00
|
|
|
retry:
|
Bias the location of pages freed for min_free_kbytes in the same MAX_ORDER_NR_PAGES blocks
The standard buddy allocator always favours the smallest block of pages.
The effect of this is that the pages free to satisfy min_free_kbytes tends
to be preserved since boot time at the same location of memory ffor a very
long time and as a contiguous block. When an administrator sets the
reserve at 16384 at boot time, it tends to be the same MAX_ORDER blocks
that remain free. This allows the occasional high atomic allocation to
succeed up until the point the blocks are split. In practice, it is
difficult to split these blocks but when they do split, the benefit of
having min_free_kbytes for contiguous blocks disappears. Additionally,
increasing min_free_kbytes once the system has been running for some time
has no guarantee of creating contiguous blocks.
On the other hand, CONFIG_PAGE_GROUP_BY_MOBILITY favours splitting large
blocks when there are no free pages of the appropriate type available. A
side-effect of this is that all blocks in memory tends to be used up and
the contiguous free blocks from boot time are not preserved like in the
vanilla allocator. This can cause a problem if a new caller is unwilling
to reclaim or does not reclaim for long enough.
A failure scenario was found for a wireless network device allocating
order-1 atomic allocations but the allocations were not intense or frequent
enough for a whole block of pages to be preserved for MIGRATE_HIGHALLOC.
This was reproduced on a desktop by booting with mem=256mb, forcing the
driver to allocate at order-1, running a bittorrent client (downloading a
debian ISO) and building a kernel with -j2.
This patch addresses the problem on the desktop machine booted with
mem=256mb. It works by setting aside a reserve of MAX_ORDER_NR_PAGES
blocks, the number of which depends on the value of min_free_kbytes. These
blocks are only fallen back to when there is no other free pages. Then the
smallest possible page is used just like the normal buddy allocator instead
of the largest possible page to preserve contiguous pages The pages in free
lists in the reserve blocks are never taken for another migrate type. The
results is that even if min_free_kbytes is set to a low value, contiguous
blocks will be preserved in the MIGRATE_RESERVE blocks.
This works better than the vanilla allocator because if min_free_kbytes is
increased, a new reserve block will be chosen based on the location of
reclaimable pages and the block will free up as contiguous pages. In the
vanilla allocator, no effort is made to target a block of pages to free as
contiguous pages and min_free_kbytes pages are scattered randomly.
This effect has been observed on the test machine. min_free_kbytes was set
initially low but it was kept as a contiguous free block within
MIGRATE_RESERVE. min_free_kbytes was then set to a higher value and over a
period of time, the free blocks were within the reserve and coalescing.
How long it takes to free up depends on how quickly LRU is rotating.
Amusingly, this means that more activity will free the blocks faster.
This mechanism potentially replaces MIGRATE_HIGHALLOC as it may be more
effective than grouping contiguous free pages together. It all depends on
whether the number of active atomic high allocations exceeds
min_free_kbytes or not. If the number of active allocations exceeds
min_free_kbytes, it's worth it but maybe in that situation, min_free_kbytes
should be set higher. Once there are no more reports of allocation
failures, a patch will be submitted that backs out MIGRATE_HIGHALLOC and
see if the reports stay missing.
Credit to Mariusz Kozlowski for discovering the problem, describing the
failure scenario and testing patches and scenarios.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:58 +00:00
|
|
|
page = __rmqueue_smallest(zone, order, migratetype);
|
2015-11-07 00:28:34 +00:00
|
|
|
if (unlikely(!page)) {
|
2020-08-07 06:26:04 +00:00
|
|
|
if (alloc_flags & ALLOC_CMA)
|
2015-04-14 22:45:15 +00:00
|
|
|
page = __rmqueue_cma_fallback(zone, order);
|
|
|
|
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
if (!page && __rmqueue_fallback(zone, order, migratetype,
|
|
|
|
alloc_flags))
|
2017-05-08 22:54:37 +00:00
|
|
|
goto retry;
|
2009-06-16 22:32:04 +00:00
|
|
|
}
|
2021-01-12 23:49:08 +00:00
|
|
|
out:
|
|
|
|
if (page)
|
|
|
|
trace_mm_page_alloc_zone_locked(page, order, migratetype);
|
2007-10-16 08:25:48 +00:00
|
|
|
return page;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2012-01-11 14:16:11 +00:00
|
|
|
/*
|
2005-04-16 22:20:36 +00:00
|
|
|
* Obtain a specified number of elements from the buddy allocator, all under
|
|
|
|
* a single hold of the lock, for efficiency. Add them to the supplied list.
|
|
|
|
* Returns the number of new pages which were placed at *list.
|
|
|
|
*/
|
2012-01-11 14:16:11 +00:00
|
|
|
static int rmqueue_bulk(struct zone *zone, unsigned int order,
|
2007-10-16 08:25:48 +00:00
|
|
|
unsigned long count, struct list_head *list,
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
int migratetype, unsigned int alloc_flags)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2016-12-13 00:44:41 +00:00
|
|
|
int i, alloced = 0;
|
2012-01-11 14:16:11 +00:00
|
|
|
|
2017-04-20 21:37:43 +00:00
|
|
|
spin_lock(&zone->lock);
|
2005-04-16 22:20:36 +00:00
|
|
|
for (i = 0; i < count; ++i) {
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
struct page *page = __rmqueue(zone, order, migratetype,
|
|
|
|
alloc_flags);
|
2006-01-06 08:11:01 +00:00
|
|
|
if (unlikely(page == NULL))
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
2007-12-18 00:20:05 +00:00
|
|
|
|
2016-05-20 00:14:35 +00:00
|
|
|
if (unlikely(check_pcp_refill(page)))
|
|
|
|
continue;
|
|
|
|
|
2007-12-18 00:20:05 +00:00
|
|
|
/*
|
2017-11-16 01:38:07 +00:00
|
|
|
* Split buddy pages returned by expand() are received here in
|
|
|
|
* physical page order. The page is added to the tail of
|
|
|
|
* caller's list. From the callers perspective, the linked list
|
|
|
|
* is ordered by page number under some conditions. This is
|
|
|
|
* useful for IO devices that can forward direction from the
|
|
|
|
* head, thus also in the physical page order. This is useful
|
|
|
|
* for IO devices that can merge IO requests if the physical
|
|
|
|
* pages are ordered properly.
|
2007-12-18 00:20:05 +00:00
|
|
|
*/
|
2017-11-16 01:38:07 +00:00
|
|
|
list_add_tail(&page->lru, list);
|
2016-12-13 00:44:41 +00:00
|
|
|
alloced++;
|
2015-09-08 22:01:25 +00:00
|
|
|
if (is_migrate_cma(get_pcppage_migratetype(page)))
|
2012-10-08 23:32:02 +00:00
|
|
|
__mod_zone_page_state(zone, NR_FREE_CMA_PAGES,
|
|
|
|
-(1 << order));
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2016-12-13 00:44:41 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* i pages were removed from the buddy list even if some leak due
|
|
|
|
* to check_pcp_refill failing so adjust NR_FREE_PAGES based
|
|
|
|
* on i. Do not confuse with 'alloced' which is the number of
|
|
|
|
* pages added to the pcp list.
|
|
|
|
*/
|
2009-06-16 22:32:13 +00:00
|
|
|
__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
|
2017-04-20 21:37:43 +00:00
|
|
|
spin_unlock(&zone->lock);
|
2016-12-13 00:44:41 +00:00
|
|
|
return alloced;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2005-06-22 00:14:57 +00:00
|
|
|
#ifdef CONFIG_NUMA
|
2006-03-10 01:33:54 +00:00
|
|
|
/*
|
2007-05-09 09:35:14 +00:00
|
|
|
* Called from the vmstat counter updater to drain pagesets of this
|
|
|
|
* currently executing processor on remote nodes after they have
|
|
|
|
* expired.
|
|
|
|
*
|
2006-03-22 08:09:08 +00:00
|
|
|
* Note that this function must be called with the thread pinned to
|
|
|
|
* a single processor.
|
2006-03-10 01:33:54 +00:00
|
|
|
*/
|
2007-05-09 09:35:14 +00:00
|
|
|
void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp)
|
2005-06-22 00:14:57 +00:00
|
|
|
{
|
|
|
|
unsigned long flags;
|
2014-08-06 23:05:15 +00:00
|
|
|
int to_drain, batch;
|
2005-06-22 00:14:57 +00:00
|
|
|
|
2007-05-09 09:35:14 +00:00
|
|
|
local_irq_save(flags);
|
2015-04-15 23:14:08 +00:00
|
|
|
batch = READ_ONCE(pcp->batch);
|
2014-08-06 23:05:15 +00:00
|
|
|
to_drain = min(pcp->count, batch);
|
2018-04-05 23:24:06 +00:00
|
|
|
if (to_drain > 0)
|
2012-07-31 23:42:53 +00:00
|
|
|
free_pcppages_bulk(zone, to_drain, pcp);
|
2007-05-09 09:35:14 +00:00
|
|
|
local_irq_restore(flags);
|
2005-06-22 00:14:57 +00:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-02-05 06:29:11 +00:00
|
|
|
/*
|
2014-12-10 23:43:01 +00:00
|
|
|
* Drain pcplists of the indicated processor and zone.
|
2008-02-05 06:29:11 +00:00
|
|
|
*
|
|
|
|
* The processor must either be the current processor and the
|
|
|
|
* thread pinned to the current processor or a processor that
|
|
|
|
* is not online.
|
|
|
|
*/
|
2014-12-10 23:43:01 +00:00
|
|
|
static void drain_pages_zone(unsigned int cpu, struct zone *zone)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-01-06 08:10:56 +00:00
|
|
|
unsigned long flags;
|
2014-12-10 23:43:01 +00:00
|
|
|
struct per_cpu_pageset *pset;
|
|
|
|
struct per_cpu_pages *pcp;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2014-12-10 23:43:01 +00:00
|
|
|
local_irq_save(flags);
|
|
|
|
pset = per_cpu_ptr(zone->pageset, cpu);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2014-12-10 23:43:01 +00:00
|
|
|
pcp = &pset->pcp;
|
2018-04-05 23:24:06 +00:00
|
|
|
if (pcp->count)
|
2014-12-10 23:43:01 +00:00
|
|
|
free_pcppages_bulk(zone, pcp->count, pcp);
|
|
|
|
local_irq_restore(flags);
|
|
|
|
}
|
2008-02-05 06:29:19 +00:00
|
|
|
|
2014-12-10 23:43:01 +00:00
|
|
|
/*
|
|
|
|
* Drain pcplists of all zones on the indicated processor.
|
|
|
|
*
|
|
|
|
* The processor must either be the current processor and the
|
|
|
|
* thread pinned to the current processor or a processor that
|
|
|
|
* is not online.
|
|
|
|
*/
|
|
|
|
static void drain_pages(unsigned int cpu)
|
|
|
|
{
|
|
|
|
struct zone *zone;
|
|
|
|
|
|
|
|
for_each_populated_zone(zone) {
|
|
|
|
drain_pages_zone(cpu, zone);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-02-05 06:29:11 +00:00
|
|
|
/*
|
|
|
|
* Spill all of this CPU's per-cpu pages back into the buddy allocator.
|
2014-12-10 23:43:01 +00:00
|
|
|
*
|
|
|
|
* The CPU has to be pinned. When zone parameter is non-NULL, spill just
|
|
|
|
* the single zone's pages.
|
2008-02-05 06:29:11 +00:00
|
|
|
*/
|
2014-12-10 23:43:01 +00:00
|
|
|
void drain_local_pages(struct zone *zone)
|
2008-02-05 06:29:11 +00:00
|
|
|
{
|
2014-12-10 23:43:01 +00:00
|
|
|
int cpu = smp_processor_id();
|
|
|
|
|
|
|
|
if (zone)
|
|
|
|
drain_pages_zone(cpu, zone);
|
|
|
|
else
|
|
|
|
drain_pages(cpu);
|
2008-02-05 06:29:11 +00:00
|
|
|
}
|
|
|
|
|
2017-02-24 22:56:32 +00:00
|
|
|
static void drain_local_pages_wq(struct work_struct *work)
|
|
|
|
{
|
2018-12-28 08:38:58 +00:00
|
|
|
struct pcpu_drain *drain;
|
|
|
|
|
|
|
|
drain = container_of(work, struct pcpu_drain, work);
|
|
|
|
|
2017-02-24 22:56:35 +00:00
|
|
|
/*
|
|
|
|
* drain_all_pages doesn't use proper cpu hotplug protection so
|
|
|
|
* we can race with cpu offline when the WQ can move this from
|
|
|
|
* a cpu pinned worker to an unbound one. We can operate on a different
|
|
|
|
* cpu which is allright but we also have to make sure to not move to
|
|
|
|
* a different one.
|
|
|
|
*/
|
|
|
|
preempt_disable();
|
2018-12-28 08:38:58 +00:00
|
|
|
drain_local_pages(drain->zone);
|
2017-02-24 22:56:35 +00:00
|
|
|
preempt_enable();
|
2017-02-24 22:56:32 +00:00
|
|
|
}
|
|
|
|
|
2008-02-05 06:29:11 +00:00
|
|
|
/*
|
mm, page_alloc: disable pcplists during memory offline
Memory offlining relies on page isolation to guarantee a forward progress
because pages cannot be reused while they are isolated. But the page
isolation itself doesn't prevent from races while freed pages are stored
on pcp lists and thus can be reused. This can be worked around by
repeated draining of pcplists, as done by commit 968318261221
("mm/memory_hotplug: drain per-cpu pages again during memory offline").
David and Michal would prefer that this race was closed in a way that
callers of page isolation who need stronger guarantees don't need to
repeatedly drain. David suggested disabling pcplists usage completely
during page isolation, instead of repeatedly draining them.
To achieve this without adding special cases in alloc/free fastpath, we
can use the same approach as boot pagesets - when pcp->high is 0, any
pcplist addition will be immediately flushed.
The race can thus be closed by setting pcp->high to 0 and draining
pcplists once, before calling start_isolate_page_range(). The draining
will serialize after processes that already disabled interrupts and read
the old value of pcp->high in free_unref_page_commit(), and processes that
have not yet disabled interrupts, will observe pcp->high == 0 when they
are rescheduled, and skip pcplists. This guarantees no stray pages on
pcplists in zones where isolation happens.
This patch thus adds zone_pcp_disable() and zone_pcp_enable() functions
that page isolation users can call before start_isolate_page_range() and
after unisolating (or offlining) the isolated pages.
Also, drain_all_pages() is optimized to only execute on cpus where
pcplists are not empty. The check can however race with a free to pcplist
that has not yet increased the pcp->count from 0 to 1. Thus make the
drain optionally skip the racy check and drain on all cpus, and use this
option in zone_pcp_disable().
As we have to avoid external updates to high and batch while pcplists are
disabled, we take pcp_batch_high_lock in zone_pcp_disable() and release it
in zone_pcp_enable(). This also synchronizes multiple users of
zone_pcp_disable()/enable().
Currently the only user of this functionality is offline_pages().
[vbabka@suse.cz: add comment, per David]
Link: https://lkml.kernel.org/r/527480ef-ed72-e1c1-52a0-1c5b0113df45@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Suggested-by: David Hildenbrand <david@redhat.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:59 +00:00
|
|
|
* The implementation of drain_all_pages(), exposing an extra parameter to
|
|
|
|
* drain on all cpus.
|
2014-12-10 23:43:01 +00:00
|
|
|
*
|
mm, page_alloc: disable pcplists during memory offline
Memory offlining relies on page isolation to guarantee a forward progress
because pages cannot be reused while they are isolated. But the page
isolation itself doesn't prevent from races while freed pages are stored
on pcp lists and thus can be reused. This can be worked around by
repeated draining of pcplists, as done by commit 968318261221
("mm/memory_hotplug: drain per-cpu pages again during memory offline").
David and Michal would prefer that this race was closed in a way that
callers of page isolation who need stronger guarantees don't need to
repeatedly drain. David suggested disabling pcplists usage completely
during page isolation, instead of repeatedly draining them.
To achieve this without adding special cases in alloc/free fastpath, we
can use the same approach as boot pagesets - when pcp->high is 0, any
pcplist addition will be immediately flushed.
The race can thus be closed by setting pcp->high to 0 and draining
pcplists once, before calling start_isolate_page_range(). The draining
will serialize after processes that already disabled interrupts and read
the old value of pcp->high in free_unref_page_commit(), and processes that
have not yet disabled interrupts, will observe pcp->high == 0 when they
are rescheduled, and skip pcplists. This guarantees no stray pages on
pcplists in zones where isolation happens.
This patch thus adds zone_pcp_disable() and zone_pcp_enable() functions
that page isolation users can call before start_isolate_page_range() and
after unisolating (or offlining) the isolated pages.
Also, drain_all_pages() is optimized to only execute on cpus where
pcplists are not empty. The check can however race with a free to pcplist
that has not yet increased the pcp->count from 0 to 1. Thus make the
drain optionally skip the racy check and drain on all cpus, and use this
option in zone_pcp_disable().
As we have to avoid external updates to high and batch while pcplists are
disabled, we take pcp_batch_high_lock in zone_pcp_disable() and release it
in zone_pcp_enable(). This also synchronizes multiple users of
zone_pcp_disable()/enable().
Currently the only user of this functionality is offline_pages().
[vbabka@suse.cz: add comment, per David]
Link: https://lkml.kernel.org/r/527480ef-ed72-e1c1-52a0-1c5b0113df45@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Suggested-by: David Hildenbrand <david@redhat.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:59 +00:00
|
|
|
* drain_all_pages() is optimized to only execute on cpus where pcplists are
|
|
|
|
* not empty. The check for non-emptiness can however race with a free to
|
|
|
|
* pcplist that has not yet increased the pcp->count from 0 to 1. Callers
|
|
|
|
* that need the guarantee that every CPU has drained can disable the
|
|
|
|
* optimizing racy check.
|
2008-02-05 06:29:11 +00:00
|
|
|
*/
|
2020-12-15 03:11:12 +00:00
|
|
|
static void __drain_all_pages(struct zone *zone, bool force_all_cpus)
|
2008-02-05 06:29:11 +00:00
|
|
|
{
|
2012-03-28 21:42:45 +00:00
|
|
|
int cpu;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate in the BSS so we wont require allocation in
|
|
|
|
* direct reclaim path for CONFIG_CPUMASK_OFFSTACK=y
|
|
|
|
*/
|
|
|
|
static cpumask_t cpus_with_pcps;
|
|
|
|
|
2017-04-07 23:05:05 +00:00
|
|
|
/*
|
|
|
|
* Make sure nobody triggers this path before mm_percpu_wq is fully
|
|
|
|
* initialized.
|
|
|
|
*/
|
|
|
|
if (WARN_ON_ONCE(!mm_percpu_wq))
|
|
|
|
return;
|
|
|
|
|
2017-02-24 22:56:56 +00:00
|
|
|
/*
|
|
|
|
* Do not drain if one is already in progress unless it's specific to
|
|
|
|
* a zone. Such callers are primarily CMA and memory hotplug and need
|
|
|
|
* the drain to be complete when the call returns.
|
|
|
|
*/
|
|
|
|
if (unlikely(!mutex_trylock(&pcpu_drain_mutex))) {
|
|
|
|
if (!zone)
|
|
|
|
return;
|
|
|
|
mutex_lock(&pcpu_drain_mutex);
|
|
|
|
}
|
2017-02-24 22:56:32 +00:00
|
|
|
|
2012-03-28 21:42:45 +00:00
|
|
|
/*
|
|
|
|
* We don't care about racing with CPU hotplug event
|
|
|
|
* as offline notification will cause the notified
|
|
|
|
* cpu to drain that CPU pcps and on_each_cpu_mask
|
|
|
|
* disables preemption as part of its processing
|
|
|
|
*/
|
|
|
|
for_each_online_cpu(cpu) {
|
2014-12-10 23:43:01 +00:00
|
|
|
struct per_cpu_pageset *pcp;
|
|
|
|
struct zone *z;
|
2012-03-28 21:42:45 +00:00
|
|
|
bool has_pcps = false;
|
2014-12-10 23:43:01 +00:00
|
|
|
|
mm, page_alloc: disable pcplists during memory offline
Memory offlining relies on page isolation to guarantee a forward progress
because pages cannot be reused while they are isolated. But the page
isolation itself doesn't prevent from races while freed pages are stored
on pcp lists and thus can be reused. This can be worked around by
repeated draining of pcplists, as done by commit 968318261221
("mm/memory_hotplug: drain per-cpu pages again during memory offline").
David and Michal would prefer that this race was closed in a way that
callers of page isolation who need stronger guarantees don't need to
repeatedly drain. David suggested disabling pcplists usage completely
during page isolation, instead of repeatedly draining them.
To achieve this without adding special cases in alloc/free fastpath, we
can use the same approach as boot pagesets - when pcp->high is 0, any
pcplist addition will be immediately flushed.
The race can thus be closed by setting pcp->high to 0 and draining
pcplists once, before calling start_isolate_page_range(). The draining
will serialize after processes that already disabled interrupts and read
the old value of pcp->high in free_unref_page_commit(), and processes that
have not yet disabled interrupts, will observe pcp->high == 0 when they
are rescheduled, and skip pcplists. This guarantees no stray pages on
pcplists in zones where isolation happens.
This patch thus adds zone_pcp_disable() and zone_pcp_enable() functions
that page isolation users can call before start_isolate_page_range() and
after unisolating (or offlining) the isolated pages.
Also, drain_all_pages() is optimized to only execute on cpus where
pcplists are not empty. The check can however race with a free to pcplist
that has not yet increased the pcp->count from 0 to 1. Thus make the
drain optionally skip the racy check and drain on all cpus, and use this
option in zone_pcp_disable().
As we have to avoid external updates to high and batch while pcplists are
disabled, we take pcp_batch_high_lock in zone_pcp_disable() and release it
in zone_pcp_enable(). This also synchronizes multiple users of
zone_pcp_disable()/enable().
Currently the only user of this functionality is offline_pages().
[vbabka@suse.cz: add comment, per David]
Link: https://lkml.kernel.org/r/527480ef-ed72-e1c1-52a0-1c5b0113df45@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Suggested-by: David Hildenbrand <david@redhat.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:59 +00:00
|
|
|
if (force_all_cpus) {
|
|
|
|
/*
|
|
|
|
* The pcp.count check is racy, some callers need a
|
|
|
|
* guarantee that no cpu is missed.
|
|
|
|
*/
|
|
|
|
has_pcps = true;
|
|
|
|
} else if (zone) {
|
2012-03-28 21:42:45 +00:00
|
|
|
pcp = per_cpu_ptr(zone->pageset, cpu);
|
2014-12-10 23:43:01 +00:00
|
|
|
if (pcp->pcp.count)
|
2012-03-28 21:42:45 +00:00
|
|
|
has_pcps = true;
|
2014-12-10 23:43:01 +00:00
|
|
|
} else {
|
|
|
|
for_each_populated_zone(z) {
|
|
|
|
pcp = per_cpu_ptr(z->pageset, cpu);
|
|
|
|
if (pcp->pcp.count) {
|
|
|
|
has_pcps = true;
|
|
|
|
break;
|
|
|
|
}
|
2012-03-28 21:42:45 +00:00
|
|
|
}
|
|
|
|
}
|
2014-12-10 23:43:01 +00:00
|
|
|
|
2012-03-28 21:42:45 +00:00
|
|
|
if (has_pcps)
|
|
|
|
cpumask_set_cpu(cpu, &cpus_with_pcps);
|
|
|
|
else
|
|
|
|
cpumask_clear_cpu(cpu, &cpus_with_pcps);
|
|
|
|
}
|
2017-02-24 22:56:32 +00:00
|
|
|
|
2017-02-24 22:56:56 +00:00
|
|
|
for_each_cpu(cpu, &cpus_with_pcps) {
|
2018-12-28 08:38:58 +00:00
|
|
|
struct pcpu_drain *drain = per_cpu_ptr(&pcpu_drain, cpu);
|
|
|
|
|
|
|
|
drain->zone = zone;
|
|
|
|
INIT_WORK(&drain->work, drain_local_pages_wq);
|
|
|
|
queue_work_on(cpu, mm_percpu_wq, &drain->work);
|
2017-02-24 22:56:32 +00:00
|
|
|
}
|
2017-02-24 22:56:56 +00:00
|
|
|
for_each_cpu(cpu, &cpus_with_pcps)
|
2018-12-28 08:38:58 +00:00
|
|
|
flush_work(&per_cpu_ptr(&pcpu_drain, cpu)->work);
|
2017-02-24 22:56:56 +00:00
|
|
|
|
|
|
|
mutex_unlock(&pcpu_drain_mutex);
|
2008-02-05 06:29:11 +00:00
|
|
|
}
|
|
|
|
|
mm, page_alloc: disable pcplists during memory offline
Memory offlining relies on page isolation to guarantee a forward progress
because pages cannot be reused while they are isolated. But the page
isolation itself doesn't prevent from races while freed pages are stored
on pcp lists and thus can be reused. This can be worked around by
repeated draining of pcplists, as done by commit 968318261221
("mm/memory_hotplug: drain per-cpu pages again during memory offline").
David and Michal would prefer that this race was closed in a way that
callers of page isolation who need stronger guarantees don't need to
repeatedly drain. David suggested disabling pcplists usage completely
during page isolation, instead of repeatedly draining them.
To achieve this without adding special cases in alloc/free fastpath, we
can use the same approach as boot pagesets - when pcp->high is 0, any
pcplist addition will be immediately flushed.
The race can thus be closed by setting pcp->high to 0 and draining
pcplists once, before calling start_isolate_page_range(). The draining
will serialize after processes that already disabled interrupts and read
the old value of pcp->high in free_unref_page_commit(), and processes that
have not yet disabled interrupts, will observe pcp->high == 0 when they
are rescheduled, and skip pcplists. This guarantees no stray pages on
pcplists in zones where isolation happens.
This patch thus adds zone_pcp_disable() and zone_pcp_enable() functions
that page isolation users can call before start_isolate_page_range() and
after unisolating (or offlining) the isolated pages.
Also, drain_all_pages() is optimized to only execute on cpus where
pcplists are not empty. The check can however race with a free to pcplist
that has not yet increased the pcp->count from 0 to 1. Thus make the
drain optionally skip the racy check and drain on all cpus, and use this
option in zone_pcp_disable().
As we have to avoid external updates to high and batch while pcplists are
disabled, we take pcp_batch_high_lock in zone_pcp_disable() and release it
in zone_pcp_enable(). This also synchronizes multiple users of
zone_pcp_disable()/enable().
Currently the only user of this functionality is offline_pages().
[vbabka@suse.cz: add comment, per David]
Link: https://lkml.kernel.org/r/527480ef-ed72-e1c1-52a0-1c5b0113df45@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Suggested-by: David Hildenbrand <david@redhat.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:59 +00:00
|
|
|
/*
|
|
|
|
* Spill all the per-cpu pages from all CPUs back into the buddy allocator.
|
|
|
|
*
|
|
|
|
* When zone parameter is non-NULL, spill just the single zone's pages.
|
|
|
|
*
|
|
|
|
* Note that this can be extremely slow as the draining happens in a workqueue.
|
|
|
|
*/
|
|
|
|
void drain_all_pages(struct zone *zone)
|
|
|
|
{
|
|
|
|
__drain_all_pages(zone, false);
|
|
|
|
}
|
|
|
|
|
2007-07-29 21:27:18 +00:00
|
|
|
#ifdef CONFIG_HIBERNATION
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2017-08-25 22:55:30 +00:00
|
|
|
/*
|
|
|
|
* Touch the watchdog for every WD_PAGE_COUNT pages.
|
|
|
|
*/
|
|
|
|
#define WD_PAGE_COUNT (128*1024)
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
void mark_free_pages(struct zone *zone)
|
|
|
|
{
|
2017-08-25 22:55:30 +00:00
|
|
|
unsigned long pfn, max_zone_pfn, page_count = WD_PAGE_COUNT;
|
2006-09-26 06:32:49 +00:00
|
|
|
unsigned long flags;
|
2014-06-04 23:10:21 +00:00
|
|
|
unsigned int order, t;
|
2016-01-14 23:20:33 +00:00
|
|
|
struct page *page;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-09-11 21:21:45 +00:00
|
|
|
if (zone_is_empty(zone))
|
2005-04-16 22:20:36 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
2006-09-26 06:32:49 +00:00
|
|
|
|
2013-02-23 00:35:23 +00:00
|
|
|
max_zone_pfn = zone_end_pfn(zone);
|
2006-09-26 06:32:49 +00:00
|
|
|
for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
|
|
|
|
if (pfn_valid(pfn)) {
|
2016-01-14 23:20:33 +00:00
|
|
|
page = pfn_to_page(pfn);
|
2016-05-20 00:12:16 +00:00
|
|
|
|
2017-08-25 22:55:30 +00:00
|
|
|
if (!--page_count) {
|
|
|
|
touch_nmi_watchdog();
|
|
|
|
page_count = WD_PAGE_COUNT;
|
|
|
|
}
|
|
|
|
|
2016-05-20 00:12:16 +00:00
|
|
|
if (page_zone(page) != zone)
|
|
|
|
continue;
|
|
|
|
|
2007-05-06 21:50:42 +00:00
|
|
|
if (!swsusp_page_is_forbidden(page))
|
|
|
|
swsusp_unset_page_free(page);
|
2006-09-26 06:32:49 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2007-10-16 08:25:48 +00:00
|
|
|
for_each_migratetype_order(order, t) {
|
2016-01-14 23:20:33 +00:00
|
|
|
list_for_each_entry(page,
|
|
|
|
&zone->free_area[order].free_list[t], lru) {
|
2006-09-26 06:32:49 +00:00
|
|
|
unsigned long i;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2016-01-14 23:20:33 +00:00
|
|
|
pfn = page_to_pfn(page);
|
2017-08-25 22:55:30 +00:00
|
|
|
for (i = 0; i < (1UL << order); i++) {
|
|
|
|
if (!--page_count) {
|
|
|
|
touch_nmi_watchdog();
|
|
|
|
page_count = WD_PAGE_COUNT;
|
|
|
|
}
|
2007-05-06 21:50:42 +00:00
|
|
|
swsusp_set_page_free(pfn_to_page(pfn + i));
|
2017-08-25 22:55:30 +00:00
|
|
|
}
|
2006-09-26 06:32:49 +00:00
|
|
|
}
|
2007-10-16 08:25:48 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
|
|
|
}
|
2007-10-16 08:25:50 +00:00
|
|
|
#endif /* CONFIG_PM */
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2017-11-16 01:37:59 +00:00
|
|
|
static bool free_unref_page_prepare(struct page *page, unsigned long pfn)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
int migratetype;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2016-05-20 00:14:32 +00:00
|
|
|
if (!free_pcp_prepare(page))
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
return false;
|
2005-11-22 05:32:20 +00:00
|
|
|
|
2014-06-04 23:10:17 +00:00
|
|
|
migratetype = get_pfnblock_migratetype(page, pfn);
|
2015-09-08 22:01:25 +00:00
|
|
|
set_pcppage_migratetype(page, migratetype);
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2017-11-16 01:37:59 +00:00
|
|
|
static void free_unref_page_commit(struct page *page, unsigned long pfn)
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
{
|
|
|
|
struct zone *zone = page_zone(page);
|
|
|
|
struct per_cpu_pages *pcp;
|
|
|
|
int migratetype;
|
|
|
|
|
|
|
|
migratetype = get_pcppage_migratetype(page);
|
2017-04-20 21:37:43 +00:00
|
|
|
__count_vm_event(PGFREE);
|
2009-06-16 22:32:08 +00:00
|
|
|
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
/*
|
|
|
|
* We only track unmovable, reclaimable and movable on pcp lists.
|
|
|
|
* Free ISOLATE pages back to the allocator because they are being
|
2017-05-03 21:52:52 +00:00
|
|
|
* offlined but treat HIGHATOMIC as movable pages so we can get those
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
* areas back if necessary. Otherwise, we may have to free
|
|
|
|
* excessively into the page allocator
|
|
|
|
*/
|
|
|
|
if (migratetype >= MIGRATE_PCPTYPES) {
|
2013-02-23 00:33:58 +00:00
|
|
|
if (unlikely(is_migrate_isolate(migratetype))) {
|
2020-10-16 03:09:35 +00:00
|
|
|
free_one_page(zone, page, pfn, 0, migratetype,
|
|
|
|
FPI_NONE);
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
return;
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
}
|
|
|
|
migratetype = MIGRATE_MOVABLE;
|
|
|
|
}
|
|
|
|
|
2010-01-05 06:34:51 +00:00
|
|
|
pcp = &this_cpu_ptr(zone->pageset)->pcp;
|
2017-11-16 01:37:59 +00:00
|
|
|
list_add(&page->lru, &pcp->lists[migratetype]);
|
2005-04-16 22:20:36 +00:00
|
|
|
pcp->count++;
|
2020-12-15 03:10:50 +00:00
|
|
|
if (pcp->count >= READ_ONCE(pcp->high))
|
|
|
|
free_pcppages_bulk(zone, READ_ONCE(pcp->batch), pcp);
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
}
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
/*
|
|
|
|
* Free a 0-order page
|
|
|
|
*/
|
2017-11-16 01:37:59 +00:00
|
|
|
void free_unref_page(struct page *page)
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
unsigned long pfn = page_to_pfn(page);
|
|
|
|
|
2017-11-16 01:37:59 +00:00
|
|
|
if (!free_unref_page_prepare(page, pfn))
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
local_irq_save(flags);
|
2017-11-16 01:37:59 +00:00
|
|
|
free_unref_page_commit(page, pfn);
|
2017-04-20 21:37:43 +00:00
|
|
|
local_irq_restore(flags);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2012-01-10 23:07:04 +00:00
|
|
|
/*
|
|
|
|
* Free a list of 0-order pages
|
|
|
|
*/
|
2017-11-16 01:37:59 +00:00
|
|
|
void free_unref_page_list(struct list_head *list)
|
2012-01-10 23:07:04 +00:00
|
|
|
{
|
|
|
|
struct page *page, *next;
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
unsigned long flags, pfn;
|
2017-12-14 23:32:55 +00:00
|
|
|
int batch_count = 0;
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
|
|
|
|
/* Prepare pages for freeing */
|
|
|
|
list_for_each_entry_safe(page, next, list, lru) {
|
|
|
|
pfn = page_to_pfn(page);
|
2017-11-16 01:37:59 +00:00
|
|
|
if (!free_unref_page_prepare(page, pfn))
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
list_del(&page->lru);
|
|
|
|
set_page_private(page, pfn);
|
|
|
|
}
|
2012-01-10 23:07:04 +00:00
|
|
|
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
local_irq_save(flags);
|
2012-01-10 23:07:04 +00:00
|
|
|
list_for_each_entry_safe(page, next, list, lru) {
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
unsigned long pfn = page_private(page);
|
|
|
|
|
|
|
|
set_page_private(page, 0);
|
2017-11-16 01:37:59 +00:00
|
|
|
trace_mm_page_free_batched(page);
|
|
|
|
free_unref_page_commit(page, pfn);
|
2017-12-14 23:32:55 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Guard against excessive IRQ disabled times when we get
|
|
|
|
* a large list of pages to free.
|
|
|
|
*/
|
|
|
|
if (++batch_count == SWAP_CLUSTER_MAX) {
|
|
|
|
local_irq_restore(flags);
|
|
|
|
batch_count = 0;
|
|
|
|
local_irq_save(flags);
|
|
|
|
}
|
2012-01-10 23:07:04 +00:00
|
|
|
}
|
mm, page_alloc: enable/disable IRQs once when freeing a list of pages
Patch series "Follow-up for speed up page cache truncation", v2.
This series is a follow-on for Jan Kara's series "Speed up page cache
truncation" series. We both ended up looking at the same problem but
saw different problems based on the same data. This series builds upon
his work.
A variety of workloads were compared on four separate machines but each
machine showed gains albeit at different levels. Minimally, some of the
differences are due to NUMA where truncating data from a remote node is
slower than a local node. The workloads checked were
o sparse truncate microbenchmark, tiny
o sparse truncate microbenchmark, large
o reaim-io disk workfile
o dbench4 (modified by mmtests to produce more stable results)
o filebench varmail configuration for small memory size
o bonnie, directory operations, working set size 2*RAM
reaim-io, dbench and filebench all showed minor gains. Truncation does
not dominate those workloads but were tested to ensure no other
regressions. They will not be reported further.
The sparse truncate microbench was written by Jan. It creates a number
of files and then times how long it takes to truncate each one. The
"tiny" configuraiton creates a number of files that easily fits in
memory and times how long it takes to truncate files with page cache.
The large configuration uses enough files to have data that is twice the
size of memory and so timings there include truncating page cache and
working set shadow entries in the radix tree.
Patches 1-4 are the most relevant parts of this series. Patches 5-8 are
optional as they are deleting code that is essentially useless but has a
negligible performance impact.
The changelogs have more information on performance but just for bonnie
delete options, the main comparison is
bonnie
4.14.0-rc5 4.14.0-rc5 4.14.0-rc5
jan-v2 vanilla mel-v2
Hmean SeqCreate ops 76.20 ( 0.00%) 75.80 ( -0.53%) 76.80 ( 0.79%)
Hmean SeqCreate read 85.00 ( 0.00%) 85.00 ( 0.00%) 85.00 ( 0.00%)
Hmean SeqCreate del 13752.31 ( 0.00%) 12090.23 ( -12.09%) 15304.84 ( 11.29%)
Hmean RandCreate ops 76.00 ( 0.00%) 75.60 ( -0.53%) 77.00 ( 1.32%)
Hmean RandCreate read 96.80 ( 0.00%) 96.80 ( 0.00%) 97.00 ( 0.21%)
Hmean RandCreate del 13233.75 ( 0.00%) 11525.35 ( -12.91%) 14446.61 ( 9.16%)
Jan's series is the baseline and the vanilla kernel is 12% slower where
as this series on top gains another 11%. This is from a different
machine than the data in the changelogs but the detailed data was not
collected as there was no substantial change in v2.
This patch (of 8):
Freeing a list of pages current enables/disables IRQs for each page
freed. This patch splits freeing a list of pages into two operations --
preparing the pages for freeing and the actual freeing. This is a
tradeoff - we're taking two passes of the list to free in exchange for
avoiding multiple enable/disable of IRQs.
sparsetruncate (tiny)
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Min Time 149.00 ( 0.00%) 141.00 ( 5.37%)
1st-qrtle Time 150.00 ( 0.00%) 142.00 ( 5.33%)
2nd-qrtle Time 151.00 ( 0.00%) 142.00 ( 5.96%)
3rd-qrtle Time 151.00 ( 0.00%) 143.00 ( 5.30%)
Max-90% Time 153.00 ( 0.00%) 144.00 ( 5.88%)
Max-95% Time 155.00 ( 0.00%) 147.00 ( 5.16%)
Max-99% Time 201.00 ( 0.00%) 195.00 ( 2.99%)
Max Time 236.00 ( 0.00%) 230.00 ( 2.54%)
Amean Time 152.65 ( 0.00%) 144.37 ( 5.43%)
Stddev Time 9.78 ( 0.00%) 10.44 ( -6.72%)
Coeff Time 6.41 ( 0.00%) 7.23 ( -12.84%)
Best99%Amean Time 152.07 ( 0.00%) 143.72 ( 5.50%)
Best95%Amean Time 150.75 ( 0.00%) 142.37 ( 5.56%)
Best90%Amean Time 150.59 ( 0.00%) 142.19 ( 5.58%)
Best75%Amean Time 150.36 ( 0.00%) 141.92 ( 5.61%)
Best50%Amean Time 150.04 ( 0.00%) 141.69 ( 5.56%)
Best25%Amean Time 149.85 ( 0.00%) 141.38 ( 5.65%)
With a tiny number of files, each file truncated has resident page cache
and it shows that time to truncate is roughtly 5-6% with some minor
jitter.
4.14.0-rc4 4.14.0-rc4
janbatch-v1r1 oneirq-v1r1
Hmean SeqCreate ops 65.27 ( 0.00%) 81.86 ( 25.43%)
Hmean SeqCreate read 39.48 ( 0.00%) 47.44 ( 20.16%)
Hmean SeqCreate del 24963.95 ( 0.00%) 26319.99 ( 5.43%)
Hmean RandCreate ops 65.47 ( 0.00%) 82.01 ( 25.26%)
Hmean RandCreate read 42.04 ( 0.00%) 51.75 ( 23.09%)
Hmean RandCreate del 23377.66 ( 0.00%) 23764.79 ( 1.66%)
As expected, there is a small gain for the delete operation.
[mgorman@techsingularity.net: use page_private and set_page_private helpers]
Link: http://lkml.kernel.org/r/20171018101547.mjycw7zreb66jzpa@techsingularity.net
Link: http://lkml.kernel.org/r/20171018075952.10627-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:37:37 +00:00
|
|
|
local_irq_restore(flags);
|
2012-01-10 23:07:04 +00:00
|
|
|
}
|
|
|
|
|
2006-03-22 08:08:05 +00:00
|
|
|
/*
|
|
|
|
* split_page takes a non-compound higher-order page, and splits it into
|
|
|
|
* n (1<<order) sub-pages: page[0..n]
|
|
|
|
* Each sub-page must be freed individually.
|
|
|
|
*
|
|
|
|
* Note: this is probably too low level an operation for use in drivers.
|
|
|
|
* Please consult with lkml before using this in your driver.
|
|
|
|
*/
|
|
|
|
void split_page(struct page *page, unsigned int order)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2014-01-23 23:52:54 +00:00
|
|
|
VM_BUG_ON_PAGE(PageCompound(page), page);
|
|
|
|
VM_BUG_ON_PAGE(!page_count(page), page);
|
2008-11-25 15:55:53 +00:00
|
|
|
|
2016-07-26 22:23:49 +00:00
|
|
|
for (i = 1; i < (1 << order); i++)
|
2006-03-22 08:08:40 +00:00
|
|
|
set_page_refcounted(page + i);
|
2020-10-16 03:05:29 +00:00
|
|
|
split_page_owner(page, 1 << order);
|
2006-03-22 08:08:05 +00:00
|
|
|
}
|
2013-03-25 22:47:38 +00:00
|
|
|
EXPORT_SYMBOL_GPL(split_page);
|
2006-03-22 08:08:05 +00:00
|
|
|
|
2014-11-13 23:19:21 +00:00
|
|
|
int __isolate_free_page(struct page *page, unsigned int order)
|
2010-05-24 21:32:27 +00:00
|
|
|
{
|
|
|
|
unsigned long watermark;
|
|
|
|
struct zone *zone;
|
2012-10-08 23:32:00 +00:00
|
|
|
int mt;
|
2010-05-24 21:32:27 +00:00
|
|
|
|
|
|
|
BUG_ON(!PageBuddy(page));
|
|
|
|
|
|
|
|
zone = page_zone(page);
|
2012-12-12 00:02:57 +00:00
|
|
|
mt = get_pageblock_migratetype(page);
|
2010-05-24 21:32:27 +00:00
|
|
|
|
2013-02-23 00:33:58 +00:00
|
|
|
if (!is_migrate_isolate(mt)) {
|
2016-10-07 23:58:00 +00:00
|
|
|
/*
|
|
|
|
* Obey watermarks as if the page was being allocated. We can
|
|
|
|
* emulate a high-order watermark check with a raised order-0
|
|
|
|
* watermark, because we already know our high-order page
|
|
|
|
* exists.
|
|
|
|
*/
|
2019-03-05 23:44:50 +00:00
|
|
|
watermark = zone->_watermark[WMARK_MIN] + (1UL << order);
|
2018-05-23 01:18:21 +00:00
|
|
|
if (!zone_watermark_ok(zone, 0, watermark, 0, ALLOC_CMA))
|
2012-12-12 00:02:57 +00:00
|
|
|
return 0;
|
|
|
|
|
2013-01-11 22:32:16 +00:00
|
|
|
__mod_zone_freepage_state(zone, -(1UL << order), mt);
|
2012-12-12 00:02:57 +00:00
|
|
|
}
|
2010-05-24 21:32:27 +00:00
|
|
|
|
|
|
|
/* Remove page from free list */
|
2019-05-14 22:41:32 +00:00
|
|
|
|
2020-04-07 03:04:49 +00:00
|
|
|
del_page_from_free_list(page, zone, order);
|
2012-10-08 23:32:00 +00:00
|
|
|
|
2016-07-28 22:45:07 +00:00
|
|
|
/*
|
|
|
|
* Set the pageblock if the isolated page is at least half of a
|
|
|
|
* pageblock
|
|
|
|
*/
|
2010-05-24 21:32:27 +00:00
|
|
|
if (order >= pageblock_order - 1) {
|
|
|
|
struct page *endpage = page + (1 << order) - 1;
|
2011-12-29 12:09:50 +00:00
|
|
|
for (; page < endpage; page += pageblock_nr_pages) {
|
|
|
|
int mt = get_pageblock_migratetype(page);
|
mm: don't steal highatomic pageblock
Patch series "use up highorder free pages before OOM", v3.
I got OOM report from production team with v4.4 kernel. It had enough
free memory but failed to allocate GFP_KERNEL order-0 page and finally
encountered OOM kill. It occured during QA process which launches
several apps, switching and so on. It happned rarely. IOW, In normal
situation, it was not a problem but if we are unluck so that several
apps uses peak memory at the same time, it can happen. If we manage to
pass the phase, the system can go working well.
I could reproduce it with my test(memory spike easily. Look at below.
The reason is free pages(19M) of DMA32 zone are reserved for
HIGHORDERATOMIC and doesn't unreserved before the OOM.
balloon invoked oom-killer: gfp_mask=0x24280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), order=0, oom_score_adj=0
balloon cpuset=/ mems_allowed=0
CPU: 1 PID: 8473 Comm: balloon Tainted: G W OE 4.8.0-rc7-00219-g3f74c9559583-dirty #3161
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
dump_header+0x5c/0x1ce
oom_kill_process+0x22e/0x400
out_of_memory+0x1ac/0x210
__alloc_pages_nodemask+0x101e/0x1040
handle_mm_fault+0xa0a/0xbf0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:383949 inactive_anon:106724 isolated_anon:0
active_file:15 inactive_file:44 isolated_file:0
unevictable:0 dirty:0 writeback:24 unstable:0
slab_reclaimable:2483 slab_unreclaimable:3326
mapped:0 shmem:0 pagetables:1906 bounce:0
free:6898 free_pcp:291 free_cma:0
Node 0 active_anon:1535796kB inactive_anon:426896kB active_file:60kB inactive_file:176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:96kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1418 all_unreclaimable? no
DMA free:8188kB min:44kB low:56kB high:68kB active_anon:7648kB inactive_anon:0kB active_file:0kB inactive_file:4kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:20kB kernel_stack:0kB pagetables:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:19404kB min:5628kB low:7624kB high:9620kB active_anon:1528148kB inactive_anon:426896kB active_file:60kB inactive_file:420kB unevictable:0kB writepending:96kB present:2080640kB managed:2030092kB mlocked:0kB slab_reclaimable:9932kB slab_unreclaimable:13284kB kernel_stack:2496kB pagetables:7624kB bounce:0kB free_pcp:900kB local_pcp:112kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 0*4kB 0*8kB 0*16kB 0*32kB 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 2*4096kB (H) = 8192kB
DMA32: 7*4kB (H) 8*8kB (H) 30*16kB (H) 31*32kB (H) 14*64kB (H) 9*128kB (H) 2*256kB (H) 2*512kB (H) 4*1024kB (H) 5*2048kB (H) 0*4096kB = 19484kB
51131 total pagecache pages
50795 pages in swap cache
Swap cache stats: add 3532405601, delete 3532354806, find 124289150/1822712228
Free swap = 8kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12658 pages reserved
0 pages cma reserved
0 pages hwpoisoned
Another example exceeded the limit by the race is
in:imklog: page allocation failure: order:0, mode:0x2280020(GFP_ATOMIC|__GFP_NOTRACK)
CPU: 0 PID: 476 Comm: in:imklog Tainted: G E 4.8.0-rc7-00217-g266ef83c51e5-dirty #3135
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
warn_alloc_failed+0xdb/0x130
__alloc_pages_nodemask+0x4d6/0xdb0
new_slab+0x339/0x490
___slab_alloc.constprop.74+0x367/0x480
__slab_alloc.constprop.73+0x20/0x40
__kmalloc+0x1a4/0x1e0
alloc_indirect.isra.14+0x1d/0x50
virtqueue_add_sgs+0x1c4/0x470
__virtblk_add_req+0xae/0x1f0
virtio_queue_rq+0x12d/0x290
__blk_mq_run_hw_queue+0x239/0x370
blk_mq_run_hw_queue+0x8f/0xb0
blk_mq_insert_requests+0x18c/0x1a0
blk_mq_flush_plug_list+0x125/0x140
blk_flush_plug_list+0xc7/0x220
blk_finish_plug+0x2c/0x40
__do_page_cache_readahead+0x196/0x230
filemap_fault+0x448/0x4f0
ext4_filemap_fault+0x36/0x50
__do_fault+0x75/0x140
handle_mm_fault+0x84d/0xbe0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:363826 inactive_anon:121283 isolated_anon:32
active_file:65 inactive_file:152 isolated_file:0
unevictable:0 dirty:0 writeback:46 unstable:0
slab_reclaimable:2778 slab_unreclaimable:3070
mapped:112 shmem:0 pagetables:1822 bounce:0
free:9469 free_pcp:231 free_cma:0
Node 0 active_anon:1455304kB inactive_anon:485132kB active_file:260kB inactive_file:608kB unevictable:0kB isolated(anon):128kB isolated(file):0kB mapped:448kB dirty:0kB writeback:184kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:13641 all_unreclaimable? no
DMA free:7748kB min:44kB low:56kB high:68kB active_anon:7944kB inactive_anon:104kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:108kB kernel_stack:0kB pagetables:4kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:30128kB min:5628kB low:7624kB high:9620kB active_anon:1447360kB inactive_anon:485028kB active_file:260kB inactive_file:608kB unevictable:0kB writepending:184kB present:2080640kB managed:2030132kB mlocked:0kB slab_reclaimable:11112kB slab_unreclaimable:12172kB kernel_stack:2400kB pagetables:7284kB bounce:0kB free_pcp:924kB local_pcp:72kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 7*4kB (UE) 3*8kB (UH) 1*16kB (M) 0*32kB 2*64kB (U) 1*128kB (M) 1*256kB (U) 0*512kB 1*1024kB (U) 1*2048kB (U) 1*4096kB (H) = 7748kB
DMA32: 10*4kB (H) 3*8kB (H) 47*16kB (H) 38*32kB (H) 5*64kB (H) 1*128kB (H) 2*256kB (H) 3*512kB (H) 3*1024kB (H) 3*2048kB (H) 4*4096kB (H) = 30128kB
2775 total pagecache pages
2536 pages in swap cache
Swap cache stats: add 206786828, delete 206784292, find 7323106/106686077
Free swap = 108744kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12648 pages reserved
0 pages cma reserved
0 pages hwpoisoned
During the investigation, I found some problems with highatomic so this
patch aims to solve the problems and the final goal is to unreserve
every highatomic free pages before the OOM kill.
This patch (of 4):
In page freeing path, migratetype is racy so that a highorderatomic page
could free into non-highorderatomic free list. If that page is
allocated, VM can change the pageblock from higorderatomic to something.
In that case, highatomic pageblock accounting is broken so it doesn't
work(e.g., VM cannot reserve highorderatomic pageblocks any more
although it doesn't reach 1% limit).
So, this patch prohibits the changing from highatomic to other type.
It's no problem because MIGRATE_HIGHATOMIC is not listed in fallback
array so stealing will only happen due to unexpected races which is
really rare. Also, such prohibiting keeps highatomic pageblock more
longer so it would be better for highorderatomic page allocation.
Link: http://lkml.kernel.org/r/1476259429-18279-2-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Sangseok Lee <sangseok.lee@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-12-13 00:42:05 +00:00
|
|
|
if (!is_migrate_isolate(mt) && !is_migrate_cma(mt)
|
2017-05-03 21:52:52 +00:00
|
|
|
&& !is_migrate_highatomic(mt))
|
2011-12-29 12:09:50 +00:00
|
|
|
set_pageblock_migratetype(page,
|
|
|
|
MIGRATE_MOVABLE);
|
|
|
|
}
|
2010-05-24 21:32:27 +00:00
|
|
|
}
|
|
|
|
|
2015-07-17 23:24:15 +00:00
|
|
|
|
2013-01-11 22:32:16 +00:00
|
|
|
return 1UL << order;
|
2012-10-08 23:29:12 +00:00
|
|
|
}
|
|
|
|
|
2020-04-07 03:04:53 +00:00
|
|
|
/**
|
|
|
|
* __putback_isolated_page - Return a now-isolated page back where we got it
|
|
|
|
* @page: Page that was isolated
|
|
|
|
* @order: Order of the isolated page
|
2020-04-10 21:32:29 +00:00
|
|
|
* @mt: The page's pageblock's migratetype
|
2020-04-07 03:04:53 +00:00
|
|
|
*
|
|
|
|
* This function is meant to return a page pulled from the free lists via
|
|
|
|
* __isolate_free_page back to the free lists they were pulled from.
|
|
|
|
*/
|
|
|
|
void __putback_isolated_page(struct page *page, unsigned int order, int mt)
|
|
|
|
{
|
|
|
|
struct zone *zone = page_zone(page);
|
|
|
|
|
|
|
|
/* zone lock should be held when this function is called */
|
|
|
|
lockdep_assert_held(&zone->lock);
|
|
|
|
|
|
|
|
/* Return isolated page to tail of freelist. */
|
mm/page_alloc: convert "report" flag of __free_one_page() to a proper flag
Patch series "mm: place pages to the freelist tail when onlining and undoing isolation", v2.
When adding separate memory blocks via add_memory*() and onlining them
immediately, the metadata (especially the memmap) of the next block will
be placed onto one of the just added+onlined block. This creates a chain
of unmovable allocations: If the last memory block cannot get
offlined+removed() so will all dependent ones. We directly have unmovable
allocations all over the place.
This can be observed quite easily using virtio-mem, however, it can also
be observed when using DIMMs. The freshly onlined pages will usually be
placed to the head of the freelists, meaning they will be allocated next,
turning the just-added memory usually immediately un-removable. The fresh
pages are cold, prefering to allocate others (that might be hot) also
feels to be the natural thing to do.
It also applies to the hyper-v balloon xen-balloon, and ppc64 dlpar: when
adding separate, successive memory blocks, each memory block will have
unmovable allocations on them - for example gigantic pages will fail to
allocate.
While the ZONE_NORMAL doesn't provide any guarantees that memory can get
offlined+removed again (any kind of fragmentation with unmovable
allocations is possible), there are many scenarios (hotplugging a lot of
memory, running workload, hotunplug some memory/as much as possible) where
we can offline+remove quite a lot with this patchset.
a) To visualize the problem, a very simple example:
Start a VM with 4GB and 8GB of virtio-mem memory:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000033fffffff 9G online yes 32-103
Memory block size: 128M
Total online memory: 12G
Total offline memory: 0B
Then try to unplug as much as possible using virtio-mem. Observe which
memory blocks are still around. Without this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
0x0000000148000000-0x000000014fffffff 128M online yes 41
0x0000000158000000-0x000000015fffffff 128M online yes 43
0x0000000168000000-0x000000016fffffff 128M online yes 45
0x0000000178000000-0x000000017fffffff 128M online yes 47
0x0000000188000000-0x0000000197ffffff 256M online yes 49-50
0x00000001a0000000-0x00000001a7ffffff 128M online yes 52
0x00000001b0000000-0x00000001b7ffffff 128M online yes 54
0x00000001c0000000-0x00000001c7ffffff 128M online yes 56
0x00000001d0000000-0x00000001d7ffffff 128M online yes 58
0x00000001e0000000-0x00000001e7ffffff 128M online yes 60
0x00000001f0000000-0x00000001f7ffffff 128M online yes 62
0x0000000200000000-0x0000000207ffffff 128M online yes 64
0x0000000210000000-0x0000000217ffffff 128M online yes 66
0x0000000220000000-0x0000000227ffffff 128M online yes 68
0x0000000230000000-0x0000000237ffffff 128M online yes 70
0x0000000240000000-0x0000000247ffffff 128M online yes 72
0x0000000250000000-0x0000000257ffffff 128M online yes 74
0x0000000260000000-0x0000000267ffffff 128M online yes 76
0x0000000270000000-0x0000000277ffffff 128M online yes 78
0x0000000280000000-0x0000000287ffffff 128M online yes 80
0x0000000290000000-0x0000000297ffffff 128M online yes 82
0x00000002a0000000-0x00000002a7ffffff 128M online yes 84
0x00000002b0000000-0x00000002b7ffffff 128M online yes 86
0x00000002c0000000-0x00000002c7ffffff 128M online yes 88
0x00000002d0000000-0x00000002d7ffffff 128M online yes 90
0x00000002e0000000-0x00000002e7ffffff 128M online yes 92
0x00000002f0000000-0x00000002f7ffffff 128M online yes 94
0x0000000300000000-0x0000000307ffffff 128M online yes 96
0x0000000310000000-0x0000000317ffffff 128M online yes 98
0x0000000320000000-0x0000000327ffffff 128M online yes 100
0x0000000330000000-0x000000033fffffff 256M online yes 102-103
Memory block size: 128M
Total online memory: 8.1G
Total offline memory: 0B
With this patch set:
[root@localhost ~]# lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x00000000bfffffff 3G online yes 0-23
0x0000000100000000-0x000000013fffffff 1G online yes 32-39
Memory block size: 128M
Total online memory: 4G
Total offline memory: 0B
All memory can get unplugged, all memory block can get removed. Of
course, no workload ran and the system was basically idle, but it
highlights the issue - the fairly deterministic chain of unmovable
allocations. When a huge page for the 2MB memmap is needed, a
just-onlined 4MB page will be split. The remaining 2MB page will be used
for the memmap of the next memory block. So one memory block will hold
the memmap of the two following memory blocks. Finally the pages of the
last-onlined memory block will get used for the next bigger allocations -
if any allocation is unmovable, all dependent memory blocks cannot get
unplugged and removed until that allocation is gone.
Note that with bigger memory blocks (e.g., 256MB), *all* memory
blocks are dependent and none can get unplugged again!
b) Experiment with memory intensive workload
I performed an experiment with an older version of this patch set (before
we used undo_isolate_page_range() in online_pages(): Hotplug 56GB to a VM
with an initial 4GB, onlining all memory to ZONE_NORMAL right from the
kernel when adding it. I then run various memory intensive workloads that
consume most system memory for a total of 45 minutes. Once finished, I
try to unplug as much memory as possible.
With this change, I am able to remove via virtio-mem (adding individual
128MB memory blocks) 413 out of 448 added memory blocks. Via individual
(256MB) DIMMs 380 out of 448 added memory blocks. (I don't have any
numbers without this patchset, but looking at the above example, it's at
most half of the 448 memory blocks for virtio-mem, and most probably none
for DIMMs).
Again, there are workloads that might behave very differently due to the
nature of ZONE_NORMAL.
This change also affects (besides memory onlining):
- Other users of undo_isolate_page_range(): Pages are always placed to the
tail.
-- When memory offlining fails
-- When memory isolation fails after having isolated some pageblocks
-- When alloc_contig_range() either succeeds or fails
- Other users of __putback_isolated_page(): Pages are always placed to the
tail.
-- Free page reporting
- Other users of __free_pages_core()
-- AFAIKs, any memory that is getting exposed to the buddy during boot.
IIUC we will now usually allocate memory from lower addresses within
a zone first (especially during boot).
- Other users of generic_online_page()
-- Hyper-V balloon
This patch (of 5):
Let's prepare for additional flags and avoid long parameter lists of
bools. Follow-up patches will also make use of the flags in
__free_pages_ok().
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-1-david@redhat.com
Link: https://lkml.kernel.org/r/20201005121534.15649-2-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:20 +00:00
|
|
|
__free_one_page(page, page_to_pfn(page), zone, order, mt,
|
mm/page_alloc: place pages to tail in __putback_isolated_page()
__putback_isolated_page() already documents that pages will be placed to
the tail of the freelist - this is, however, not the case for "order >=
MAX_ORDER - 2" (see buddy_merge_likely()) - which should be the case for
all existing users.
This change affects two users:
- free page reporting
- page isolation, when undoing the isolation (including memory onlining).
This behavior is desirable for pages that haven't really been touched
lately, so exactly the two users that don't actually read/write page
content, but rather move untouched pages.
The new behavior is especially desirable for memory onlining, where we
allow allocation of newly onlined pages via undo_isolate_page_range() in
online_pages(). Right now, we always place them to the head of the
freelist, resulting in undesireable behavior: Assume we add individual
memory chunks via add_memory() and online them right away to the NORMAL
zone. We create a dependency chain of unmovable allocations e.g., via the
memmap. The memmap of the next chunk will be placed onto previous chunks
- if the last block cannot get offlined+removed, all dependent ones cannot
get offlined+removed. While this can already be observed with individual
DIMMs, it's more of an issue for virtio-mem (and I suspect also ppc
DLPAR).
Document that this should only be used for optimizations, and no code
should rely on this behavior for correction (if the order of the freelists
ever changes).
We won't care about page shuffling: memory onlining already properly
shuffles after onlining. free page reporting doesn't care about
physically contiguous ranges, and there are already cases where page
isolation will simply move (physically close) free pages to (currently)
the head of the freelists via move_freepages_block() instead of shuffling.
If this becomes ever relevant, we should shuffle the whole zone when
undoing isolation of larger ranges, and after free_contig_range().
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@linux.alibaba.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Scott Cheloha <cheloha@linux.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Link: https://lkml.kernel.org/r/20201005121534.15649-3-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:09:26 +00:00
|
|
|
FPI_SKIP_REPORT_NOTIFY | FPI_TO_TAIL);
|
2020-04-07 03:04:53 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:13:27 +00:00
|
|
|
/*
|
|
|
|
* Update NUMA hit/miss statistics
|
|
|
|
*
|
|
|
|
* Must be called with interrupts disabled.
|
|
|
|
*/
|
2017-01-11 00:57:42 +00:00
|
|
|
static inline void zone_statistics(struct zone *preferred_zone, struct zone *z)
|
2016-05-20 00:13:27 +00:00
|
|
|
{
|
|
|
|
#ifdef CONFIG_NUMA
|
mm: change the call sites of numa statistics items
Patch series "Separate NUMA statistics from zone statistics", v2.
Each page allocation updates a set of per-zone statistics with a call to
zone_statistics(). As discussed in 2017 MM summit, these are a
substantial source of overhead in the page allocator and are very rarely
consumed. This significant overhead in cache bouncing caused by zone
counters (NUMA associated counters) update in parallel in multi-threaded
page allocation (pointed out by Dave Hansen).
A link to the MM summit slides:
http://people.netfilter.org/hawk/presentations/MM-summit2017/MM-summit2017-JesperBrouer.pdf
To mitigate this overhead, this patchset separates NUMA statistics from
zone statistics framework, and update NUMA counter threshold to a fixed
size of MAX_U16 - 2, as a small threshold greatly increases the update
frequency of the global counter from local per cpu counter (suggested by
Ying Huang). The rationality is that these statistics counters don't
need to be read often, unlike other VM counters, so it's not a problem
to use a large threshold and make readers more expensive.
With this patchset, we see 31.3% drop of CPU cycles(537-->369, see
below) for per single page allocation and reclaim on Jesper's
page_bench03 benchmark. Meanwhile, this patchset keeps the same style
of virtual memory statistics with little end-user-visible effects (only
move the numa stats to show behind zone page stats, see the first patch
for details).
I did an experiment of single page allocation and reclaim concurrently
using Jesper's page_bench03 benchmark on a 2-Socket Broadwell-based
server (88 processors with 126G memory) with different size of threshold
of pcp counter.
Benchmark provided by Jesper D Brouer(increase loop times to 10000000):
https://github.com/netoptimizer/prototype-kernel/tree/master/kernel/mm/bench
Threshold CPU cycles Throughput(88 threads)
32 799 241760478
64 640 301628829
125 537 358906028 <==> system by default
256 468 412397590
512 428 450550704
4096 399 482520943
20000 394 489009617
30000 395 488017817
65533 369(-31.3%) 521661345(+45.3%) <==> with this patchset
N/A 342(-36.3%) 562900157(+56.8%) <==> disable zone_statistics
This patch (of 3):
In this patch, NUMA statistics is separated from zone statistics
framework, all the call sites of NUMA stats are changed to use
numa-stats-specific functions, it does not have any functionality change
except that the number of NUMA stats is shown behind zone page stats
when users *read* the zone info.
E.g. cat /proc/zoneinfo
***Base*** ***With this patch***
nr_free_pages 3976 nr_free_pages 3976
nr_zone_inactive_anon 0 nr_zone_inactive_anon 0
nr_zone_active_anon 0 nr_zone_active_anon 0
nr_zone_inactive_file 0 nr_zone_inactive_file 0
nr_zone_active_file 0 nr_zone_active_file 0
nr_zone_unevictable 0 nr_zone_unevictable 0
nr_zone_write_pending 0 nr_zone_write_pending 0
nr_mlock 0 nr_mlock 0
nr_page_table_pages 0 nr_page_table_pages 0
nr_kernel_stack 0 nr_kernel_stack 0
nr_bounce 0 nr_bounce 0
nr_zspages 0 nr_zspages 0
numa_hit 0 *nr_free_cma 0*
numa_miss 0 numa_hit 0
numa_foreign 0 numa_miss 0
numa_interleave 0 numa_foreign 0
numa_local 0 numa_interleave 0
numa_other 0 numa_local 0
*nr_free_cma 0* numa_other 0
... ...
vm stats threshold: 10 vm stats threshold: 10
... ...
The next patch updates the numa stats counter size and threshold.
[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/1503568801-21305-2-git-send-email-kemi.wang@intel.com
Signed-off-by: Kemi Wang <kemi.wang@intel.com>
Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christopher Lameter <cl@linux.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Ying Huang <ying.huang@intel.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Tim Chen <tim.c.chen@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:12:48 +00:00
|
|
|
enum numa_stat_item local_stat = NUMA_LOCAL;
|
2016-05-20 00:13:27 +00:00
|
|
|
|
2017-11-16 01:38:22 +00:00
|
|
|
/* skip numa counters update if numa stats is disabled */
|
|
|
|
if (!static_branch_likely(&vm_numa_stat_key))
|
|
|
|
return;
|
|
|
|
|
2018-08-22 04:53:32 +00:00
|
|
|
if (zone_to_nid(z) != numa_node_id())
|
2016-05-20 00:13:27 +00:00
|
|
|
local_stat = NUMA_OTHER;
|
|
|
|
|
2018-08-22 04:53:32 +00:00
|
|
|
if (zone_to_nid(z) == zone_to_nid(preferred_zone))
|
mm: change the call sites of numa statistics items
Patch series "Separate NUMA statistics from zone statistics", v2.
Each page allocation updates a set of per-zone statistics with a call to
zone_statistics(). As discussed in 2017 MM summit, these are a
substantial source of overhead in the page allocator and are very rarely
consumed. This significant overhead in cache bouncing caused by zone
counters (NUMA associated counters) update in parallel in multi-threaded
page allocation (pointed out by Dave Hansen).
A link to the MM summit slides:
http://people.netfilter.org/hawk/presentations/MM-summit2017/MM-summit2017-JesperBrouer.pdf
To mitigate this overhead, this patchset separates NUMA statistics from
zone statistics framework, and update NUMA counter threshold to a fixed
size of MAX_U16 - 2, as a small threshold greatly increases the update
frequency of the global counter from local per cpu counter (suggested by
Ying Huang). The rationality is that these statistics counters don't
need to be read often, unlike other VM counters, so it's not a problem
to use a large threshold and make readers more expensive.
With this patchset, we see 31.3% drop of CPU cycles(537-->369, see
below) for per single page allocation and reclaim on Jesper's
page_bench03 benchmark. Meanwhile, this patchset keeps the same style
of virtual memory statistics with little end-user-visible effects (only
move the numa stats to show behind zone page stats, see the first patch
for details).
I did an experiment of single page allocation and reclaim concurrently
using Jesper's page_bench03 benchmark on a 2-Socket Broadwell-based
server (88 processors with 126G memory) with different size of threshold
of pcp counter.
Benchmark provided by Jesper D Brouer(increase loop times to 10000000):
https://github.com/netoptimizer/prototype-kernel/tree/master/kernel/mm/bench
Threshold CPU cycles Throughput(88 threads)
32 799 241760478
64 640 301628829
125 537 358906028 <==> system by default
256 468 412397590
512 428 450550704
4096 399 482520943
20000 394 489009617
30000 395 488017817
65533 369(-31.3%) 521661345(+45.3%) <==> with this patchset
N/A 342(-36.3%) 562900157(+56.8%) <==> disable zone_statistics
This patch (of 3):
In this patch, NUMA statistics is separated from zone statistics
framework, all the call sites of NUMA stats are changed to use
numa-stats-specific functions, it does not have any functionality change
except that the number of NUMA stats is shown behind zone page stats
when users *read* the zone info.
E.g. cat /proc/zoneinfo
***Base*** ***With this patch***
nr_free_pages 3976 nr_free_pages 3976
nr_zone_inactive_anon 0 nr_zone_inactive_anon 0
nr_zone_active_anon 0 nr_zone_active_anon 0
nr_zone_inactive_file 0 nr_zone_inactive_file 0
nr_zone_active_file 0 nr_zone_active_file 0
nr_zone_unevictable 0 nr_zone_unevictable 0
nr_zone_write_pending 0 nr_zone_write_pending 0
nr_mlock 0 nr_mlock 0
nr_page_table_pages 0 nr_page_table_pages 0
nr_kernel_stack 0 nr_kernel_stack 0
nr_bounce 0 nr_bounce 0
nr_zspages 0 nr_zspages 0
numa_hit 0 *nr_free_cma 0*
numa_miss 0 numa_hit 0
numa_foreign 0 numa_miss 0
numa_interleave 0 numa_foreign 0
numa_local 0 numa_interleave 0
numa_other 0 numa_local 0
*nr_free_cma 0* numa_other 0
... ...
vm stats threshold: 10 vm stats threshold: 10
... ...
The next patch updates the numa stats counter size and threshold.
[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/1503568801-21305-2-git-send-email-kemi.wang@intel.com
Signed-off-by: Kemi Wang <kemi.wang@intel.com>
Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christopher Lameter <cl@linux.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Ying Huang <ying.huang@intel.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Tim Chen <tim.c.chen@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:12:48 +00:00
|
|
|
__inc_numa_state(z, NUMA_HIT);
|
2017-01-11 00:57:39 +00:00
|
|
|
else {
|
mm: change the call sites of numa statistics items
Patch series "Separate NUMA statistics from zone statistics", v2.
Each page allocation updates a set of per-zone statistics with a call to
zone_statistics(). As discussed in 2017 MM summit, these are a
substantial source of overhead in the page allocator and are very rarely
consumed. This significant overhead in cache bouncing caused by zone
counters (NUMA associated counters) update in parallel in multi-threaded
page allocation (pointed out by Dave Hansen).
A link to the MM summit slides:
http://people.netfilter.org/hawk/presentations/MM-summit2017/MM-summit2017-JesperBrouer.pdf
To mitigate this overhead, this patchset separates NUMA statistics from
zone statistics framework, and update NUMA counter threshold to a fixed
size of MAX_U16 - 2, as a small threshold greatly increases the update
frequency of the global counter from local per cpu counter (suggested by
Ying Huang). The rationality is that these statistics counters don't
need to be read often, unlike other VM counters, so it's not a problem
to use a large threshold and make readers more expensive.
With this patchset, we see 31.3% drop of CPU cycles(537-->369, see
below) for per single page allocation and reclaim on Jesper's
page_bench03 benchmark. Meanwhile, this patchset keeps the same style
of virtual memory statistics with little end-user-visible effects (only
move the numa stats to show behind zone page stats, see the first patch
for details).
I did an experiment of single page allocation and reclaim concurrently
using Jesper's page_bench03 benchmark on a 2-Socket Broadwell-based
server (88 processors with 126G memory) with different size of threshold
of pcp counter.
Benchmark provided by Jesper D Brouer(increase loop times to 10000000):
https://github.com/netoptimizer/prototype-kernel/tree/master/kernel/mm/bench
Threshold CPU cycles Throughput(88 threads)
32 799 241760478
64 640 301628829
125 537 358906028 <==> system by default
256 468 412397590
512 428 450550704
4096 399 482520943
20000 394 489009617
30000 395 488017817
65533 369(-31.3%) 521661345(+45.3%) <==> with this patchset
N/A 342(-36.3%) 562900157(+56.8%) <==> disable zone_statistics
This patch (of 3):
In this patch, NUMA statistics is separated from zone statistics
framework, all the call sites of NUMA stats are changed to use
numa-stats-specific functions, it does not have any functionality change
except that the number of NUMA stats is shown behind zone page stats
when users *read* the zone info.
E.g. cat /proc/zoneinfo
***Base*** ***With this patch***
nr_free_pages 3976 nr_free_pages 3976
nr_zone_inactive_anon 0 nr_zone_inactive_anon 0
nr_zone_active_anon 0 nr_zone_active_anon 0
nr_zone_inactive_file 0 nr_zone_inactive_file 0
nr_zone_active_file 0 nr_zone_active_file 0
nr_zone_unevictable 0 nr_zone_unevictable 0
nr_zone_write_pending 0 nr_zone_write_pending 0
nr_mlock 0 nr_mlock 0
nr_page_table_pages 0 nr_page_table_pages 0
nr_kernel_stack 0 nr_kernel_stack 0
nr_bounce 0 nr_bounce 0
nr_zspages 0 nr_zspages 0
numa_hit 0 *nr_free_cma 0*
numa_miss 0 numa_hit 0
numa_foreign 0 numa_miss 0
numa_interleave 0 numa_foreign 0
numa_local 0 numa_interleave 0
numa_other 0 numa_local 0
*nr_free_cma 0* numa_other 0
... ...
vm stats threshold: 10 vm stats threshold: 10
... ...
The next patch updates the numa stats counter size and threshold.
[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/1503568801-21305-2-git-send-email-kemi.wang@intel.com
Signed-off-by: Kemi Wang <kemi.wang@intel.com>
Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christopher Lameter <cl@linux.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Ying Huang <ying.huang@intel.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Tim Chen <tim.c.chen@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:12:48 +00:00
|
|
|
__inc_numa_state(z, NUMA_MISS);
|
|
|
|
__inc_numa_state(preferred_zone, NUMA_FOREIGN);
|
2016-05-20 00:13:27 +00:00
|
|
|
}
|
mm: change the call sites of numa statistics items
Patch series "Separate NUMA statistics from zone statistics", v2.
Each page allocation updates a set of per-zone statistics with a call to
zone_statistics(). As discussed in 2017 MM summit, these are a
substantial source of overhead in the page allocator and are very rarely
consumed. This significant overhead in cache bouncing caused by zone
counters (NUMA associated counters) update in parallel in multi-threaded
page allocation (pointed out by Dave Hansen).
A link to the MM summit slides:
http://people.netfilter.org/hawk/presentations/MM-summit2017/MM-summit2017-JesperBrouer.pdf
To mitigate this overhead, this patchset separates NUMA statistics from
zone statistics framework, and update NUMA counter threshold to a fixed
size of MAX_U16 - 2, as a small threshold greatly increases the update
frequency of the global counter from local per cpu counter (suggested by
Ying Huang). The rationality is that these statistics counters don't
need to be read often, unlike other VM counters, so it's not a problem
to use a large threshold and make readers more expensive.
With this patchset, we see 31.3% drop of CPU cycles(537-->369, see
below) for per single page allocation and reclaim on Jesper's
page_bench03 benchmark. Meanwhile, this patchset keeps the same style
of virtual memory statistics with little end-user-visible effects (only
move the numa stats to show behind zone page stats, see the first patch
for details).
I did an experiment of single page allocation and reclaim concurrently
using Jesper's page_bench03 benchmark on a 2-Socket Broadwell-based
server (88 processors with 126G memory) with different size of threshold
of pcp counter.
Benchmark provided by Jesper D Brouer(increase loop times to 10000000):
https://github.com/netoptimizer/prototype-kernel/tree/master/kernel/mm/bench
Threshold CPU cycles Throughput(88 threads)
32 799 241760478
64 640 301628829
125 537 358906028 <==> system by default
256 468 412397590
512 428 450550704
4096 399 482520943
20000 394 489009617
30000 395 488017817
65533 369(-31.3%) 521661345(+45.3%) <==> with this patchset
N/A 342(-36.3%) 562900157(+56.8%) <==> disable zone_statistics
This patch (of 3):
In this patch, NUMA statistics is separated from zone statistics
framework, all the call sites of NUMA stats are changed to use
numa-stats-specific functions, it does not have any functionality change
except that the number of NUMA stats is shown behind zone page stats
when users *read* the zone info.
E.g. cat /proc/zoneinfo
***Base*** ***With this patch***
nr_free_pages 3976 nr_free_pages 3976
nr_zone_inactive_anon 0 nr_zone_inactive_anon 0
nr_zone_active_anon 0 nr_zone_active_anon 0
nr_zone_inactive_file 0 nr_zone_inactive_file 0
nr_zone_active_file 0 nr_zone_active_file 0
nr_zone_unevictable 0 nr_zone_unevictable 0
nr_zone_write_pending 0 nr_zone_write_pending 0
nr_mlock 0 nr_mlock 0
nr_page_table_pages 0 nr_page_table_pages 0
nr_kernel_stack 0 nr_kernel_stack 0
nr_bounce 0 nr_bounce 0
nr_zspages 0 nr_zspages 0
numa_hit 0 *nr_free_cma 0*
numa_miss 0 numa_hit 0
numa_foreign 0 numa_miss 0
numa_interleave 0 numa_foreign 0
numa_local 0 numa_interleave 0
numa_other 0 numa_local 0
*nr_free_cma 0* numa_other 0
... ...
vm stats threshold: 10 vm stats threshold: 10
... ...
The next patch updates the numa stats counter size and threshold.
[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/1503568801-21305-2-git-send-email-kemi.wang@intel.com
Signed-off-by: Kemi Wang <kemi.wang@intel.com>
Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christopher Lameter <cl@linux.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Ying Huang <ying.huang@intel.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Tim Chen <tim.c.chen@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:12:48 +00:00
|
|
|
__inc_numa_state(z, local_stat);
|
2016-05-20 00:13:27 +00:00
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2017-02-24 22:56:26 +00:00
|
|
|
/* Remove page from the per-cpu list, caller must protect the list */
|
|
|
|
static struct page *__rmqueue_pcplist(struct zone *zone, int migratetype,
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
unsigned int alloc_flags,
|
2017-11-16 01:38:03 +00:00
|
|
|
struct per_cpu_pages *pcp,
|
2017-02-24 22:56:26 +00:00
|
|
|
struct list_head *list)
|
|
|
|
{
|
|
|
|
struct page *page;
|
|
|
|
|
|
|
|
do {
|
|
|
|
if (list_empty(list)) {
|
|
|
|
pcp->count += rmqueue_bulk(zone, 0,
|
2020-12-15 03:10:50 +00:00
|
|
|
READ_ONCE(pcp->batch), list,
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
migratetype, alloc_flags);
|
2017-02-24 22:56:26 +00:00
|
|
|
if (unlikely(list_empty(list)))
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2017-11-16 01:38:03 +00:00
|
|
|
page = list_first_entry(list, struct page, lru);
|
2017-02-24 22:56:26 +00:00
|
|
|
list_del(&page->lru);
|
|
|
|
pcp->count--;
|
|
|
|
} while (check_new_pcp(page));
|
|
|
|
|
|
|
|
return page;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Lock and remove page from the per-cpu list */
|
|
|
|
static struct page *rmqueue_pcplist(struct zone *preferred_zone,
|
2019-05-14 00:22:40 +00:00
|
|
|
struct zone *zone, gfp_t gfp_flags,
|
|
|
|
int migratetype, unsigned int alloc_flags)
|
2017-02-24 22:56:26 +00:00
|
|
|
{
|
|
|
|
struct per_cpu_pages *pcp;
|
|
|
|
struct list_head *list;
|
|
|
|
struct page *page;
|
2017-04-20 21:37:43 +00:00
|
|
|
unsigned long flags;
|
2017-02-24 22:56:26 +00:00
|
|
|
|
2017-04-20 21:37:43 +00:00
|
|
|
local_irq_save(flags);
|
2017-02-24 22:56:26 +00:00
|
|
|
pcp = &this_cpu_ptr(zone->pageset)->pcp;
|
|
|
|
list = &pcp->lists[migratetype];
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
page = __rmqueue_pcplist(zone, migratetype, alloc_flags, pcp, list);
|
2017-02-24 22:56:26 +00:00
|
|
|
if (page) {
|
2019-05-14 00:22:40 +00:00
|
|
|
__count_zid_vm_events(PGALLOC, page_zonenum(page), 1);
|
2017-02-24 22:56:26 +00:00
|
|
|
zone_statistics(preferred_zone, zone);
|
|
|
|
}
|
2017-04-20 21:37:43 +00:00
|
|
|
local_irq_restore(flags);
|
2017-02-24 22:56:26 +00:00
|
|
|
return page;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
mm: set page->pfmemalloc in prep_new_page()
The possibility of replacing the numerous parameters of alloc_pages*
functions with a single structure has been discussed when Minchan proposed
to expand the x86 kernel stack [1]. This series implements the change,
along with few more cleanups/microoptimizations.
The series is based on next-20150108 and I used gcc 4.8.3 20140627 on
openSUSE 13.2 for compiling. Config includess NUMA and COMPACTION.
The core change is the introduction of a new struct alloc_context, which looks
like this:
struct alloc_context {
struct zonelist *zonelist;
nodemask_t *nodemask;
struct zone *preferred_zone;
int classzone_idx;
int migratetype;
enum zone_type high_zoneidx;
};
All the contents is mostly constant, except that __alloc_pages_slowpath()
changes preferred_zone, classzone_idx and potentially zonelist. But
that's not a problem in case control returns to retry_cpuset: in
__alloc_pages_nodemask(), those will be reset to initial values again
(although it's a bit subtle). On the other hand, gfp_flags and alloc_info
mutate so much that it doesn't make sense to put them into alloc_context.
Still, the result is one parameter instead of up to 7. This is all in
Patch 2.
Patch 3 is a step to expand alloc_context usage out of page_alloc.c
itself. The function try_to_compact_pages() can also much benefit from
the parameter reduction, but it means the struct definition has to be
moved to a shared header.
Patch 1 should IMHO be included even if the rest is deemed not useful
enough. It improves maintainability and also has some code/stack
reduction. Patch 4 is OTOH a tiny optimization.
Overall bloat-o-meter results:
add/remove: 0/0 grow/shrink: 0/4 up/down: 0/-460 (-460)
function old new delta
nr_free_zone_pages 129 115 -14
__alloc_pages_direct_compact 329 256 -73
get_page_from_freelist 2670 2576 -94
__alloc_pages_nodemask 2564 2285 -279
try_to_compact_pages 582 579 -3
Overall stack sizes per ./scripts/checkstack.pl:
old new delta
get_page_from_freelist: 184 184 0
__alloc_pages_nodemask 248 200 -48
__alloc_pages_direct_c 40 - -40
try_to_compact_pages 72 72 0
-88
[1] http://marc.info/?l=linux-mm&m=140142462528257&w=2
This patch (of 4):
prep_new_page() sets almost everything in the struct page of the page
being allocated, except page->pfmemalloc. This is not obvious and has at
least once led to a bug where page->pfmemalloc was forgotten to be set
correctly, see commit 8fb74b9fb2b1 ("mm: compaction: partially revert
capture of suitable high-order page").
This patch moves the pfmemalloc setting to prep_new_page(), which means it
needs to gain alloc_flags parameter. The call to prep_new_page is moved
from buffered_rmqueue() to get_page_from_freelist(), which also leads to
simpler code. An obsolete comment for buffered_rmqueue() is replaced.
In addition to better maintainability there is a small reduction of code
and stack usage for get_page_from_freelist(), which inlines the other
functions involved.
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-145 (-145)
function old new delta
get_page_from_freelist 2670 2525 -145
Stack usage is reduced from 184 to 168 bytes.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:25:38 +00:00
|
|
|
* Allocate a page from the given zone. Use pcplists for order-0 allocations.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2009-06-16 22:32:05 +00:00
|
|
|
static inline
|
2017-02-24 22:56:26 +00:00
|
|
|
struct page *rmqueue(struct zone *preferred_zone,
|
2014-06-04 23:10:21 +00:00
|
|
|
struct zone *zone, unsigned int order,
|
2016-05-20 00:13:38 +00:00
|
|
|
gfp_t gfp_flags, unsigned int alloc_flags,
|
|
|
|
int migratetype)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
unsigned long flags;
|
2005-11-22 05:32:20 +00:00
|
|
|
struct page *page;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2017-04-20 21:37:43 +00:00
|
|
|
if (likely(order == 0)) {
|
2020-10-03 05:21:45 +00:00
|
|
|
/*
|
|
|
|
* MIGRATE_MOVABLE pcplist could have the pages on CMA area and
|
|
|
|
* we need to skip it when CMA area isn't allowed.
|
|
|
|
*/
|
|
|
|
if (!IS_ENABLED(CONFIG_CMA) || alloc_flags & ALLOC_CMA ||
|
|
|
|
migratetype != MIGRATE_MOVABLE) {
|
|
|
|
page = rmqueue_pcplist(preferred_zone, zone, gfp_flags,
|
2019-05-14 00:22:40 +00:00
|
|
|
migratetype, alloc_flags);
|
2020-10-03 05:21:45 +00:00
|
|
|
goto out;
|
|
|
|
}
|
2017-02-24 22:56:26 +00:00
|
|
|
}
|
2016-06-03 21:55:52 +00:00
|
|
|
|
2017-02-24 22:56:26 +00:00
|
|
|
/*
|
|
|
|
* We most definitely don't want callers attempting to
|
|
|
|
* allocate greater than order-1 page units with __GFP_NOFAIL.
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE((gfp_flags & __GFP_NOFAIL) && (order > 1));
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
2015-11-07 00:28:37 +00:00
|
|
|
|
2017-02-24 22:56:26 +00:00
|
|
|
do {
|
|
|
|
page = NULL;
|
2020-10-03 05:21:45 +00:00
|
|
|
/*
|
|
|
|
* order-0 request can reach here when the pcplist is skipped
|
|
|
|
* due to non-CMA allocation context. HIGHATOMIC area is
|
|
|
|
* reserved for high-order atomic allocation, so order-0
|
|
|
|
* request should skip it.
|
|
|
|
*/
|
|
|
|
if (order > 0 && alloc_flags & ALLOC_HARDER) {
|
2017-02-24 22:56:26 +00:00
|
|
|
page = __rmqueue_smallest(zone, order, MIGRATE_HIGHATOMIC);
|
|
|
|
if (page)
|
|
|
|
trace_mm_page_alloc_zone_locked(page, order, migratetype);
|
|
|
|
}
|
2006-01-06 08:11:20 +00:00
|
|
|
if (!page)
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
page = __rmqueue(zone, order, migratetype, alloc_flags);
|
2017-02-24 22:56:26 +00:00
|
|
|
} while (page && check_new_pages(page, order));
|
|
|
|
spin_unlock(&zone->lock);
|
|
|
|
if (!page)
|
|
|
|
goto failed;
|
|
|
|
__mod_zone_freepage_state(zone, -(1 << order),
|
|
|
|
get_pcppage_migratetype(page));
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2016-07-28 22:46:56 +00:00
|
|
|
__count_zid_vm_events(PGALLOC, page_zonenum(page), 1 << order);
|
2017-01-11 00:57:42 +00:00
|
|
|
zone_statistics(preferred_zone, zone);
|
2006-01-06 08:11:20 +00:00
|
|
|
local_irq_restore(flags);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2017-02-24 22:56:26 +00:00
|
|
|
out:
|
mm, page_alloc: do not wake kswapd with zone lock held
syzbot reported the following regression in the latest merge window and
it was confirmed by Qian Cai that a similar bug was visible from a
different context.
======================================================
WARNING: possible circular locking dependency detected
4.20.0+ #297 Not tainted
------------------------------------------------------
syz-executor0/8529 is trying to acquire lock:
000000005e7fb829 (&pgdat->kswapd_wait){....}, at:
__wake_up_common_lock+0x19e/0x330 kernel/sched/wait.c:120
but task is already holding lock:
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: spin_lock
include/linux/spinlock.h:329 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: rmqueue_bulk
mm/page_alloc.c:2548 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: __rmqueue_pcplist
mm/page_alloc.c:3021 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: rmqueue_pcplist
mm/page_alloc.c:3050 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at: rmqueue
mm/page_alloc.c:3072 [inline]
000000009bb7bae0 (&(&zone->lock)->rlock){-.-.}, at:
get_page_from_freelist+0x1bae/0x52a0 mm/page_alloc.c:3491
It appears to be a false positive in that the only way the lock ordering
should be inverted is if kswapd is waking itself and the wakeup
allocates debugging objects which should already be allocated if it's
kswapd doing the waking. Nevertheless, the possibility exists and so
it's best to avoid the problem.
This patch flags a zone as needing a kswapd using the, surprisingly,
unused zone flag field. The flag is read without the lock held to do
the wakeup. It's possible that the flag setting context is not the same
as the flag clearing context or for small races to occur. However, each
race possibility is harmless and there is no visible degredation in
fragmentation treatment.
While zone->flag could have continued to be unused, there is potential
for moving some existing fields into the flags field instead.
Particularly read-mostly ones like zone->initialized and
zone->contiguous.
Link: http://lkml.kernel.org/r/20190103225712.GJ31517@techsingularity.net
Fixes: 1c30844d2dfe ("mm: reclaim small amounts of memory when an external fragmentation event occurs")
Reported-by: syzbot+93d94a001cfbce9e60e1@syzkaller.appspotmail.com
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Qian Cai <cai@lca.pw>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-08 23:23:39 +00:00
|
|
|
/* Separate test+clear to avoid unnecessary atomics */
|
|
|
|
if (test_bit(ZONE_BOOSTED_WATERMARK, &zone->flags)) {
|
|
|
|
clear_bit(ZONE_BOOSTED_WATERMARK, &zone->flags);
|
|
|
|
wakeup_kswapd(zone, 0, 0, zone_idx(zone));
|
|
|
|
}
|
|
|
|
|
2017-02-24 22:56:26 +00:00
|
|
|
VM_BUG_ON_PAGE(page && bad_range(zone, page), page);
|
2005-04-16 22:20:36 +00:00
|
|
|
return page;
|
2006-01-06 08:11:20 +00:00
|
|
|
|
|
|
|
failed:
|
|
|
|
local_irq_restore(flags);
|
|
|
|
return NULL;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2006-12-08 10:39:45 +00:00
|
|
|
#ifdef CONFIG_FAIL_PAGE_ALLOC
|
|
|
|
|
2011-07-26 23:09:03 +00:00
|
|
|
static struct {
|
2006-12-08 10:39:45 +00:00
|
|
|
struct fault_attr attr;
|
|
|
|
|
2015-09-26 22:04:07 +00:00
|
|
|
bool ignore_gfp_highmem;
|
2015-11-07 00:28:28 +00:00
|
|
|
bool ignore_gfp_reclaim;
|
2007-07-16 06:40:23 +00:00
|
|
|
u32 min_order;
|
2006-12-08 10:39:45 +00:00
|
|
|
} fail_page_alloc = {
|
|
|
|
.attr = FAULT_ATTR_INITIALIZER,
|
2015-11-07 00:28:28 +00:00
|
|
|
.ignore_gfp_reclaim = true,
|
2015-09-26 22:04:07 +00:00
|
|
|
.ignore_gfp_highmem = true,
|
2007-07-16 06:40:23 +00:00
|
|
|
.min_order = 1,
|
2006-12-08 10:39:45 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
static int __init setup_fail_page_alloc(char *str)
|
|
|
|
{
|
|
|
|
return setup_fault_attr(&fail_page_alloc.attr, str);
|
|
|
|
}
|
|
|
|
__setup("fail_page_alloc=", setup_fail_page_alloc);
|
|
|
|
|
2018-12-28 08:39:23 +00:00
|
|
|
static bool __should_fail_alloc_page(gfp_t gfp_mask, unsigned int order)
|
2006-12-08 10:39:45 +00:00
|
|
|
{
|
2007-07-16 06:40:23 +00:00
|
|
|
if (order < fail_page_alloc.min_order)
|
2012-07-31 23:41:51 +00:00
|
|
|
return false;
|
2006-12-08 10:39:45 +00:00
|
|
|
if (gfp_mask & __GFP_NOFAIL)
|
2012-07-31 23:41:51 +00:00
|
|
|
return false;
|
2006-12-08 10:39:45 +00:00
|
|
|
if (fail_page_alloc.ignore_gfp_highmem && (gfp_mask & __GFP_HIGHMEM))
|
2012-07-31 23:41:51 +00:00
|
|
|
return false;
|
2015-11-07 00:28:28 +00:00
|
|
|
if (fail_page_alloc.ignore_gfp_reclaim &&
|
|
|
|
(gfp_mask & __GFP_DIRECT_RECLAIM))
|
2012-07-31 23:41:51 +00:00
|
|
|
return false;
|
2006-12-08 10:39:45 +00:00
|
|
|
|
|
|
|
return should_fail(&fail_page_alloc.attr, 1 << order);
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
|
|
|
|
|
|
|
|
static int __init fail_page_alloc_debugfs(void)
|
|
|
|
{
|
2018-06-14 22:27:58 +00:00
|
|
|
umode_t mode = S_IFREG | 0600;
|
2006-12-08 10:39:45 +00:00
|
|
|
struct dentry *dir;
|
|
|
|
|
2011-08-03 23:21:01 +00:00
|
|
|
dir = fault_create_debugfs_attr("fail_page_alloc", NULL,
|
|
|
|
&fail_page_alloc.attr);
|
2011-07-26 23:09:03 +00:00
|
|
|
|
2019-03-05 23:46:09 +00:00
|
|
|
debugfs_create_bool("ignore-gfp-wait", mode, dir,
|
|
|
|
&fail_page_alloc.ignore_gfp_reclaim);
|
|
|
|
debugfs_create_bool("ignore-gfp-highmem", mode, dir,
|
|
|
|
&fail_page_alloc.ignore_gfp_highmem);
|
|
|
|
debugfs_create_u32("min-order", mode, dir, &fail_page_alloc.min_order);
|
2006-12-08 10:39:45 +00:00
|
|
|
|
2019-03-05 23:46:09 +00:00
|
|
|
return 0;
|
2006-12-08 10:39:45 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
late_initcall(fail_page_alloc_debugfs);
|
|
|
|
|
|
|
|
#endif /* CONFIG_FAULT_INJECTION_DEBUG_FS */
|
|
|
|
|
|
|
|
#else /* CONFIG_FAIL_PAGE_ALLOC */
|
|
|
|
|
2018-12-28 08:39:23 +00:00
|
|
|
static inline bool __should_fail_alloc_page(gfp_t gfp_mask, unsigned int order)
|
2006-12-08 10:39:45 +00:00
|
|
|
{
|
2012-07-31 23:41:51 +00:00
|
|
|
return false;
|
2006-12-08 10:39:45 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* CONFIG_FAIL_PAGE_ALLOC */
|
|
|
|
|
2020-08-27 22:01:10 +00:00
|
|
|
noinline bool should_fail_alloc_page(gfp_t gfp_mask, unsigned int order)
|
2018-12-28 08:39:23 +00:00
|
|
|
{
|
|
|
|
return __should_fail_alloc_page(gfp_mask, order);
|
|
|
|
}
|
|
|
|
ALLOW_ERROR_INJECTION(should_fail_alloc_page, TRUE);
|
|
|
|
|
page_alloc: consider highatomic reserve in watermark fast
zone_watermark_fast was introduced by commit 48ee5f3696f6 ("mm,
page_alloc: shortcut watermark checks for order-0 pages"). The commit
simply checks if free pages is bigger than watermark without additional
calculation such like reducing watermark.
It considered free cma pages but it did not consider highatomic reserved.
This may incur exhaustion of free pages except high order atomic free
pages.
Assume that reserved_highatomic pageblock is bigger than watermark min,
and there are only few free pages except high order atomic free. Because
zone_watermark_fast passes the allocation without considering high order
atomic free, normal reclaimable allocation like GFP_HIGHUSER will consume
all the free pages. Then finally order-0 atomic allocation may fail on
allocation.
This means watermark min is not protected against non-atomic allocation.
The order-0 atomic allocation with ALLOC_HARDER unwantedly can be failed.
Additionally the __GFP_MEMALLOC allocation with ALLOC_NO_WATERMARKS also
can be failed.
To avoid the problem, zone_watermark_fast should consider highatomic
reserve. If the actual size of high atomic free is counted accurately
like cma free, we may use it. On this patch just use
nr_reserved_highatomic. Additionally introduce
__zone_watermark_unusable_free to factor out common parts between
zone_watermark_fast and __zone_watermark_ok.
This is an example of ALLOC_HARDER allocation failure using v4.19 based
kernel.
Binder:9343_3: page allocation failure: order:0, mode:0x480020(GFP_ATOMIC), nodemask=(null)
Call trace:
[<ffffff8008f40f8c>] dump_stack+0xb8/0xf0
[<ffffff8008223320>] warn_alloc+0xd8/0x12c
[<ffffff80082245e4>] __alloc_pages_nodemask+0x120c/0x1250
[<ffffff800827f6e8>] new_slab+0x128/0x604
[<ffffff800827b0cc>] ___slab_alloc+0x508/0x670
[<ffffff800827ba00>] __kmalloc+0x2f8/0x310
[<ffffff80084ac3e0>] context_struct_to_string+0x104/0x1cc
[<ffffff80084ad8fc>] security_sid_to_context_core+0x74/0x144
[<ffffff80084ad880>] security_sid_to_context+0x10/0x18
[<ffffff800849bd80>] selinux_secid_to_secctx+0x20/0x28
[<ffffff800849109c>] security_secid_to_secctx+0x3c/0x70
[<ffffff8008bfe118>] binder_transaction+0xe68/0x454c
Mem-Info:
active_anon:102061 inactive_anon:81551 isolated_anon:0
active_file:59102 inactive_file:68924 isolated_file:64
unevictable:611 dirty:63 writeback:0 unstable:0
slab_reclaimable:13324 slab_unreclaimable:44354
mapped:83015 shmem:4858 pagetables:26316 bounce:0
free:2727 free_pcp:1035 free_cma:178
Node 0 active_anon:408244kB inactive_anon:326204kB active_file:236408kB inactive_file:275696kB unevictable:2444kB isolated(anon):0kB isolated(file):256kB mapped:332060kB dirty:252kB writeback:0kB shmem:19432kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:10908kB min:6192kB low:44388kB high:47060kB active_anon:409160kB inactive_anon:325924kB active_file:235820kB inactive_file:276628kB unevictable:2444kB writepending:252kB present:3076096kB managed:2673676kB mlocked:2444kB kernel_stack:62512kB pagetables:105264kB bounce:0kB free_pcp:4140kB local_pcp:40kB free_cma:712kB
lowmem_reserve[]: 0 0
Normal: 505*4kB (H) 357*8kB (H) 201*16kB (H) 65*32kB (H) 1*64kB (H) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 10236kB
138826 total pagecache pages
5460 pages in swap cache
Swap cache stats: add 8273090, delete 8267506, find 1004381/4060142
This is an example of ALLOC_NO_WATERMARKS allocation failure using v4.14
based kernel.
kswapd0: page allocation failure: order:0, mode:0x140000a(GFP_NOIO|__GFP_HIGHMEM|__GFP_MOVABLE), nodemask=(null)
kswapd0 cpuset=/ mems_allowed=0
CPU: 4 PID: 1221 Comm: kswapd0 Not tainted 4.14.113-18770262-userdebug #1
Call trace:
[<0000000000000000>] dump_backtrace+0x0/0x248
[<0000000000000000>] show_stack+0x18/0x20
[<0000000000000000>] __dump_stack+0x20/0x28
[<0000000000000000>] dump_stack+0x68/0x90
[<0000000000000000>] warn_alloc+0x104/0x198
[<0000000000000000>] __alloc_pages_nodemask+0xdc0/0xdf0
[<0000000000000000>] zs_malloc+0x148/0x3d0
[<0000000000000000>] zram_bvec_rw+0x410/0x798
[<0000000000000000>] zram_rw_page+0x88/0xdc
[<0000000000000000>] bdev_write_page+0x70/0xbc
[<0000000000000000>] __swap_writepage+0x58/0x37c
[<0000000000000000>] swap_writepage+0x40/0x4c
[<0000000000000000>] shrink_page_list+0xc30/0xf48
[<0000000000000000>] shrink_inactive_list+0x2b0/0x61c
[<0000000000000000>] shrink_node_memcg+0x23c/0x618
[<0000000000000000>] shrink_node+0x1c8/0x304
[<0000000000000000>] kswapd+0x680/0x7c4
[<0000000000000000>] kthread+0x110/0x120
[<0000000000000000>] ret_from_fork+0x10/0x18
Mem-Info:
active_anon:111826 inactive_anon:65557 isolated_anon:0\x0a active_file:44260 inactive_file:83422 isolated_file:0\x0a unevictable:4158 dirty:117 writeback:0 unstable:0\x0a slab_reclaimable:13943 slab_unreclaimable:43315\x0a mapped:102511 shmem:3299 pagetables:19566 bounce:0\x0a free:3510 free_pcp:553 free_cma:0
Node 0 active_anon:447304kB inactive_anon:262228kB active_file:177040kB inactive_file:333688kB unevictable:16632kB isolated(anon):0kB isolated(file):0kB mapped:410044kB d irty:468kB writeback:0kB shmem:13196kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:14040kB min:7440kB low:94500kB high:98136kB reserved_highatomic:32768KB active_anon:447336kB inactive_anon:261668kB active_file:177572kB inactive_file:333768k B unevictable:16632kB writepending:480kB present:4081664kB managed:3637088kB mlocked:16632kB kernel_stack:47072kB pagetables:78264kB bounce:0kB free_pcp:2280kB local_pcp:720kB free_cma:0kB [ 4738.329607] lowmem_reserve[]: 0 0
Normal: 860*4kB (H) 453*8kB (H) 180*16kB (H) 26*32kB (H) 34*64kB (H) 6*128kB (H) 2*256kB (H) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 14232kB
This is trace log which shows GFP_HIGHUSER consumes free pages right
before ALLOC_NO_WATERMARKS.
<...>-22275 [006] .... 889.213383: mm_page_alloc: page=00000000d2be5665 pfn=970744 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213385: mm_page_alloc: page=000000004b2335c2 pfn=970745 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213387: mm_page_alloc: page=00000000017272e1 pfn=970278 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213389: mm_page_alloc: page=00000000c4be79fb pfn=970279 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213391: mm_page_alloc: page=00000000f8a51d4f pfn=970260 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213393: mm_page_alloc: page=000000006ba8f5ac pfn=970261 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213395: mm_page_alloc: page=00000000819f1cd3 pfn=970196 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213396: mm_page_alloc: page=00000000f6b72a64 pfn=970197 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
kswapd0-1207 [005] ...1 889.213398: mm_page_alloc: page= (null) pfn=0 order=0 migratetype=1 nr_free=3650 gfp_flags=GFP_NOWAIT|__GFP_HIGHMEM|__GFP_NOWARN|__GFP_MOVABLE
[jaewon31.kim@samsung.com: remove redundant code for high-order]
Link: http://lkml.kernel.org/r/20200623035242.27232-1-jaewon31.kim@samsung.com
Reported-by: Yong-Taek Lee <ytk.lee@samsung.com>
Suggested-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yong-Taek Lee <ytk.lee@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Link: http://lkml.kernel.org/r/20200619235958.11283-1-jaewon31.kim@samsung.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:20 +00:00
|
|
|
static inline long __zone_watermark_unusable_free(struct zone *z,
|
|
|
|
unsigned int order, unsigned int alloc_flags)
|
|
|
|
{
|
|
|
|
const bool alloc_harder = (alloc_flags & (ALLOC_HARDER|ALLOC_OOM));
|
|
|
|
long unusable_free = (1 << order) - 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the caller does not have rights to ALLOC_HARDER then subtract
|
|
|
|
* the high-atomic reserves. This will over-estimate the size of the
|
|
|
|
* atomic reserve but it avoids a search.
|
|
|
|
*/
|
|
|
|
if (likely(!alloc_harder))
|
|
|
|
unusable_free += z->nr_reserved_highatomic;
|
|
|
|
|
|
|
|
#ifdef CONFIG_CMA
|
|
|
|
/* If allocation can't use CMA areas don't use free CMA pages */
|
|
|
|
if (!(alloc_flags & ALLOC_CMA))
|
|
|
|
unusable_free += zone_page_state(z, NR_FREE_CMA_PAGES);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
return unusable_free;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
2015-11-07 00:28:40 +00:00
|
|
|
* Return true if free base pages are above 'mark'. For high-order checks it
|
|
|
|
* will return true of the order-0 watermark is reached and there is at least
|
|
|
|
* one free page of a suitable size. Checking now avoids taking the zone lock
|
|
|
|
* to check in the allocation paths if no pages are free.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2016-05-20 23:57:12 +00:00
|
|
|
bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
|
2020-06-03 22:59:01 +00:00
|
|
|
int highest_zoneidx, unsigned int alloc_flags,
|
2016-05-20 23:57:12 +00:00
|
|
|
long free_pages)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2007-02-10 09:43:02 +00:00
|
|
|
long min = mark;
|
2005-04-16 22:20:36 +00:00
|
|
|
int o;
|
2017-09-06 23:24:50 +00:00
|
|
|
const bool alloc_harder = (alloc_flags & (ALLOC_HARDER|ALLOC_OOM));
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2015-11-07 00:28:37 +00:00
|
|
|
/* free_pages may go negative - that's OK */
|
page_alloc: consider highatomic reserve in watermark fast
zone_watermark_fast was introduced by commit 48ee5f3696f6 ("mm,
page_alloc: shortcut watermark checks for order-0 pages"). The commit
simply checks if free pages is bigger than watermark without additional
calculation such like reducing watermark.
It considered free cma pages but it did not consider highatomic reserved.
This may incur exhaustion of free pages except high order atomic free
pages.
Assume that reserved_highatomic pageblock is bigger than watermark min,
and there are only few free pages except high order atomic free. Because
zone_watermark_fast passes the allocation without considering high order
atomic free, normal reclaimable allocation like GFP_HIGHUSER will consume
all the free pages. Then finally order-0 atomic allocation may fail on
allocation.
This means watermark min is not protected against non-atomic allocation.
The order-0 atomic allocation with ALLOC_HARDER unwantedly can be failed.
Additionally the __GFP_MEMALLOC allocation with ALLOC_NO_WATERMARKS also
can be failed.
To avoid the problem, zone_watermark_fast should consider highatomic
reserve. If the actual size of high atomic free is counted accurately
like cma free, we may use it. On this patch just use
nr_reserved_highatomic. Additionally introduce
__zone_watermark_unusable_free to factor out common parts between
zone_watermark_fast and __zone_watermark_ok.
This is an example of ALLOC_HARDER allocation failure using v4.19 based
kernel.
Binder:9343_3: page allocation failure: order:0, mode:0x480020(GFP_ATOMIC), nodemask=(null)
Call trace:
[<ffffff8008f40f8c>] dump_stack+0xb8/0xf0
[<ffffff8008223320>] warn_alloc+0xd8/0x12c
[<ffffff80082245e4>] __alloc_pages_nodemask+0x120c/0x1250
[<ffffff800827f6e8>] new_slab+0x128/0x604
[<ffffff800827b0cc>] ___slab_alloc+0x508/0x670
[<ffffff800827ba00>] __kmalloc+0x2f8/0x310
[<ffffff80084ac3e0>] context_struct_to_string+0x104/0x1cc
[<ffffff80084ad8fc>] security_sid_to_context_core+0x74/0x144
[<ffffff80084ad880>] security_sid_to_context+0x10/0x18
[<ffffff800849bd80>] selinux_secid_to_secctx+0x20/0x28
[<ffffff800849109c>] security_secid_to_secctx+0x3c/0x70
[<ffffff8008bfe118>] binder_transaction+0xe68/0x454c
Mem-Info:
active_anon:102061 inactive_anon:81551 isolated_anon:0
active_file:59102 inactive_file:68924 isolated_file:64
unevictable:611 dirty:63 writeback:0 unstable:0
slab_reclaimable:13324 slab_unreclaimable:44354
mapped:83015 shmem:4858 pagetables:26316 bounce:0
free:2727 free_pcp:1035 free_cma:178
Node 0 active_anon:408244kB inactive_anon:326204kB active_file:236408kB inactive_file:275696kB unevictable:2444kB isolated(anon):0kB isolated(file):256kB mapped:332060kB dirty:252kB writeback:0kB shmem:19432kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:10908kB min:6192kB low:44388kB high:47060kB active_anon:409160kB inactive_anon:325924kB active_file:235820kB inactive_file:276628kB unevictable:2444kB writepending:252kB present:3076096kB managed:2673676kB mlocked:2444kB kernel_stack:62512kB pagetables:105264kB bounce:0kB free_pcp:4140kB local_pcp:40kB free_cma:712kB
lowmem_reserve[]: 0 0
Normal: 505*4kB (H) 357*8kB (H) 201*16kB (H) 65*32kB (H) 1*64kB (H) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 10236kB
138826 total pagecache pages
5460 pages in swap cache
Swap cache stats: add 8273090, delete 8267506, find 1004381/4060142
This is an example of ALLOC_NO_WATERMARKS allocation failure using v4.14
based kernel.
kswapd0: page allocation failure: order:0, mode:0x140000a(GFP_NOIO|__GFP_HIGHMEM|__GFP_MOVABLE), nodemask=(null)
kswapd0 cpuset=/ mems_allowed=0
CPU: 4 PID: 1221 Comm: kswapd0 Not tainted 4.14.113-18770262-userdebug #1
Call trace:
[<0000000000000000>] dump_backtrace+0x0/0x248
[<0000000000000000>] show_stack+0x18/0x20
[<0000000000000000>] __dump_stack+0x20/0x28
[<0000000000000000>] dump_stack+0x68/0x90
[<0000000000000000>] warn_alloc+0x104/0x198
[<0000000000000000>] __alloc_pages_nodemask+0xdc0/0xdf0
[<0000000000000000>] zs_malloc+0x148/0x3d0
[<0000000000000000>] zram_bvec_rw+0x410/0x798
[<0000000000000000>] zram_rw_page+0x88/0xdc
[<0000000000000000>] bdev_write_page+0x70/0xbc
[<0000000000000000>] __swap_writepage+0x58/0x37c
[<0000000000000000>] swap_writepage+0x40/0x4c
[<0000000000000000>] shrink_page_list+0xc30/0xf48
[<0000000000000000>] shrink_inactive_list+0x2b0/0x61c
[<0000000000000000>] shrink_node_memcg+0x23c/0x618
[<0000000000000000>] shrink_node+0x1c8/0x304
[<0000000000000000>] kswapd+0x680/0x7c4
[<0000000000000000>] kthread+0x110/0x120
[<0000000000000000>] ret_from_fork+0x10/0x18
Mem-Info:
active_anon:111826 inactive_anon:65557 isolated_anon:0\x0a active_file:44260 inactive_file:83422 isolated_file:0\x0a unevictable:4158 dirty:117 writeback:0 unstable:0\x0a slab_reclaimable:13943 slab_unreclaimable:43315\x0a mapped:102511 shmem:3299 pagetables:19566 bounce:0\x0a free:3510 free_pcp:553 free_cma:0
Node 0 active_anon:447304kB inactive_anon:262228kB active_file:177040kB inactive_file:333688kB unevictable:16632kB isolated(anon):0kB isolated(file):0kB mapped:410044kB d irty:468kB writeback:0kB shmem:13196kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:14040kB min:7440kB low:94500kB high:98136kB reserved_highatomic:32768KB active_anon:447336kB inactive_anon:261668kB active_file:177572kB inactive_file:333768k B unevictable:16632kB writepending:480kB present:4081664kB managed:3637088kB mlocked:16632kB kernel_stack:47072kB pagetables:78264kB bounce:0kB free_pcp:2280kB local_pcp:720kB free_cma:0kB [ 4738.329607] lowmem_reserve[]: 0 0
Normal: 860*4kB (H) 453*8kB (H) 180*16kB (H) 26*32kB (H) 34*64kB (H) 6*128kB (H) 2*256kB (H) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 14232kB
This is trace log which shows GFP_HIGHUSER consumes free pages right
before ALLOC_NO_WATERMARKS.
<...>-22275 [006] .... 889.213383: mm_page_alloc: page=00000000d2be5665 pfn=970744 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213385: mm_page_alloc: page=000000004b2335c2 pfn=970745 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213387: mm_page_alloc: page=00000000017272e1 pfn=970278 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213389: mm_page_alloc: page=00000000c4be79fb pfn=970279 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213391: mm_page_alloc: page=00000000f8a51d4f pfn=970260 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213393: mm_page_alloc: page=000000006ba8f5ac pfn=970261 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213395: mm_page_alloc: page=00000000819f1cd3 pfn=970196 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213396: mm_page_alloc: page=00000000f6b72a64 pfn=970197 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
kswapd0-1207 [005] ...1 889.213398: mm_page_alloc: page= (null) pfn=0 order=0 migratetype=1 nr_free=3650 gfp_flags=GFP_NOWAIT|__GFP_HIGHMEM|__GFP_NOWARN|__GFP_MOVABLE
[jaewon31.kim@samsung.com: remove redundant code for high-order]
Link: http://lkml.kernel.org/r/20200623035242.27232-1-jaewon31.kim@samsung.com
Reported-by: Yong-Taek Lee <ytk.lee@samsung.com>
Suggested-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yong-Taek Lee <ytk.lee@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Link: http://lkml.kernel.org/r/20200619235958.11283-1-jaewon31.kim@samsung.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:20 +00:00
|
|
|
free_pages -= __zone_watermark_unusable_free(z, order, alloc_flags);
|
2015-11-07 00:28:37 +00:00
|
|
|
|
2005-11-14 00:06:43 +00:00
|
|
|
if (alloc_flags & ALLOC_HIGH)
|
2005-04-16 22:20:36 +00:00
|
|
|
min -= min / 2;
|
2015-11-07 00:28:37 +00:00
|
|
|
|
page_alloc: consider highatomic reserve in watermark fast
zone_watermark_fast was introduced by commit 48ee5f3696f6 ("mm,
page_alloc: shortcut watermark checks for order-0 pages"). The commit
simply checks if free pages is bigger than watermark without additional
calculation such like reducing watermark.
It considered free cma pages but it did not consider highatomic reserved.
This may incur exhaustion of free pages except high order atomic free
pages.
Assume that reserved_highatomic pageblock is bigger than watermark min,
and there are only few free pages except high order atomic free. Because
zone_watermark_fast passes the allocation without considering high order
atomic free, normal reclaimable allocation like GFP_HIGHUSER will consume
all the free pages. Then finally order-0 atomic allocation may fail on
allocation.
This means watermark min is not protected against non-atomic allocation.
The order-0 atomic allocation with ALLOC_HARDER unwantedly can be failed.
Additionally the __GFP_MEMALLOC allocation with ALLOC_NO_WATERMARKS also
can be failed.
To avoid the problem, zone_watermark_fast should consider highatomic
reserve. If the actual size of high atomic free is counted accurately
like cma free, we may use it. On this patch just use
nr_reserved_highatomic. Additionally introduce
__zone_watermark_unusable_free to factor out common parts between
zone_watermark_fast and __zone_watermark_ok.
This is an example of ALLOC_HARDER allocation failure using v4.19 based
kernel.
Binder:9343_3: page allocation failure: order:0, mode:0x480020(GFP_ATOMIC), nodemask=(null)
Call trace:
[<ffffff8008f40f8c>] dump_stack+0xb8/0xf0
[<ffffff8008223320>] warn_alloc+0xd8/0x12c
[<ffffff80082245e4>] __alloc_pages_nodemask+0x120c/0x1250
[<ffffff800827f6e8>] new_slab+0x128/0x604
[<ffffff800827b0cc>] ___slab_alloc+0x508/0x670
[<ffffff800827ba00>] __kmalloc+0x2f8/0x310
[<ffffff80084ac3e0>] context_struct_to_string+0x104/0x1cc
[<ffffff80084ad8fc>] security_sid_to_context_core+0x74/0x144
[<ffffff80084ad880>] security_sid_to_context+0x10/0x18
[<ffffff800849bd80>] selinux_secid_to_secctx+0x20/0x28
[<ffffff800849109c>] security_secid_to_secctx+0x3c/0x70
[<ffffff8008bfe118>] binder_transaction+0xe68/0x454c
Mem-Info:
active_anon:102061 inactive_anon:81551 isolated_anon:0
active_file:59102 inactive_file:68924 isolated_file:64
unevictable:611 dirty:63 writeback:0 unstable:0
slab_reclaimable:13324 slab_unreclaimable:44354
mapped:83015 shmem:4858 pagetables:26316 bounce:0
free:2727 free_pcp:1035 free_cma:178
Node 0 active_anon:408244kB inactive_anon:326204kB active_file:236408kB inactive_file:275696kB unevictable:2444kB isolated(anon):0kB isolated(file):256kB mapped:332060kB dirty:252kB writeback:0kB shmem:19432kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:10908kB min:6192kB low:44388kB high:47060kB active_anon:409160kB inactive_anon:325924kB active_file:235820kB inactive_file:276628kB unevictable:2444kB writepending:252kB present:3076096kB managed:2673676kB mlocked:2444kB kernel_stack:62512kB pagetables:105264kB bounce:0kB free_pcp:4140kB local_pcp:40kB free_cma:712kB
lowmem_reserve[]: 0 0
Normal: 505*4kB (H) 357*8kB (H) 201*16kB (H) 65*32kB (H) 1*64kB (H) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 10236kB
138826 total pagecache pages
5460 pages in swap cache
Swap cache stats: add 8273090, delete 8267506, find 1004381/4060142
This is an example of ALLOC_NO_WATERMARKS allocation failure using v4.14
based kernel.
kswapd0: page allocation failure: order:0, mode:0x140000a(GFP_NOIO|__GFP_HIGHMEM|__GFP_MOVABLE), nodemask=(null)
kswapd0 cpuset=/ mems_allowed=0
CPU: 4 PID: 1221 Comm: kswapd0 Not tainted 4.14.113-18770262-userdebug #1
Call trace:
[<0000000000000000>] dump_backtrace+0x0/0x248
[<0000000000000000>] show_stack+0x18/0x20
[<0000000000000000>] __dump_stack+0x20/0x28
[<0000000000000000>] dump_stack+0x68/0x90
[<0000000000000000>] warn_alloc+0x104/0x198
[<0000000000000000>] __alloc_pages_nodemask+0xdc0/0xdf0
[<0000000000000000>] zs_malloc+0x148/0x3d0
[<0000000000000000>] zram_bvec_rw+0x410/0x798
[<0000000000000000>] zram_rw_page+0x88/0xdc
[<0000000000000000>] bdev_write_page+0x70/0xbc
[<0000000000000000>] __swap_writepage+0x58/0x37c
[<0000000000000000>] swap_writepage+0x40/0x4c
[<0000000000000000>] shrink_page_list+0xc30/0xf48
[<0000000000000000>] shrink_inactive_list+0x2b0/0x61c
[<0000000000000000>] shrink_node_memcg+0x23c/0x618
[<0000000000000000>] shrink_node+0x1c8/0x304
[<0000000000000000>] kswapd+0x680/0x7c4
[<0000000000000000>] kthread+0x110/0x120
[<0000000000000000>] ret_from_fork+0x10/0x18
Mem-Info:
active_anon:111826 inactive_anon:65557 isolated_anon:0\x0a active_file:44260 inactive_file:83422 isolated_file:0\x0a unevictable:4158 dirty:117 writeback:0 unstable:0\x0a slab_reclaimable:13943 slab_unreclaimable:43315\x0a mapped:102511 shmem:3299 pagetables:19566 bounce:0\x0a free:3510 free_pcp:553 free_cma:0
Node 0 active_anon:447304kB inactive_anon:262228kB active_file:177040kB inactive_file:333688kB unevictable:16632kB isolated(anon):0kB isolated(file):0kB mapped:410044kB d irty:468kB writeback:0kB shmem:13196kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:14040kB min:7440kB low:94500kB high:98136kB reserved_highatomic:32768KB active_anon:447336kB inactive_anon:261668kB active_file:177572kB inactive_file:333768k B unevictable:16632kB writepending:480kB present:4081664kB managed:3637088kB mlocked:16632kB kernel_stack:47072kB pagetables:78264kB bounce:0kB free_pcp:2280kB local_pcp:720kB free_cma:0kB [ 4738.329607] lowmem_reserve[]: 0 0
Normal: 860*4kB (H) 453*8kB (H) 180*16kB (H) 26*32kB (H) 34*64kB (H) 6*128kB (H) 2*256kB (H) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 14232kB
This is trace log which shows GFP_HIGHUSER consumes free pages right
before ALLOC_NO_WATERMARKS.
<...>-22275 [006] .... 889.213383: mm_page_alloc: page=00000000d2be5665 pfn=970744 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213385: mm_page_alloc: page=000000004b2335c2 pfn=970745 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213387: mm_page_alloc: page=00000000017272e1 pfn=970278 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213389: mm_page_alloc: page=00000000c4be79fb pfn=970279 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213391: mm_page_alloc: page=00000000f8a51d4f pfn=970260 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213393: mm_page_alloc: page=000000006ba8f5ac pfn=970261 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213395: mm_page_alloc: page=00000000819f1cd3 pfn=970196 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213396: mm_page_alloc: page=00000000f6b72a64 pfn=970197 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
kswapd0-1207 [005] ...1 889.213398: mm_page_alloc: page= (null) pfn=0 order=0 migratetype=1 nr_free=3650 gfp_flags=GFP_NOWAIT|__GFP_HIGHMEM|__GFP_NOWARN|__GFP_MOVABLE
[jaewon31.kim@samsung.com: remove redundant code for high-order]
Link: http://lkml.kernel.org/r/20200623035242.27232-1-jaewon31.kim@samsung.com
Reported-by: Yong-Taek Lee <ytk.lee@samsung.com>
Suggested-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yong-Taek Lee <ytk.lee@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Link: http://lkml.kernel.org/r/20200619235958.11283-1-jaewon31.kim@samsung.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:20 +00:00
|
|
|
if (unlikely(alloc_harder)) {
|
2017-09-06 23:24:50 +00:00
|
|
|
/*
|
|
|
|
* OOM victims can try even harder than normal ALLOC_HARDER
|
|
|
|
* users on the grounds that it's definitely going to be in
|
|
|
|
* the exit path shortly and free memory. Any allocation it
|
|
|
|
* makes during the free path will be small and short-lived.
|
|
|
|
*/
|
|
|
|
if (alloc_flags & ALLOC_OOM)
|
|
|
|
min -= min / 2;
|
|
|
|
else
|
|
|
|
min -= min / 4;
|
|
|
|
}
|
|
|
|
|
2015-11-07 00:28:40 +00:00
|
|
|
/*
|
|
|
|
* Check watermarks for an order-0 allocation request. If these
|
|
|
|
* are not met, then a high-order request also cannot go ahead
|
|
|
|
* even if a suitable page happened to be free.
|
|
|
|
*/
|
2020-06-03 22:59:01 +00:00
|
|
|
if (free_pages <= min + z->lowmem_reserve[highest_zoneidx])
|
mm: page allocator: adjust the per-cpu counter threshold when memory is low
Commit aa45484 ("calculate a better estimate of NR_FREE_PAGES when memory
is low") noted that watermarks were based on the vmstat NR_FREE_PAGES. To
avoid synchronization overhead, these counters are maintained on a per-cpu
basis and drained both periodically and when a threshold is above a
threshold. On large CPU systems, the difference between the estimate and
real value of NR_FREE_PAGES can be very high. The system can get into a
case where pages are allocated far below the min watermark potentially
causing livelock issues. The commit solved the problem by taking a better
reading of NR_FREE_PAGES when memory was low.
Unfortately, as reported by Shaohua Li this accurate reading can consume a
large amount of CPU time on systems with many sockets due to cache line
bouncing. This patch takes a different approach. For large machines
where counter drift might be unsafe and while kswapd is awake, the per-cpu
thresholds for the target pgdat are reduced to limit the level of drift to
what should be a safe level. This incurs a performance penalty in heavy
memory pressure by a factor that depends on the workload and the machine
but the machine should function correctly without accidentally exhausting
all memory on a node. There is an additional cost when kswapd wakes and
sleeps but the event is not expected to be frequent - in Shaohua's test
case, there was one recorded sleep and wake event at least.
To ensure that kswapd wakes up, a safe version of zone_watermark_ok() is
introduced that takes a more accurate reading of NR_FREE_PAGES when called
from wakeup_kswapd, when deciding whether it is really safe to go back to
sleep in sleeping_prematurely() and when deciding if a zone is really
balanced or not in balance_pgdat(). We are still using an expensive
function but limiting how often it is called.
When the test case is reproduced, the time spent in the watermark
functions is reduced. The following report is on the percentage of time
spent cumulatively spent in the functions zone_nr_free_pages(),
zone_watermark_ok(), __zone_watermark_ok(), zone_watermark_ok_safe(),
zone_page_state_snapshot(), zone_page_state().
vanilla 11.6615%
disable-threshold 0.2584%
David said:
: We had to pull aa454840 "mm: page allocator: calculate a better estimate
: of NR_FREE_PAGES when memory is low and kswapd is awake" from 2.6.36
: internally because tests showed that it would cause the machine to stall
: as the result of heavy kswapd activity. I merged it back with this fix as
: it is pending in the -mm tree and it solves the issue we were seeing, so I
: definitely think this should be pushed to -stable (and I would seriously
: consider it for 2.6.37 inclusion even at this late date).
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reported-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Tested-by: Nicolas Bareil <nico@chdir.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: <stable@kernel.org> [2.6.37.1, 2.6.36.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:45:41 +00:00
|
|
|
return false;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2015-11-07 00:28:40 +00:00
|
|
|
/* If this is an order-0 request then the watermark is fine */
|
|
|
|
if (!order)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
/* For a high-order request, check at least one suitable page is free */
|
|
|
|
for (o = order; o < MAX_ORDER; o++) {
|
|
|
|
struct free_area *area = &z->free_area[o];
|
|
|
|
int mt;
|
|
|
|
|
|
|
|
if (!area->nr_free)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
for (mt = 0; mt < MIGRATE_PCPTYPES; mt++) {
|
2019-05-14 22:41:32 +00:00
|
|
|
if (!free_area_empty(area, mt))
|
2015-11-07 00:28:40 +00:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_CMA
|
2018-05-23 01:18:21 +00:00
|
|
|
if ((alloc_flags & ALLOC_CMA) &&
|
2019-05-14 22:41:32 +00:00
|
|
|
!free_area_empty(area, MIGRATE_CMA)) {
|
2015-11-07 00:28:40 +00:00
|
|
|
return true;
|
2018-05-23 01:18:21 +00:00
|
|
|
}
|
2015-11-07 00:28:40 +00:00
|
|
|
#endif
|
2020-04-02 04:09:50 +00:00
|
|
|
if (alloc_harder && !free_area_empty(area, MIGRATE_HIGHATOMIC))
|
mm, page_alloc: fix potential false positive in __zone_watermark_ok
Since commit 97a16fc82a7c ("mm, page_alloc: only enforce watermarks for
order-0 allocations"), __zone_watermark_ok() check for high-order
allocations will shortcut per-migratetype free list checks for
ALLOC_HARDER allocations, and return true as long as there's free page
of any migratetype. The intention is that ALLOC_HARDER can allocate
from MIGRATE_HIGHATOMIC free lists, while normal allocations can't.
However, as a side effect, the watermark check will then also return
true when there are pages only on the MIGRATE_ISOLATE list, or (prior to
CMA conversion to ZONE_MOVABLE) on the MIGRATE_CMA list. Since the
allocation cannot actually obtain isolated pages, and might not be able
to obtain CMA pages, this can result in a false positive.
The condition should be rare and perhaps the outcome is not a fatal one.
Still, it's better if the watermark check is correct. There also
shouldn't be a performance tradeoff here.
Link: http://lkml.kernel.org/r/20171102125001.23708-1-vbabka@suse.cz
Fixes: 97a16fc82a7c ("mm, page_alloc: only enforce watermarks for order-0 allocations")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:38:30 +00:00
|
|
|
return true;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2015-11-07 00:28:40 +00:00
|
|
|
return false;
|
mm: page allocator: adjust the per-cpu counter threshold when memory is low
Commit aa45484 ("calculate a better estimate of NR_FREE_PAGES when memory
is low") noted that watermarks were based on the vmstat NR_FREE_PAGES. To
avoid synchronization overhead, these counters are maintained on a per-cpu
basis and drained both periodically and when a threshold is above a
threshold. On large CPU systems, the difference between the estimate and
real value of NR_FREE_PAGES can be very high. The system can get into a
case where pages are allocated far below the min watermark potentially
causing livelock issues. The commit solved the problem by taking a better
reading of NR_FREE_PAGES when memory was low.
Unfortately, as reported by Shaohua Li this accurate reading can consume a
large amount of CPU time on systems with many sockets due to cache line
bouncing. This patch takes a different approach. For large machines
where counter drift might be unsafe and while kswapd is awake, the per-cpu
thresholds for the target pgdat are reduced to limit the level of drift to
what should be a safe level. This incurs a performance penalty in heavy
memory pressure by a factor that depends on the workload and the machine
but the machine should function correctly without accidentally exhausting
all memory on a node. There is an additional cost when kswapd wakes and
sleeps but the event is not expected to be frequent - in Shaohua's test
case, there was one recorded sleep and wake event at least.
To ensure that kswapd wakes up, a safe version of zone_watermark_ok() is
introduced that takes a more accurate reading of NR_FREE_PAGES when called
from wakeup_kswapd, when deciding whether it is really safe to go back to
sleep in sleeping_prematurely() and when deciding if a zone is really
balanced or not in balance_pgdat(). We are still using an expensive
function but limiting how often it is called.
When the test case is reproduced, the time spent in the watermark
functions is reduced. The following report is on the percentage of time
spent cumulatively spent in the functions zone_nr_free_pages(),
zone_watermark_ok(), __zone_watermark_ok(), zone_watermark_ok_safe(),
zone_page_state_snapshot(), zone_page_state().
vanilla 11.6615%
disable-threshold 0.2584%
David said:
: We had to pull aa454840 "mm: page allocator: calculate a better estimate
: of NR_FREE_PAGES when memory is low and kswapd is awake" from 2.6.36
: internally because tests showed that it would cause the machine to stall
: as the result of heavy kswapd activity. I merged it back with this fix as
: it is pending in the -mm tree and it solves the issue we were seeing, so I
: definitely think this should be pushed to -stable (and I would seriously
: consider it for 2.6.37 inclusion even at this late date).
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reported-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Tested-by: Nicolas Bareil <nico@chdir.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: <stable@kernel.org> [2.6.37.1, 2.6.36.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:45:41 +00:00
|
|
|
}
|
|
|
|
|
2014-06-04 23:10:21 +00:00
|
|
|
bool zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
|
2020-06-03 22:59:01 +00:00
|
|
|
int highest_zoneidx, unsigned int alloc_flags)
|
mm: page allocator: adjust the per-cpu counter threshold when memory is low
Commit aa45484 ("calculate a better estimate of NR_FREE_PAGES when memory
is low") noted that watermarks were based on the vmstat NR_FREE_PAGES. To
avoid synchronization overhead, these counters are maintained on a per-cpu
basis and drained both periodically and when a threshold is above a
threshold. On large CPU systems, the difference between the estimate and
real value of NR_FREE_PAGES can be very high. The system can get into a
case where pages are allocated far below the min watermark potentially
causing livelock issues. The commit solved the problem by taking a better
reading of NR_FREE_PAGES when memory was low.
Unfortately, as reported by Shaohua Li this accurate reading can consume a
large amount of CPU time on systems with many sockets due to cache line
bouncing. This patch takes a different approach. For large machines
where counter drift might be unsafe and while kswapd is awake, the per-cpu
thresholds for the target pgdat are reduced to limit the level of drift to
what should be a safe level. This incurs a performance penalty in heavy
memory pressure by a factor that depends on the workload and the machine
but the machine should function correctly without accidentally exhausting
all memory on a node. There is an additional cost when kswapd wakes and
sleeps but the event is not expected to be frequent - in Shaohua's test
case, there was one recorded sleep and wake event at least.
To ensure that kswapd wakes up, a safe version of zone_watermark_ok() is
introduced that takes a more accurate reading of NR_FREE_PAGES when called
from wakeup_kswapd, when deciding whether it is really safe to go back to
sleep in sleeping_prematurely() and when deciding if a zone is really
balanced or not in balance_pgdat(). We are still using an expensive
function but limiting how often it is called.
When the test case is reproduced, the time spent in the watermark
functions is reduced. The following report is on the percentage of time
spent cumulatively spent in the functions zone_nr_free_pages(),
zone_watermark_ok(), __zone_watermark_ok(), zone_watermark_ok_safe(),
zone_page_state_snapshot(), zone_page_state().
vanilla 11.6615%
disable-threshold 0.2584%
David said:
: We had to pull aa454840 "mm: page allocator: calculate a better estimate
: of NR_FREE_PAGES when memory is low and kswapd is awake" from 2.6.36
: internally because tests showed that it would cause the machine to stall
: as the result of heavy kswapd activity. I merged it back with this fix as
: it is pending in the -mm tree and it solves the issue we were seeing, so I
: definitely think this should be pushed to -stable (and I would seriously
: consider it for 2.6.37 inclusion even at this late date).
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reported-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Tested-by: Nicolas Bareil <nico@chdir.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: <stable@kernel.org> [2.6.37.1, 2.6.36.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:45:41 +00:00
|
|
|
{
|
2020-06-03 22:59:01 +00:00
|
|
|
return __zone_watermark_ok(z, order, mark, highest_zoneidx, alloc_flags,
|
mm: page allocator: adjust the per-cpu counter threshold when memory is low
Commit aa45484 ("calculate a better estimate of NR_FREE_PAGES when memory
is low") noted that watermarks were based on the vmstat NR_FREE_PAGES. To
avoid synchronization overhead, these counters are maintained on a per-cpu
basis and drained both periodically and when a threshold is above a
threshold. On large CPU systems, the difference between the estimate and
real value of NR_FREE_PAGES can be very high. The system can get into a
case where pages are allocated far below the min watermark potentially
causing livelock issues. The commit solved the problem by taking a better
reading of NR_FREE_PAGES when memory was low.
Unfortately, as reported by Shaohua Li this accurate reading can consume a
large amount of CPU time on systems with many sockets due to cache line
bouncing. This patch takes a different approach. For large machines
where counter drift might be unsafe and while kswapd is awake, the per-cpu
thresholds for the target pgdat are reduced to limit the level of drift to
what should be a safe level. This incurs a performance penalty in heavy
memory pressure by a factor that depends on the workload and the machine
but the machine should function correctly without accidentally exhausting
all memory on a node. There is an additional cost when kswapd wakes and
sleeps but the event is not expected to be frequent - in Shaohua's test
case, there was one recorded sleep and wake event at least.
To ensure that kswapd wakes up, a safe version of zone_watermark_ok() is
introduced that takes a more accurate reading of NR_FREE_PAGES when called
from wakeup_kswapd, when deciding whether it is really safe to go back to
sleep in sleeping_prematurely() and when deciding if a zone is really
balanced or not in balance_pgdat(). We are still using an expensive
function but limiting how often it is called.
When the test case is reproduced, the time spent in the watermark
functions is reduced. The following report is on the percentage of time
spent cumulatively spent in the functions zone_nr_free_pages(),
zone_watermark_ok(), __zone_watermark_ok(), zone_watermark_ok_safe(),
zone_page_state_snapshot(), zone_page_state().
vanilla 11.6615%
disable-threshold 0.2584%
David said:
: We had to pull aa454840 "mm: page allocator: calculate a better estimate
: of NR_FREE_PAGES when memory is low and kswapd is awake" from 2.6.36
: internally because tests showed that it would cause the machine to stall
: as the result of heavy kswapd activity. I merged it back with this fix as
: it is pending in the -mm tree and it solves the issue we were seeing, so I
: definitely think this should be pushed to -stable (and I would seriously
: consider it for 2.6.37 inclusion even at this late date).
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reported-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Tested-by: Nicolas Bareil <nico@chdir.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: <stable@kernel.org> [2.6.37.1, 2.6.36.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:45:41 +00:00
|
|
|
zone_page_state(z, NR_FREE_PAGES));
|
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:07 +00:00
|
|
|
static inline bool zone_watermark_fast(struct zone *z, unsigned int order,
|
2020-06-03 22:59:01 +00:00
|
|
|
unsigned long mark, int highest_zoneidx,
|
mm, page_alloc: skip ->waternark_boost for atomic order-0 allocations
When boosting is enabled, it is observed that rate of atomic order-0
allocation failures are high due to the fact that free levels in the
system are checked with ->watermark_boost offset. This is not a problem
for sleepable allocations but for atomic allocations which looks like
regression.
This problem is seen frequently on system setup of Android kernel running
on Snapdragon hardware with 4GB RAM size. When no extfrag event occurred
in the system, ->watermark_boost factor is zero, thus the watermark
configurations in the system are:
_watermark = (
[WMARK_MIN] = 1272, --> ~5MB
[WMARK_LOW] = 9067, --> ~36MB
[WMARK_HIGH] = 9385), --> ~38MB
watermark_boost = 0
After launching some memory hungry applications in Android which can cause
extfrag events in the system to an extent that ->watermark_boost can be
set to max i.e. default boost factor makes it to 150% of high watermark.
_watermark = (
[WMARK_MIN] = 1272, --> ~5MB
[WMARK_LOW] = 9067, --> ~36MB
[WMARK_HIGH] = 9385), --> ~38MB
watermark_boost = 14077, -->~57MB
With default system configuration, for an atomic order-0 allocation to
succeed, having free memory of ~2MB will suffice. But boosting makes the
min_wmark to ~61MB thus for an atomic order-0 allocation to be successful
system should have minimum of ~23MB of free memory(from calculations of
zone_watermark_ok(), min = 3/4(min/2)). But failures are observed despite
system is having ~20MB of free memory. In the testing, this is
reproducible as early as first 300secs since boot and with furtherlowram
configurations(<2GB) it is observed as early as first 150secs since boot.
These failures can be avoided by excluding the ->watermark_boost in
watermark caluculations for atomic order-0 allocations.
[akpm@linux-foundation.org: fix comment grammar, reflow comment]
[charante@codeaurora.org: fix suggested by Mel Gorman]
Link: http://lkml.kernel.org/r/31556793-57b1-1c21-1a9d-22674d9bd938@codeaurora.org
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Link: http://lkml.kernel.org/r/1589882284-21010-1-git-send-email-charante@codeaurora.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:24 +00:00
|
|
|
unsigned int alloc_flags, gfp_t gfp_mask)
|
2016-05-20 00:14:07 +00:00
|
|
|
{
|
page_alloc: consider highatomic reserve in watermark fast
zone_watermark_fast was introduced by commit 48ee5f3696f6 ("mm,
page_alloc: shortcut watermark checks for order-0 pages"). The commit
simply checks if free pages is bigger than watermark without additional
calculation such like reducing watermark.
It considered free cma pages but it did not consider highatomic reserved.
This may incur exhaustion of free pages except high order atomic free
pages.
Assume that reserved_highatomic pageblock is bigger than watermark min,
and there are only few free pages except high order atomic free. Because
zone_watermark_fast passes the allocation without considering high order
atomic free, normal reclaimable allocation like GFP_HIGHUSER will consume
all the free pages. Then finally order-0 atomic allocation may fail on
allocation.
This means watermark min is not protected against non-atomic allocation.
The order-0 atomic allocation with ALLOC_HARDER unwantedly can be failed.
Additionally the __GFP_MEMALLOC allocation with ALLOC_NO_WATERMARKS also
can be failed.
To avoid the problem, zone_watermark_fast should consider highatomic
reserve. If the actual size of high atomic free is counted accurately
like cma free, we may use it. On this patch just use
nr_reserved_highatomic. Additionally introduce
__zone_watermark_unusable_free to factor out common parts between
zone_watermark_fast and __zone_watermark_ok.
This is an example of ALLOC_HARDER allocation failure using v4.19 based
kernel.
Binder:9343_3: page allocation failure: order:0, mode:0x480020(GFP_ATOMIC), nodemask=(null)
Call trace:
[<ffffff8008f40f8c>] dump_stack+0xb8/0xf0
[<ffffff8008223320>] warn_alloc+0xd8/0x12c
[<ffffff80082245e4>] __alloc_pages_nodemask+0x120c/0x1250
[<ffffff800827f6e8>] new_slab+0x128/0x604
[<ffffff800827b0cc>] ___slab_alloc+0x508/0x670
[<ffffff800827ba00>] __kmalloc+0x2f8/0x310
[<ffffff80084ac3e0>] context_struct_to_string+0x104/0x1cc
[<ffffff80084ad8fc>] security_sid_to_context_core+0x74/0x144
[<ffffff80084ad880>] security_sid_to_context+0x10/0x18
[<ffffff800849bd80>] selinux_secid_to_secctx+0x20/0x28
[<ffffff800849109c>] security_secid_to_secctx+0x3c/0x70
[<ffffff8008bfe118>] binder_transaction+0xe68/0x454c
Mem-Info:
active_anon:102061 inactive_anon:81551 isolated_anon:0
active_file:59102 inactive_file:68924 isolated_file:64
unevictable:611 dirty:63 writeback:0 unstable:0
slab_reclaimable:13324 slab_unreclaimable:44354
mapped:83015 shmem:4858 pagetables:26316 bounce:0
free:2727 free_pcp:1035 free_cma:178
Node 0 active_anon:408244kB inactive_anon:326204kB active_file:236408kB inactive_file:275696kB unevictable:2444kB isolated(anon):0kB isolated(file):256kB mapped:332060kB dirty:252kB writeback:0kB shmem:19432kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:10908kB min:6192kB low:44388kB high:47060kB active_anon:409160kB inactive_anon:325924kB active_file:235820kB inactive_file:276628kB unevictable:2444kB writepending:252kB present:3076096kB managed:2673676kB mlocked:2444kB kernel_stack:62512kB pagetables:105264kB bounce:0kB free_pcp:4140kB local_pcp:40kB free_cma:712kB
lowmem_reserve[]: 0 0
Normal: 505*4kB (H) 357*8kB (H) 201*16kB (H) 65*32kB (H) 1*64kB (H) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 10236kB
138826 total pagecache pages
5460 pages in swap cache
Swap cache stats: add 8273090, delete 8267506, find 1004381/4060142
This is an example of ALLOC_NO_WATERMARKS allocation failure using v4.14
based kernel.
kswapd0: page allocation failure: order:0, mode:0x140000a(GFP_NOIO|__GFP_HIGHMEM|__GFP_MOVABLE), nodemask=(null)
kswapd0 cpuset=/ mems_allowed=0
CPU: 4 PID: 1221 Comm: kswapd0 Not tainted 4.14.113-18770262-userdebug #1
Call trace:
[<0000000000000000>] dump_backtrace+0x0/0x248
[<0000000000000000>] show_stack+0x18/0x20
[<0000000000000000>] __dump_stack+0x20/0x28
[<0000000000000000>] dump_stack+0x68/0x90
[<0000000000000000>] warn_alloc+0x104/0x198
[<0000000000000000>] __alloc_pages_nodemask+0xdc0/0xdf0
[<0000000000000000>] zs_malloc+0x148/0x3d0
[<0000000000000000>] zram_bvec_rw+0x410/0x798
[<0000000000000000>] zram_rw_page+0x88/0xdc
[<0000000000000000>] bdev_write_page+0x70/0xbc
[<0000000000000000>] __swap_writepage+0x58/0x37c
[<0000000000000000>] swap_writepage+0x40/0x4c
[<0000000000000000>] shrink_page_list+0xc30/0xf48
[<0000000000000000>] shrink_inactive_list+0x2b0/0x61c
[<0000000000000000>] shrink_node_memcg+0x23c/0x618
[<0000000000000000>] shrink_node+0x1c8/0x304
[<0000000000000000>] kswapd+0x680/0x7c4
[<0000000000000000>] kthread+0x110/0x120
[<0000000000000000>] ret_from_fork+0x10/0x18
Mem-Info:
active_anon:111826 inactive_anon:65557 isolated_anon:0\x0a active_file:44260 inactive_file:83422 isolated_file:0\x0a unevictable:4158 dirty:117 writeback:0 unstable:0\x0a slab_reclaimable:13943 slab_unreclaimable:43315\x0a mapped:102511 shmem:3299 pagetables:19566 bounce:0\x0a free:3510 free_pcp:553 free_cma:0
Node 0 active_anon:447304kB inactive_anon:262228kB active_file:177040kB inactive_file:333688kB unevictable:16632kB isolated(anon):0kB isolated(file):0kB mapped:410044kB d irty:468kB writeback:0kB shmem:13196kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:14040kB min:7440kB low:94500kB high:98136kB reserved_highatomic:32768KB active_anon:447336kB inactive_anon:261668kB active_file:177572kB inactive_file:333768k B unevictable:16632kB writepending:480kB present:4081664kB managed:3637088kB mlocked:16632kB kernel_stack:47072kB pagetables:78264kB bounce:0kB free_pcp:2280kB local_pcp:720kB free_cma:0kB [ 4738.329607] lowmem_reserve[]: 0 0
Normal: 860*4kB (H) 453*8kB (H) 180*16kB (H) 26*32kB (H) 34*64kB (H) 6*128kB (H) 2*256kB (H) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 14232kB
This is trace log which shows GFP_HIGHUSER consumes free pages right
before ALLOC_NO_WATERMARKS.
<...>-22275 [006] .... 889.213383: mm_page_alloc: page=00000000d2be5665 pfn=970744 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213385: mm_page_alloc: page=000000004b2335c2 pfn=970745 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213387: mm_page_alloc: page=00000000017272e1 pfn=970278 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213389: mm_page_alloc: page=00000000c4be79fb pfn=970279 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213391: mm_page_alloc: page=00000000f8a51d4f pfn=970260 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213393: mm_page_alloc: page=000000006ba8f5ac pfn=970261 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213395: mm_page_alloc: page=00000000819f1cd3 pfn=970196 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213396: mm_page_alloc: page=00000000f6b72a64 pfn=970197 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
kswapd0-1207 [005] ...1 889.213398: mm_page_alloc: page= (null) pfn=0 order=0 migratetype=1 nr_free=3650 gfp_flags=GFP_NOWAIT|__GFP_HIGHMEM|__GFP_NOWARN|__GFP_MOVABLE
[jaewon31.kim@samsung.com: remove redundant code for high-order]
Link: http://lkml.kernel.org/r/20200623035242.27232-1-jaewon31.kim@samsung.com
Reported-by: Yong-Taek Lee <ytk.lee@samsung.com>
Suggested-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yong-Taek Lee <ytk.lee@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Link: http://lkml.kernel.org/r/20200619235958.11283-1-jaewon31.kim@samsung.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:20 +00:00
|
|
|
long free_pages;
|
2018-05-23 01:18:21 +00:00
|
|
|
|
page_alloc: consider highatomic reserve in watermark fast
zone_watermark_fast was introduced by commit 48ee5f3696f6 ("mm,
page_alloc: shortcut watermark checks for order-0 pages"). The commit
simply checks if free pages is bigger than watermark without additional
calculation such like reducing watermark.
It considered free cma pages but it did not consider highatomic reserved.
This may incur exhaustion of free pages except high order atomic free
pages.
Assume that reserved_highatomic pageblock is bigger than watermark min,
and there are only few free pages except high order atomic free. Because
zone_watermark_fast passes the allocation without considering high order
atomic free, normal reclaimable allocation like GFP_HIGHUSER will consume
all the free pages. Then finally order-0 atomic allocation may fail on
allocation.
This means watermark min is not protected against non-atomic allocation.
The order-0 atomic allocation with ALLOC_HARDER unwantedly can be failed.
Additionally the __GFP_MEMALLOC allocation with ALLOC_NO_WATERMARKS also
can be failed.
To avoid the problem, zone_watermark_fast should consider highatomic
reserve. If the actual size of high atomic free is counted accurately
like cma free, we may use it. On this patch just use
nr_reserved_highatomic. Additionally introduce
__zone_watermark_unusable_free to factor out common parts between
zone_watermark_fast and __zone_watermark_ok.
This is an example of ALLOC_HARDER allocation failure using v4.19 based
kernel.
Binder:9343_3: page allocation failure: order:0, mode:0x480020(GFP_ATOMIC), nodemask=(null)
Call trace:
[<ffffff8008f40f8c>] dump_stack+0xb8/0xf0
[<ffffff8008223320>] warn_alloc+0xd8/0x12c
[<ffffff80082245e4>] __alloc_pages_nodemask+0x120c/0x1250
[<ffffff800827f6e8>] new_slab+0x128/0x604
[<ffffff800827b0cc>] ___slab_alloc+0x508/0x670
[<ffffff800827ba00>] __kmalloc+0x2f8/0x310
[<ffffff80084ac3e0>] context_struct_to_string+0x104/0x1cc
[<ffffff80084ad8fc>] security_sid_to_context_core+0x74/0x144
[<ffffff80084ad880>] security_sid_to_context+0x10/0x18
[<ffffff800849bd80>] selinux_secid_to_secctx+0x20/0x28
[<ffffff800849109c>] security_secid_to_secctx+0x3c/0x70
[<ffffff8008bfe118>] binder_transaction+0xe68/0x454c
Mem-Info:
active_anon:102061 inactive_anon:81551 isolated_anon:0
active_file:59102 inactive_file:68924 isolated_file:64
unevictable:611 dirty:63 writeback:0 unstable:0
slab_reclaimable:13324 slab_unreclaimable:44354
mapped:83015 shmem:4858 pagetables:26316 bounce:0
free:2727 free_pcp:1035 free_cma:178
Node 0 active_anon:408244kB inactive_anon:326204kB active_file:236408kB inactive_file:275696kB unevictable:2444kB isolated(anon):0kB isolated(file):256kB mapped:332060kB dirty:252kB writeback:0kB shmem:19432kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:10908kB min:6192kB low:44388kB high:47060kB active_anon:409160kB inactive_anon:325924kB active_file:235820kB inactive_file:276628kB unevictable:2444kB writepending:252kB present:3076096kB managed:2673676kB mlocked:2444kB kernel_stack:62512kB pagetables:105264kB bounce:0kB free_pcp:4140kB local_pcp:40kB free_cma:712kB
lowmem_reserve[]: 0 0
Normal: 505*4kB (H) 357*8kB (H) 201*16kB (H) 65*32kB (H) 1*64kB (H) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 10236kB
138826 total pagecache pages
5460 pages in swap cache
Swap cache stats: add 8273090, delete 8267506, find 1004381/4060142
This is an example of ALLOC_NO_WATERMARKS allocation failure using v4.14
based kernel.
kswapd0: page allocation failure: order:0, mode:0x140000a(GFP_NOIO|__GFP_HIGHMEM|__GFP_MOVABLE), nodemask=(null)
kswapd0 cpuset=/ mems_allowed=0
CPU: 4 PID: 1221 Comm: kswapd0 Not tainted 4.14.113-18770262-userdebug #1
Call trace:
[<0000000000000000>] dump_backtrace+0x0/0x248
[<0000000000000000>] show_stack+0x18/0x20
[<0000000000000000>] __dump_stack+0x20/0x28
[<0000000000000000>] dump_stack+0x68/0x90
[<0000000000000000>] warn_alloc+0x104/0x198
[<0000000000000000>] __alloc_pages_nodemask+0xdc0/0xdf0
[<0000000000000000>] zs_malloc+0x148/0x3d0
[<0000000000000000>] zram_bvec_rw+0x410/0x798
[<0000000000000000>] zram_rw_page+0x88/0xdc
[<0000000000000000>] bdev_write_page+0x70/0xbc
[<0000000000000000>] __swap_writepage+0x58/0x37c
[<0000000000000000>] swap_writepage+0x40/0x4c
[<0000000000000000>] shrink_page_list+0xc30/0xf48
[<0000000000000000>] shrink_inactive_list+0x2b0/0x61c
[<0000000000000000>] shrink_node_memcg+0x23c/0x618
[<0000000000000000>] shrink_node+0x1c8/0x304
[<0000000000000000>] kswapd+0x680/0x7c4
[<0000000000000000>] kthread+0x110/0x120
[<0000000000000000>] ret_from_fork+0x10/0x18
Mem-Info:
active_anon:111826 inactive_anon:65557 isolated_anon:0\x0a active_file:44260 inactive_file:83422 isolated_file:0\x0a unevictable:4158 dirty:117 writeback:0 unstable:0\x0a slab_reclaimable:13943 slab_unreclaimable:43315\x0a mapped:102511 shmem:3299 pagetables:19566 bounce:0\x0a free:3510 free_pcp:553 free_cma:0
Node 0 active_anon:447304kB inactive_anon:262228kB active_file:177040kB inactive_file:333688kB unevictable:16632kB isolated(anon):0kB isolated(file):0kB mapped:410044kB d irty:468kB writeback:0kB shmem:13196kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:14040kB min:7440kB low:94500kB high:98136kB reserved_highatomic:32768KB active_anon:447336kB inactive_anon:261668kB active_file:177572kB inactive_file:333768k B unevictable:16632kB writepending:480kB present:4081664kB managed:3637088kB mlocked:16632kB kernel_stack:47072kB pagetables:78264kB bounce:0kB free_pcp:2280kB local_pcp:720kB free_cma:0kB [ 4738.329607] lowmem_reserve[]: 0 0
Normal: 860*4kB (H) 453*8kB (H) 180*16kB (H) 26*32kB (H) 34*64kB (H) 6*128kB (H) 2*256kB (H) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 14232kB
This is trace log which shows GFP_HIGHUSER consumes free pages right
before ALLOC_NO_WATERMARKS.
<...>-22275 [006] .... 889.213383: mm_page_alloc: page=00000000d2be5665 pfn=970744 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213385: mm_page_alloc: page=000000004b2335c2 pfn=970745 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213387: mm_page_alloc: page=00000000017272e1 pfn=970278 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213389: mm_page_alloc: page=00000000c4be79fb pfn=970279 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213391: mm_page_alloc: page=00000000f8a51d4f pfn=970260 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213393: mm_page_alloc: page=000000006ba8f5ac pfn=970261 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213395: mm_page_alloc: page=00000000819f1cd3 pfn=970196 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213396: mm_page_alloc: page=00000000f6b72a64 pfn=970197 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
kswapd0-1207 [005] ...1 889.213398: mm_page_alloc: page= (null) pfn=0 order=0 migratetype=1 nr_free=3650 gfp_flags=GFP_NOWAIT|__GFP_HIGHMEM|__GFP_NOWARN|__GFP_MOVABLE
[jaewon31.kim@samsung.com: remove redundant code for high-order]
Link: http://lkml.kernel.org/r/20200623035242.27232-1-jaewon31.kim@samsung.com
Reported-by: Yong-Taek Lee <ytk.lee@samsung.com>
Suggested-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yong-Taek Lee <ytk.lee@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Link: http://lkml.kernel.org/r/20200619235958.11283-1-jaewon31.kim@samsung.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:20 +00:00
|
|
|
free_pages = zone_page_state(z, NR_FREE_PAGES);
|
2016-05-20 00:14:07 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Fast check for order-0 only. If this fails then the reserves
|
page_alloc: consider highatomic reserve in watermark fast
zone_watermark_fast was introduced by commit 48ee5f3696f6 ("mm,
page_alloc: shortcut watermark checks for order-0 pages"). The commit
simply checks if free pages is bigger than watermark without additional
calculation such like reducing watermark.
It considered free cma pages but it did not consider highatomic reserved.
This may incur exhaustion of free pages except high order atomic free
pages.
Assume that reserved_highatomic pageblock is bigger than watermark min,
and there are only few free pages except high order atomic free. Because
zone_watermark_fast passes the allocation without considering high order
atomic free, normal reclaimable allocation like GFP_HIGHUSER will consume
all the free pages. Then finally order-0 atomic allocation may fail on
allocation.
This means watermark min is not protected against non-atomic allocation.
The order-0 atomic allocation with ALLOC_HARDER unwantedly can be failed.
Additionally the __GFP_MEMALLOC allocation with ALLOC_NO_WATERMARKS also
can be failed.
To avoid the problem, zone_watermark_fast should consider highatomic
reserve. If the actual size of high atomic free is counted accurately
like cma free, we may use it. On this patch just use
nr_reserved_highatomic. Additionally introduce
__zone_watermark_unusable_free to factor out common parts between
zone_watermark_fast and __zone_watermark_ok.
This is an example of ALLOC_HARDER allocation failure using v4.19 based
kernel.
Binder:9343_3: page allocation failure: order:0, mode:0x480020(GFP_ATOMIC), nodemask=(null)
Call trace:
[<ffffff8008f40f8c>] dump_stack+0xb8/0xf0
[<ffffff8008223320>] warn_alloc+0xd8/0x12c
[<ffffff80082245e4>] __alloc_pages_nodemask+0x120c/0x1250
[<ffffff800827f6e8>] new_slab+0x128/0x604
[<ffffff800827b0cc>] ___slab_alloc+0x508/0x670
[<ffffff800827ba00>] __kmalloc+0x2f8/0x310
[<ffffff80084ac3e0>] context_struct_to_string+0x104/0x1cc
[<ffffff80084ad8fc>] security_sid_to_context_core+0x74/0x144
[<ffffff80084ad880>] security_sid_to_context+0x10/0x18
[<ffffff800849bd80>] selinux_secid_to_secctx+0x20/0x28
[<ffffff800849109c>] security_secid_to_secctx+0x3c/0x70
[<ffffff8008bfe118>] binder_transaction+0xe68/0x454c
Mem-Info:
active_anon:102061 inactive_anon:81551 isolated_anon:0
active_file:59102 inactive_file:68924 isolated_file:64
unevictable:611 dirty:63 writeback:0 unstable:0
slab_reclaimable:13324 slab_unreclaimable:44354
mapped:83015 shmem:4858 pagetables:26316 bounce:0
free:2727 free_pcp:1035 free_cma:178
Node 0 active_anon:408244kB inactive_anon:326204kB active_file:236408kB inactive_file:275696kB unevictable:2444kB isolated(anon):0kB isolated(file):256kB mapped:332060kB dirty:252kB writeback:0kB shmem:19432kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:10908kB min:6192kB low:44388kB high:47060kB active_anon:409160kB inactive_anon:325924kB active_file:235820kB inactive_file:276628kB unevictable:2444kB writepending:252kB present:3076096kB managed:2673676kB mlocked:2444kB kernel_stack:62512kB pagetables:105264kB bounce:0kB free_pcp:4140kB local_pcp:40kB free_cma:712kB
lowmem_reserve[]: 0 0
Normal: 505*4kB (H) 357*8kB (H) 201*16kB (H) 65*32kB (H) 1*64kB (H) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 10236kB
138826 total pagecache pages
5460 pages in swap cache
Swap cache stats: add 8273090, delete 8267506, find 1004381/4060142
This is an example of ALLOC_NO_WATERMARKS allocation failure using v4.14
based kernel.
kswapd0: page allocation failure: order:0, mode:0x140000a(GFP_NOIO|__GFP_HIGHMEM|__GFP_MOVABLE), nodemask=(null)
kswapd0 cpuset=/ mems_allowed=0
CPU: 4 PID: 1221 Comm: kswapd0 Not tainted 4.14.113-18770262-userdebug #1
Call trace:
[<0000000000000000>] dump_backtrace+0x0/0x248
[<0000000000000000>] show_stack+0x18/0x20
[<0000000000000000>] __dump_stack+0x20/0x28
[<0000000000000000>] dump_stack+0x68/0x90
[<0000000000000000>] warn_alloc+0x104/0x198
[<0000000000000000>] __alloc_pages_nodemask+0xdc0/0xdf0
[<0000000000000000>] zs_malloc+0x148/0x3d0
[<0000000000000000>] zram_bvec_rw+0x410/0x798
[<0000000000000000>] zram_rw_page+0x88/0xdc
[<0000000000000000>] bdev_write_page+0x70/0xbc
[<0000000000000000>] __swap_writepage+0x58/0x37c
[<0000000000000000>] swap_writepage+0x40/0x4c
[<0000000000000000>] shrink_page_list+0xc30/0xf48
[<0000000000000000>] shrink_inactive_list+0x2b0/0x61c
[<0000000000000000>] shrink_node_memcg+0x23c/0x618
[<0000000000000000>] shrink_node+0x1c8/0x304
[<0000000000000000>] kswapd+0x680/0x7c4
[<0000000000000000>] kthread+0x110/0x120
[<0000000000000000>] ret_from_fork+0x10/0x18
Mem-Info:
active_anon:111826 inactive_anon:65557 isolated_anon:0\x0a active_file:44260 inactive_file:83422 isolated_file:0\x0a unevictable:4158 dirty:117 writeback:0 unstable:0\x0a slab_reclaimable:13943 slab_unreclaimable:43315\x0a mapped:102511 shmem:3299 pagetables:19566 bounce:0\x0a free:3510 free_pcp:553 free_cma:0
Node 0 active_anon:447304kB inactive_anon:262228kB active_file:177040kB inactive_file:333688kB unevictable:16632kB isolated(anon):0kB isolated(file):0kB mapped:410044kB d irty:468kB writeback:0kB shmem:13196kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:14040kB min:7440kB low:94500kB high:98136kB reserved_highatomic:32768KB active_anon:447336kB inactive_anon:261668kB active_file:177572kB inactive_file:333768k B unevictable:16632kB writepending:480kB present:4081664kB managed:3637088kB mlocked:16632kB kernel_stack:47072kB pagetables:78264kB bounce:0kB free_pcp:2280kB local_pcp:720kB free_cma:0kB [ 4738.329607] lowmem_reserve[]: 0 0
Normal: 860*4kB (H) 453*8kB (H) 180*16kB (H) 26*32kB (H) 34*64kB (H) 6*128kB (H) 2*256kB (H) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 14232kB
This is trace log which shows GFP_HIGHUSER consumes free pages right
before ALLOC_NO_WATERMARKS.
<...>-22275 [006] .... 889.213383: mm_page_alloc: page=00000000d2be5665 pfn=970744 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213385: mm_page_alloc: page=000000004b2335c2 pfn=970745 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213387: mm_page_alloc: page=00000000017272e1 pfn=970278 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213389: mm_page_alloc: page=00000000c4be79fb pfn=970279 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213391: mm_page_alloc: page=00000000f8a51d4f pfn=970260 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213393: mm_page_alloc: page=000000006ba8f5ac pfn=970261 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213395: mm_page_alloc: page=00000000819f1cd3 pfn=970196 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213396: mm_page_alloc: page=00000000f6b72a64 pfn=970197 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
kswapd0-1207 [005] ...1 889.213398: mm_page_alloc: page= (null) pfn=0 order=0 migratetype=1 nr_free=3650 gfp_flags=GFP_NOWAIT|__GFP_HIGHMEM|__GFP_NOWARN|__GFP_MOVABLE
[jaewon31.kim@samsung.com: remove redundant code for high-order]
Link: http://lkml.kernel.org/r/20200623035242.27232-1-jaewon31.kim@samsung.com
Reported-by: Yong-Taek Lee <ytk.lee@samsung.com>
Suggested-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yong-Taek Lee <ytk.lee@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Link: http://lkml.kernel.org/r/20200619235958.11283-1-jaewon31.kim@samsung.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:20 +00:00
|
|
|
* need to be calculated.
|
2016-05-20 00:14:07 +00:00
|
|
|
*/
|
page_alloc: consider highatomic reserve in watermark fast
zone_watermark_fast was introduced by commit 48ee5f3696f6 ("mm,
page_alloc: shortcut watermark checks for order-0 pages"). The commit
simply checks if free pages is bigger than watermark without additional
calculation such like reducing watermark.
It considered free cma pages but it did not consider highatomic reserved.
This may incur exhaustion of free pages except high order atomic free
pages.
Assume that reserved_highatomic pageblock is bigger than watermark min,
and there are only few free pages except high order atomic free. Because
zone_watermark_fast passes the allocation without considering high order
atomic free, normal reclaimable allocation like GFP_HIGHUSER will consume
all the free pages. Then finally order-0 atomic allocation may fail on
allocation.
This means watermark min is not protected against non-atomic allocation.
The order-0 atomic allocation with ALLOC_HARDER unwantedly can be failed.
Additionally the __GFP_MEMALLOC allocation with ALLOC_NO_WATERMARKS also
can be failed.
To avoid the problem, zone_watermark_fast should consider highatomic
reserve. If the actual size of high atomic free is counted accurately
like cma free, we may use it. On this patch just use
nr_reserved_highatomic. Additionally introduce
__zone_watermark_unusable_free to factor out common parts between
zone_watermark_fast and __zone_watermark_ok.
This is an example of ALLOC_HARDER allocation failure using v4.19 based
kernel.
Binder:9343_3: page allocation failure: order:0, mode:0x480020(GFP_ATOMIC), nodemask=(null)
Call trace:
[<ffffff8008f40f8c>] dump_stack+0xb8/0xf0
[<ffffff8008223320>] warn_alloc+0xd8/0x12c
[<ffffff80082245e4>] __alloc_pages_nodemask+0x120c/0x1250
[<ffffff800827f6e8>] new_slab+0x128/0x604
[<ffffff800827b0cc>] ___slab_alloc+0x508/0x670
[<ffffff800827ba00>] __kmalloc+0x2f8/0x310
[<ffffff80084ac3e0>] context_struct_to_string+0x104/0x1cc
[<ffffff80084ad8fc>] security_sid_to_context_core+0x74/0x144
[<ffffff80084ad880>] security_sid_to_context+0x10/0x18
[<ffffff800849bd80>] selinux_secid_to_secctx+0x20/0x28
[<ffffff800849109c>] security_secid_to_secctx+0x3c/0x70
[<ffffff8008bfe118>] binder_transaction+0xe68/0x454c
Mem-Info:
active_anon:102061 inactive_anon:81551 isolated_anon:0
active_file:59102 inactive_file:68924 isolated_file:64
unevictable:611 dirty:63 writeback:0 unstable:0
slab_reclaimable:13324 slab_unreclaimable:44354
mapped:83015 shmem:4858 pagetables:26316 bounce:0
free:2727 free_pcp:1035 free_cma:178
Node 0 active_anon:408244kB inactive_anon:326204kB active_file:236408kB inactive_file:275696kB unevictable:2444kB isolated(anon):0kB isolated(file):256kB mapped:332060kB dirty:252kB writeback:0kB shmem:19432kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:10908kB min:6192kB low:44388kB high:47060kB active_anon:409160kB inactive_anon:325924kB active_file:235820kB inactive_file:276628kB unevictable:2444kB writepending:252kB present:3076096kB managed:2673676kB mlocked:2444kB kernel_stack:62512kB pagetables:105264kB bounce:0kB free_pcp:4140kB local_pcp:40kB free_cma:712kB
lowmem_reserve[]: 0 0
Normal: 505*4kB (H) 357*8kB (H) 201*16kB (H) 65*32kB (H) 1*64kB (H) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 10236kB
138826 total pagecache pages
5460 pages in swap cache
Swap cache stats: add 8273090, delete 8267506, find 1004381/4060142
This is an example of ALLOC_NO_WATERMARKS allocation failure using v4.14
based kernel.
kswapd0: page allocation failure: order:0, mode:0x140000a(GFP_NOIO|__GFP_HIGHMEM|__GFP_MOVABLE), nodemask=(null)
kswapd0 cpuset=/ mems_allowed=0
CPU: 4 PID: 1221 Comm: kswapd0 Not tainted 4.14.113-18770262-userdebug #1
Call trace:
[<0000000000000000>] dump_backtrace+0x0/0x248
[<0000000000000000>] show_stack+0x18/0x20
[<0000000000000000>] __dump_stack+0x20/0x28
[<0000000000000000>] dump_stack+0x68/0x90
[<0000000000000000>] warn_alloc+0x104/0x198
[<0000000000000000>] __alloc_pages_nodemask+0xdc0/0xdf0
[<0000000000000000>] zs_malloc+0x148/0x3d0
[<0000000000000000>] zram_bvec_rw+0x410/0x798
[<0000000000000000>] zram_rw_page+0x88/0xdc
[<0000000000000000>] bdev_write_page+0x70/0xbc
[<0000000000000000>] __swap_writepage+0x58/0x37c
[<0000000000000000>] swap_writepage+0x40/0x4c
[<0000000000000000>] shrink_page_list+0xc30/0xf48
[<0000000000000000>] shrink_inactive_list+0x2b0/0x61c
[<0000000000000000>] shrink_node_memcg+0x23c/0x618
[<0000000000000000>] shrink_node+0x1c8/0x304
[<0000000000000000>] kswapd+0x680/0x7c4
[<0000000000000000>] kthread+0x110/0x120
[<0000000000000000>] ret_from_fork+0x10/0x18
Mem-Info:
active_anon:111826 inactive_anon:65557 isolated_anon:0\x0a active_file:44260 inactive_file:83422 isolated_file:0\x0a unevictable:4158 dirty:117 writeback:0 unstable:0\x0a slab_reclaimable:13943 slab_unreclaimable:43315\x0a mapped:102511 shmem:3299 pagetables:19566 bounce:0\x0a free:3510 free_pcp:553 free_cma:0
Node 0 active_anon:447304kB inactive_anon:262228kB active_file:177040kB inactive_file:333688kB unevictable:16632kB isolated(anon):0kB isolated(file):0kB mapped:410044kB d irty:468kB writeback:0kB shmem:13196kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
Normal free:14040kB min:7440kB low:94500kB high:98136kB reserved_highatomic:32768KB active_anon:447336kB inactive_anon:261668kB active_file:177572kB inactive_file:333768k B unevictable:16632kB writepending:480kB present:4081664kB managed:3637088kB mlocked:16632kB kernel_stack:47072kB pagetables:78264kB bounce:0kB free_pcp:2280kB local_pcp:720kB free_cma:0kB [ 4738.329607] lowmem_reserve[]: 0 0
Normal: 860*4kB (H) 453*8kB (H) 180*16kB (H) 26*32kB (H) 34*64kB (H) 6*128kB (H) 2*256kB (H) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 14232kB
This is trace log which shows GFP_HIGHUSER consumes free pages right
before ALLOC_NO_WATERMARKS.
<...>-22275 [006] .... 889.213383: mm_page_alloc: page=00000000d2be5665 pfn=970744 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213385: mm_page_alloc: page=000000004b2335c2 pfn=970745 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213387: mm_page_alloc: page=00000000017272e1 pfn=970278 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213389: mm_page_alloc: page=00000000c4be79fb pfn=970279 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213391: mm_page_alloc: page=00000000f8a51d4f pfn=970260 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213393: mm_page_alloc: page=000000006ba8f5ac pfn=970261 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213395: mm_page_alloc: page=00000000819f1cd3 pfn=970196 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
<...>-22275 [006] .... 889.213396: mm_page_alloc: page=00000000f6b72a64 pfn=970197 order=0 migratetype=0 nr_free=3650 gfp_flags=GFP_HIGHUSER|__GFP_ZERO
kswapd0-1207 [005] ...1 889.213398: mm_page_alloc: page= (null) pfn=0 order=0 migratetype=1 nr_free=3650 gfp_flags=GFP_NOWAIT|__GFP_HIGHMEM|__GFP_NOWARN|__GFP_MOVABLE
[jaewon31.kim@samsung.com: remove redundant code for high-order]
Link: http://lkml.kernel.org/r/20200623035242.27232-1-jaewon31.kim@samsung.com
Reported-by: Yong-Taek Lee <ytk.lee@samsung.com>
Suggested-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yong-Taek Lee <ytk.lee@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Link: http://lkml.kernel.org/r/20200619235958.11283-1-jaewon31.kim@samsung.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:20 +00:00
|
|
|
if (!order) {
|
|
|
|
long fast_free;
|
|
|
|
|
|
|
|
fast_free = free_pages;
|
|
|
|
fast_free -= __zone_watermark_unusable_free(z, 0, alloc_flags);
|
|
|
|
if (fast_free > mark + z->lowmem_reserve[highest_zoneidx])
|
|
|
|
return true;
|
|
|
|
}
|
2016-05-20 00:14:07 +00:00
|
|
|
|
mm, page_alloc: skip ->waternark_boost for atomic order-0 allocations
When boosting is enabled, it is observed that rate of atomic order-0
allocation failures are high due to the fact that free levels in the
system are checked with ->watermark_boost offset. This is not a problem
for sleepable allocations but for atomic allocations which looks like
regression.
This problem is seen frequently on system setup of Android kernel running
on Snapdragon hardware with 4GB RAM size. When no extfrag event occurred
in the system, ->watermark_boost factor is zero, thus the watermark
configurations in the system are:
_watermark = (
[WMARK_MIN] = 1272, --> ~5MB
[WMARK_LOW] = 9067, --> ~36MB
[WMARK_HIGH] = 9385), --> ~38MB
watermark_boost = 0
After launching some memory hungry applications in Android which can cause
extfrag events in the system to an extent that ->watermark_boost can be
set to max i.e. default boost factor makes it to 150% of high watermark.
_watermark = (
[WMARK_MIN] = 1272, --> ~5MB
[WMARK_LOW] = 9067, --> ~36MB
[WMARK_HIGH] = 9385), --> ~38MB
watermark_boost = 14077, -->~57MB
With default system configuration, for an atomic order-0 allocation to
succeed, having free memory of ~2MB will suffice. But boosting makes the
min_wmark to ~61MB thus for an atomic order-0 allocation to be successful
system should have minimum of ~23MB of free memory(from calculations of
zone_watermark_ok(), min = 3/4(min/2)). But failures are observed despite
system is having ~20MB of free memory. In the testing, this is
reproducible as early as first 300secs since boot and with furtherlowram
configurations(<2GB) it is observed as early as first 150secs since boot.
These failures can be avoided by excluding the ->watermark_boost in
watermark caluculations for atomic order-0 allocations.
[akpm@linux-foundation.org: fix comment grammar, reflow comment]
[charante@codeaurora.org: fix suggested by Mel Gorman]
Link: http://lkml.kernel.org/r/31556793-57b1-1c21-1a9d-22674d9bd938@codeaurora.org
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Link: http://lkml.kernel.org/r/1589882284-21010-1-git-send-email-charante@codeaurora.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:24 +00:00
|
|
|
if (__zone_watermark_ok(z, order, mark, highest_zoneidx, alloc_flags,
|
|
|
|
free_pages))
|
|
|
|
return true;
|
|
|
|
/*
|
|
|
|
* Ignore watermark boosting for GFP_ATOMIC order-0 allocations
|
|
|
|
* when checking the min watermark. The min watermark is the
|
|
|
|
* point where boosting is ignored so that kswapd is woken up
|
|
|
|
* when below the low watermark.
|
|
|
|
*/
|
|
|
|
if (unlikely(!order && (gfp_mask & __GFP_ATOMIC) && z->watermark_boost
|
|
|
|
&& ((alloc_flags & ALLOC_WMARK_MASK) == WMARK_MIN))) {
|
|
|
|
mark = z->_watermark[WMARK_MIN];
|
|
|
|
return __zone_watermark_ok(z, order, mark, highest_zoneidx,
|
|
|
|
alloc_flags, free_pages);
|
|
|
|
}
|
|
|
|
|
|
|
|
return false;
|
2016-05-20 00:14:07 +00:00
|
|
|
}
|
|
|
|
|
2014-06-04 23:10:21 +00:00
|
|
|
bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
|
2020-06-03 22:59:01 +00:00
|
|
|
unsigned long mark, int highest_zoneidx)
|
mm: page allocator: adjust the per-cpu counter threshold when memory is low
Commit aa45484 ("calculate a better estimate of NR_FREE_PAGES when memory
is low") noted that watermarks were based on the vmstat NR_FREE_PAGES. To
avoid synchronization overhead, these counters are maintained on a per-cpu
basis and drained both periodically and when a threshold is above a
threshold. On large CPU systems, the difference between the estimate and
real value of NR_FREE_PAGES can be very high. The system can get into a
case where pages are allocated far below the min watermark potentially
causing livelock issues. The commit solved the problem by taking a better
reading of NR_FREE_PAGES when memory was low.
Unfortately, as reported by Shaohua Li this accurate reading can consume a
large amount of CPU time on systems with many sockets due to cache line
bouncing. This patch takes a different approach. For large machines
where counter drift might be unsafe and while kswapd is awake, the per-cpu
thresholds for the target pgdat are reduced to limit the level of drift to
what should be a safe level. This incurs a performance penalty in heavy
memory pressure by a factor that depends on the workload and the machine
but the machine should function correctly without accidentally exhausting
all memory on a node. There is an additional cost when kswapd wakes and
sleeps but the event is not expected to be frequent - in Shaohua's test
case, there was one recorded sleep and wake event at least.
To ensure that kswapd wakes up, a safe version of zone_watermark_ok() is
introduced that takes a more accurate reading of NR_FREE_PAGES when called
from wakeup_kswapd, when deciding whether it is really safe to go back to
sleep in sleeping_prematurely() and when deciding if a zone is really
balanced or not in balance_pgdat(). We are still using an expensive
function but limiting how often it is called.
When the test case is reproduced, the time spent in the watermark
functions is reduced. The following report is on the percentage of time
spent cumulatively spent in the functions zone_nr_free_pages(),
zone_watermark_ok(), __zone_watermark_ok(), zone_watermark_ok_safe(),
zone_page_state_snapshot(), zone_page_state().
vanilla 11.6615%
disable-threshold 0.2584%
David said:
: We had to pull aa454840 "mm: page allocator: calculate a better estimate
: of NR_FREE_PAGES when memory is low and kswapd is awake" from 2.6.36
: internally because tests showed that it would cause the machine to stall
: as the result of heavy kswapd activity. I merged it back with this fix as
: it is pending in the -mm tree and it solves the issue we were seeing, so I
: definitely think this should be pushed to -stable (and I would seriously
: consider it for 2.6.37 inclusion even at this late date).
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reported-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Tested-by: Nicolas Bareil <nico@chdir.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: <stable@kernel.org> [2.6.37.1, 2.6.36.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:45:41 +00:00
|
|
|
{
|
|
|
|
long free_pages = zone_page_state(z, NR_FREE_PAGES);
|
|
|
|
|
|
|
|
if (z->percpu_drift_mark && free_pages < z->percpu_drift_mark)
|
|
|
|
free_pages = zone_page_state_snapshot(z, NR_FREE_PAGES);
|
|
|
|
|
2020-06-03 22:59:01 +00:00
|
|
|
return __zone_watermark_ok(z, order, mark, highest_zoneidx, 0,
|
mm: page allocator: adjust the per-cpu counter threshold when memory is low
Commit aa45484 ("calculate a better estimate of NR_FREE_PAGES when memory
is low") noted that watermarks were based on the vmstat NR_FREE_PAGES. To
avoid synchronization overhead, these counters are maintained on a per-cpu
basis and drained both periodically and when a threshold is above a
threshold. On large CPU systems, the difference between the estimate and
real value of NR_FREE_PAGES can be very high. The system can get into a
case where pages are allocated far below the min watermark potentially
causing livelock issues. The commit solved the problem by taking a better
reading of NR_FREE_PAGES when memory was low.
Unfortately, as reported by Shaohua Li this accurate reading can consume a
large amount of CPU time on systems with many sockets due to cache line
bouncing. This patch takes a different approach. For large machines
where counter drift might be unsafe and while kswapd is awake, the per-cpu
thresholds for the target pgdat are reduced to limit the level of drift to
what should be a safe level. This incurs a performance penalty in heavy
memory pressure by a factor that depends on the workload and the machine
but the machine should function correctly without accidentally exhausting
all memory on a node. There is an additional cost when kswapd wakes and
sleeps but the event is not expected to be frequent - in Shaohua's test
case, there was one recorded sleep and wake event at least.
To ensure that kswapd wakes up, a safe version of zone_watermark_ok() is
introduced that takes a more accurate reading of NR_FREE_PAGES when called
from wakeup_kswapd, when deciding whether it is really safe to go back to
sleep in sleeping_prematurely() and when deciding if a zone is really
balanced or not in balance_pgdat(). We are still using an expensive
function but limiting how often it is called.
When the test case is reproduced, the time spent in the watermark
functions is reduced. The following report is on the percentage of time
spent cumulatively spent in the functions zone_nr_free_pages(),
zone_watermark_ok(), __zone_watermark_ok(), zone_watermark_ok_safe(),
zone_page_state_snapshot(), zone_page_state().
vanilla 11.6615%
disable-threshold 0.2584%
David said:
: We had to pull aa454840 "mm: page allocator: calculate a better estimate
: of NR_FREE_PAGES when memory is low and kswapd is awake" from 2.6.36
: internally because tests showed that it would cause the machine to stall
: as the result of heavy kswapd activity. I merged it back with this fix as
: it is pending in the -mm tree and it solves the issue we were seeing, so I
: definitely think this should be pushed to -stable (and I would seriously
: consider it for 2.6.37 inclusion even at this late date).
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reported-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Tested-by: Nicolas Bareil <nico@chdir.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: <stable@kernel.org> [2.6.37.1, 2.6.36.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:45:41 +00:00
|
|
|
free_pages);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
[PATCH] memory page_alloc zonelist caching speedup
Optimize the critical zonelist scanning for free pages in the kernel memory
allocator by caching the zones that were found to be full recently, and
skipping them.
Remembers the zones in a zonelist that were short of free memory in the
last second. And it stashes a zone-to-node table in the zonelist struct,
to optimize that conversion (minimize its cache footprint.)
Recent changes:
This differs in a significant way from a similar patch that I
posted a week ago. Now, instead of having a nodemask_t of
recently full nodes, I have a bitmask of recently full zones.
This solves a problem that last weeks patch had, which on
systems with multiple zones per node (such as DMA zone) would
take seeing any of these zones full as meaning that all zones
on that node were full.
Also I changed names - from "zonelist faster" to "zonelist cache",
as that seemed to better convey what we're doing here - caching
some of the key zonelist state (for faster access.)
See below for some performance benchmark results. After all that
discussion with David on why I didn't need them, I went and got
some ;). I wanted to verify that I had not hurt the normal case
of memory allocation noticeably. At least for my one little
microbenchmark, I found (1) the normal case wasn't affected, and
(2) workloads that forced scanning across multiple nodes for
memory improved up to 10% fewer System CPU cycles and lower
elapsed clock time ('sys' and 'real'). Good. See details, below.
I didn't have the logic in get_page_from_freelist() for various
full nodes and zone reclaim failures correct. That should be
fixed up now - notice the new goto labels zonelist_scan,
this_zone_full, and try_next_zone, in get_page_from_freelist().
There are two reasons I persued this alternative, over some earlier
proposals that would have focused on optimizing the fake numa
emulation case by caching the last useful zone:
1) Contrary to what I said before, we (SGI, on large ia64 sn2 systems)
have seen real customer loads where the cost to scan the zonelist
was a problem, due to many nodes being full of memory before
we got to a node we could use. Or at least, I think we have.
This was related to me by another engineer, based on experiences
from some time past. So this is not guaranteed. Most likely, though.
The following approach should help such real numa systems just as
much as it helps fake numa systems, or any combination thereof.
2) The effort to distinguish fake from real numa, using node_distance,
so that we could cache a fake numa node and optimize choosing
it over equivalent distance fake nodes, while continuing to
properly scan all real nodes in distance order, was going to
require a nasty blob of zonelist and node distance munging.
The following approach has no new dependency on node distances or
zone sorting.
See comment in the patch below for a description of what it actually does.
Technical details of note (or controversy):
- See the use of "zlc_active" and "did_zlc_setup" below, to delay
adding any work for this new mechanism until we've looked at the
first zone in zonelist. I figured the odds of the first zone
having the memory we needed were high enough that we should just
look there, first, then get fancy only if we need to keep looking.
- Some odd hackery was needed to add items to struct zonelist, while
not tripping up the custom zonelists built by the mm/mempolicy.c
code for MPOL_BIND. My usual wordy comments below explain this.
Search for "MPOL_BIND".
- Some per-node data in the struct zonelist is now modified frequently,
with no locking. Multiple CPU cores on a node could hit and mangle
this data. The theory is that this is just performance hint data,
and the memory allocator will work just fine despite any such mangling.
The fields at risk are the struct 'zonelist_cache' fields 'fullzones'
(a bitmask) and 'last_full_zap' (unsigned long jiffies). It should
all be self correcting after at most a one second delay.
- This still does a linear scan of the same lengths as before. All
I've optimized is making the scan faster, not algorithmically
shorter. It is now able to scan a compact array of 'unsigned
short' in the case of many full nodes, so one cache line should
cover quite a few nodes, rather than each node hitting another
one or two new and distinct cache lines.
- If both Andi and Nick don't find this too complicated, I will be
(pleasantly) flabbergasted.
- I removed the comment claiming we only use one cachline's worth of
zonelist. We seem, at least in the fake numa case, to have put the
lie to that claim.
- I pay no attention to the various watermarks and such in this performance
hint. A node could be marked full for one watermark, and then skipped
over when searching for a page using a different watermark. I think
that's actually quite ok, as it will tend to slightly increase the
spreading of memory over other nodes, away from a memory stressed node.
===============
Performance - some benchmark results and analysis:
This benchmark runs a memory hog program that uses multiple
threads to touch alot of memory as quickly as it can.
Multiple runs were made, touching 12, 38, 64 or 90 GBytes out of
the total 96 GBytes on the system, and using 1, 19, 37, or 55
threads (on a 56 CPU system.) System, user and real (elapsed)
timings were recorded for each run, shown in units of seconds,
in the table below.
Two kernels were tested - 2.6.18-mm3 and the same kernel with
this zonelist caching patch added. The table also shows the
percentage improvement the zonelist caching sys time is over
(lower than) the stock *-mm kernel.
number 2.6.18-mm3 zonelist-cache delta (< 0 good) percent
GBs N ------------ -------------- ---------------- systime
mem threads sys user real sys user real sys user real better
12 1 153 24 177 151 24 176 -2 0 -1 1%
12 19 99 22 8 99 22 8 0 0 0 0%
12 37 111 25 6 112 25 6 1 0 0 -0%
12 55 115 25 5 110 23 5 -5 -2 0 4%
38 1 502 74 576 497 73 570 -5 -1 -6 0%
38 19 426 78 48 373 76 39 -53 -2 -9 12%
38 37 544 83 36 547 82 36 3 -1 0 -0%
38 55 501 77 23 511 80 24 10 3 1 -1%
64 1 917 125 1042 890 124 1014 -27 -1 -28 2%
64 19 1118 138 119 965 141 103 -153 3 -16 13%
64 37 1202 151 94 1136 150 81 -66 -1 -13 5%
64 55 1118 141 61 1072 140 58 -46 -1 -3 4%
90 1 1342 177 1519 1275 174 1450 -67 -3 -69 4%
90 19 2392 199 192 2116 189 176 -276 -10 -16 11%
90 37 3313 238 175 2972 225 145 -341 -13 -30 10%
90 55 1948 210 104 1843 213 100 -105 3 -4 5%
Notes:
1) This test ran a memory hog program that started a specified number N of
threads, and had each thread allocate and touch 1/N'th of
the total memory to be used in the test run in a single loop,
writing a constant word to memory, one store every 4096 bytes.
Watching this test during some earlier trial runs, I would see
each of these threads sit down on one CPU and stay there, for
the remainder of the pass, a different CPU for each thread.
2) The 'real' column is not comparable to the 'sys' or 'user' columns.
The 'real' column is seconds wall clock time elapsed, from beginning
to end of that test pass. The 'sys' and 'user' columns are total
CPU seconds spent on that test pass. For a 19 thread test run,
for example, the sum of 'sys' and 'user' could be up to 19 times the
number of 'real' elapsed wall clock seconds.
3) Tests were run on a fresh, single-user boot, to minimize the amount
of memory already in use at the start of the test, and to minimize
the amount of background activity that might interfere.
4) Tests were done on a 56 CPU, 28 Node system with 96 GBytes of RAM.
5) Notice that the 'real' time gets large for the single thread runs, even
though the measured 'sys' and 'user' times are modest. I'm not sure what
that means - probably something to do with it being slow for one thread to
be accessing memory along ways away. Perhaps the fake numa system, running
ostensibly the same workload, would not show this substantial degradation
of 'real' time for one thread on many nodes -- lets hope not.
6) The high thread count passes (one thread per CPU - on 55 of 56 CPUs)
ran quite efficiently, as one might expect. Each pair of threads needed
to allocate and touch the memory on the node the two threads shared, a
pleasantly parallizable workload.
7) The intermediate thread count passes, when asking for alot of memory forcing
them to go to a few neighboring nodes, improved the most with this zonelist
caching patch.
Conclusions:
* This zonelist cache patch probably makes little difference one way or the
other for most workloads on real numa hardware, if those workloads avoid
heavy off node allocations.
* For memory intensive workloads requiring substantial off-node allocations
on real numa hardware, this patch improves both kernel and elapsed timings
up to ten per-cent.
* For fake numa systems, I'm optimistic, but will have to leave that up to
Rohit Seth to actually test (once I get him a 2.6.18 backport.)
Signed-off-by: Paul Jackson <pj@sgi.com>
Cc: Rohit Seth <rohitseth@google.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Cc: David Rientjes <rientjes@cs.washington.edu>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:31:48 +00:00
|
|
|
#ifdef CONFIG_NUMA
|
2012-10-08 23:33:24 +00:00
|
|
|
static bool zone_allows_reclaim(struct zone *local_zone, struct zone *zone)
|
|
|
|
{
|
mm/page_alloc: fix nodes for reclaim in fast path
When @node_reclaim_node isn't 0, the page allocator tries to reclaim
pages if the amount of free memory in the zones are below the low
watermark. On Power platform, none of NUMA nodes are scanned for page
reclaim because no nodes match the condition in zone_allows_reclaim().
On Power platform, RECLAIM_DISTANCE is set to 10 which is the distance
of Node-A to Node-A. So the preferred node even won't be scanned for
page reclaim.
__alloc_pages_nodemask()
get_page_from_freelist()
zone_allows_reclaim()
Anton proposed the test code as below:
# cat alloc.c
:
int main(int argc, char *argv[])
{
void *p;
unsigned long size;
unsigned long start, end;
start = time(NULL);
size = strtoul(argv[1], NULL, 0);
printf("To allocate %ldGB memory\n", size);
size <<= 30;
p = malloc(size);
assert(p);
memset(p, 0, size);
end = time(NULL);
printf("Used time: %ld seconds\n", end - start);
sleep(3600);
return 0;
}
The system I use for testing has two NUMA nodes. Both have 128GB
memory. In below scnario, the page caches on node#0 should be reclaimed
when it encounters pressure to accommodate request of allocation.
# echo 2 > /proc/sys/vm/zone_reclaim_mode; \
sync; \
echo 3 > /proc/sys/vm/drop_caches; \
# taskset -c 0 cat file.32G > /dev/null; \
grep FilePages /sys/devices/system/node/node0/meminfo
Node 0 FilePages: 33619712 kB
# taskset -c 0 ./alloc 128
# grep FilePages /sys/devices/system/node/node0/meminfo
Node 0 FilePages: 33619840 kB
# grep MemFree /sys/devices/system/node/node0/meminfo
Node 0 MemFree: 186816 kB
With the patch applied, the pagecache on node-0 is reclaimed when its
free memory is running out. It's the expected behaviour.
# echo 2 > /proc/sys/vm/zone_reclaim_mode; \
sync; \
echo 3 > /proc/sys/vm/drop_caches
# taskset -c 0 cat file.32G > /dev/null; \
grep FilePages /sys/devices/system/node/node0/meminfo
Node 0 FilePages: 33605568 kB
# taskset -c 0 ./alloc 128
# grep FilePages /sys/devices/system/node/node0/meminfo
Node 0 FilePages: 1379520 kB
# grep MemFree /sys/devices/system/node/node0/meminfo
Node 0 MemFree: 317120 kB
Fixes: 5f7a75acdb24 ("mm: page_alloc: do not cache reclaim distances")
Link: http://lkml.kernel.org/r/1486532455-29613-1-git-send-email-gwshan@linux.vnet.ibm.com
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Anton Blanchard <anton@samba.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: <stable@vger.kernel.org> [3.16+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-02-24 22:59:33 +00:00
|
|
|
return node_distance(zone_to_nid(local_zone), zone_to_nid(zone)) <=
|
2019-08-08 19:53:01 +00:00
|
|
|
node_reclaim_distance;
|
2012-10-08 23:33:24 +00:00
|
|
|
}
|
[PATCH] memory page_alloc zonelist caching speedup
Optimize the critical zonelist scanning for free pages in the kernel memory
allocator by caching the zones that were found to be full recently, and
skipping them.
Remembers the zones in a zonelist that were short of free memory in the
last second. And it stashes a zone-to-node table in the zonelist struct,
to optimize that conversion (minimize its cache footprint.)
Recent changes:
This differs in a significant way from a similar patch that I
posted a week ago. Now, instead of having a nodemask_t of
recently full nodes, I have a bitmask of recently full zones.
This solves a problem that last weeks patch had, which on
systems with multiple zones per node (such as DMA zone) would
take seeing any of these zones full as meaning that all zones
on that node were full.
Also I changed names - from "zonelist faster" to "zonelist cache",
as that seemed to better convey what we're doing here - caching
some of the key zonelist state (for faster access.)
See below for some performance benchmark results. After all that
discussion with David on why I didn't need them, I went and got
some ;). I wanted to verify that I had not hurt the normal case
of memory allocation noticeably. At least for my one little
microbenchmark, I found (1) the normal case wasn't affected, and
(2) workloads that forced scanning across multiple nodes for
memory improved up to 10% fewer System CPU cycles and lower
elapsed clock time ('sys' and 'real'). Good. See details, below.
I didn't have the logic in get_page_from_freelist() for various
full nodes and zone reclaim failures correct. That should be
fixed up now - notice the new goto labels zonelist_scan,
this_zone_full, and try_next_zone, in get_page_from_freelist().
There are two reasons I persued this alternative, over some earlier
proposals that would have focused on optimizing the fake numa
emulation case by caching the last useful zone:
1) Contrary to what I said before, we (SGI, on large ia64 sn2 systems)
have seen real customer loads where the cost to scan the zonelist
was a problem, due to many nodes being full of memory before
we got to a node we could use. Or at least, I think we have.
This was related to me by another engineer, based on experiences
from some time past. So this is not guaranteed. Most likely, though.
The following approach should help such real numa systems just as
much as it helps fake numa systems, or any combination thereof.
2) The effort to distinguish fake from real numa, using node_distance,
so that we could cache a fake numa node and optimize choosing
it over equivalent distance fake nodes, while continuing to
properly scan all real nodes in distance order, was going to
require a nasty blob of zonelist and node distance munging.
The following approach has no new dependency on node distances or
zone sorting.
See comment in the patch below for a description of what it actually does.
Technical details of note (or controversy):
- See the use of "zlc_active" and "did_zlc_setup" below, to delay
adding any work for this new mechanism until we've looked at the
first zone in zonelist. I figured the odds of the first zone
having the memory we needed were high enough that we should just
look there, first, then get fancy only if we need to keep looking.
- Some odd hackery was needed to add items to struct zonelist, while
not tripping up the custom zonelists built by the mm/mempolicy.c
code for MPOL_BIND. My usual wordy comments below explain this.
Search for "MPOL_BIND".
- Some per-node data in the struct zonelist is now modified frequently,
with no locking. Multiple CPU cores on a node could hit and mangle
this data. The theory is that this is just performance hint data,
and the memory allocator will work just fine despite any such mangling.
The fields at risk are the struct 'zonelist_cache' fields 'fullzones'
(a bitmask) and 'last_full_zap' (unsigned long jiffies). It should
all be self correcting after at most a one second delay.
- This still does a linear scan of the same lengths as before. All
I've optimized is making the scan faster, not algorithmically
shorter. It is now able to scan a compact array of 'unsigned
short' in the case of many full nodes, so one cache line should
cover quite a few nodes, rather than each node hitting another
one or two new and distinct cache lines.
- If both Andi and Nick don't find this too complicated, I will be
(pleasantly) flabbergasted.
- I removed the comment claiming we only use one cachline's worth of
zonelist. We seem, at least in the fake numa case, to have put the
lie to that claim.
- I pay no attention to the various watermarks and such in this performance
hint. A node could be marked full for one watermark, and then skipped
over when searching for a page using a different watermark. I think
that's actually quite ok, as it will tend to slightly increase the
spreading of memory over other nodes, away from a memory stressed node.
===============
Performance - some benchmark results and analysis:
This benchmark runs a memory hog program that uses multiple
threads to touch alot of memory as quickly as it can.
Multiple runs were made, touching 12, 38, 64 or 90 GBytes out of
the total 96 GBytes on the system, and using 1, 19, 37, or 55
threads (on a 56 CPU system.) System, user and real (elapsed)
timings were recorded for each run, shown in units of seconds,
in the table below.
Two kernels were tested - 2.6.18-mm3 and the same kernel with
this zonelist caching patch added. The table also shows the
percentage improvement the zonelist caching sys time is over
(lower than) the stock *-mm kernel.
number 2.6.18-mm3 zonelist-cache delta (< 0 good) percent
GBs N ------------ -------------- ---------------- systime
mem threads sys user real sys user real sys user real better
12 1 153 24 177 151 24 176 -2 0 -1 1%
12 19 99 22 8 99 22 8 0 0 0 0%
12 37 111 25 6 112 25 6 1 0 0 -0%
12 55 115 25 5 110 23 5 -5 -2 0 4%
38 1 502 74 576 497 73 570 -5 -1 -6 0%
38 19 426 78 48 373 76 39 -53 -2 -9 12%
38 37 544 83 36 547 82 36 3 -1 0 -0%
38 55 501 77 23 511 80 24 10 3 1 -1%
64 1 917 125 1042 890 124 1014 -27 -1 -28 2%
64 19 1118 138 119 965 141 103 -153 3 -16 13%
64 37 1202 151 94 1136 150 81 -66 -1 -13 5%
64 55 1118 141 61 1072 140 58 -46 -1 -3 4%
90 1 1342 177 1519 1275 174 1450 -67 -3 -69 4%
90 19 2392 199 192 2116 189 176 -276 -10 -16 11%
90 37 3313 238 175 2972 225 145 -341 -13 -30 10%
90 55 1948 210 104 1843 213 100 -105 3 -4 5%
Notes:
1) This test ran a memory hog program that started a specified number N of
threads, and had each thread allocate and touch 1/N'th of
the total memory to be used in the test run in a single loop,
writing a constant word to memory, one store every 4096 bytes.
Watching this test during some earlier trial runs, I would see
each of these threads sit down on one CPU and stay there, for
the remainder of the pass, a different CPU for each thread.
2) The 'real' column is not comparable to the 'sys' or 'user' columns.
The 'real' column is seconds wall clock time elapsed, from beginning
to end of that test pass. The 'sys' and 'user' columns are total
CPU seconds spent on that test pass. For a 19 thread test run,
for example, the sum of 'sys' and 'user' could be up to 19 times the
number of 'real' elapsed wall clock seconds.
3) Tests were run on a fresh, single-user boot, to minimize the amount
of memory already in use at the start of the test, and to minimize
the amount of background activity that might interfere.
4) Tests were done on a 56 CPU, 28 Node system with 96 GBytes of RAM.
5) Notice that the 'real' time gets large for the single thread runs, even
though the measured 'sys' and 'user' times are modest. I'm not sure what
that means - probably something to do with it being slow for one thread to
be accessing memory along ways away. Perhaps the fake numa system, running
ostensibly the same workload, would not show this substantial degradation
of 'real' time for one thread on many nodes -- lets hope not.
6) The high thread count passes (one thread per CPU - on 55 of 56 CPUs)
ran quite efficiently, as one might expect. Each pair of threads needed
to allocate and touch the memory on the node the two threads shared, a
pleasantly parallizable workload.
7) The intermediate thread count passes, when asking for alot of memory forcing
them to go to a few neighboring nodes, improved the most with this zonelist
caching patch.
Conclusions:
* This zonelist cache patch probably makes little difference one way or the
other for most workloads on real numa hardware, if those workloads avoid
heavy off node allocations.
* For memory intensive workloads requiring substantial off-node allocations
on real numa hardware, this patch improves both kernel and elapsed timings
up to ten per-cent.
* For fake numa systems, I'm optimistic, but will have to leave that up to
Rohit Seth to actually test (once I get him a 2.6.18 backport.)
Signed-off-by: Paul Jackson <pj@sgi.com>
Cc: Rohit Seth <rohitseth@google.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Cc: David Rientjes <rientjes@cs.washington.edu>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:31:48 +00:00
|
|
|
#else /* CONFIG_NUMA */
|
2012-10-08 23:33:24 +00:00
|
|
|
static bool zone_allows_reclaim(struct zone *local_zone, struct zone *zone)
|
|
|
|
{
|
|
|
|
return true;
|
|
|
|
}
|
[PATCH] memory page_alloc zonelist caching speedup
Optimize the critical zonelist scanning for free pages in the kernel memory
allocator by caching the zones that were found to be full recently, and
skipping them.
Remembers the zones in a zonelist that were short of free memory in the
last second. And it stashes a zone-to-node table in the zonelist struct,
to optimize that conversion (minimize its cache footprint.)
Recent changes:
This differs in a significant way from a similar patch that I
posted a week ago. Now, instead of having a nodemask_t of
recently full nodes, I have a bitmask of recently full zones.
This solves a problem that last weeks patch had, which on
systems with multiple zones per node (such as DMA zone) would
take seeing any of these zones full as meaning that all zones
on that node were full.
Also I changed names - from "zonelist faster" to "zonelist cache",
as that seemed to better convey what we're doing here - caching
some of the key zonelist state (for faster access.)
See below for some performance benchmark results. After all that
discussion with David on why I didn't need them, I went and got
some ;). I wanted to verify that I had not hurt the normal case
of memory allocation noticeably. At least for my one little
microbenchmark, I found (1) the normal case wasn't affected, and
(2) workloads that forced scanning across multiple nodes for
memory improved up to 10% fewer System CPU cycles and lower
elapsed clock time ('sys' and 'real'). Good. See details, below.
I didn't have the logic in get_page_from_freelist() for various
full nodes and zone reclaim failures correct. That should be
fixed up now - notice the new goto labels zonelist_scan,
this_zone_full, and try_next_zone, in get_page_from_freelist().
There are two reasons I persued this alternative, over some earlier
proposals that would have focused on optimizing the fake numa
emulation case by caching the last useful zone:
1) Contrary to what I said before, we (SGI, on large ia64 sn2 systems)
have seen real customer loads where the cost to scan the zonelist
was a problem, due to many nodes being full of memory before
we got to a node we could use. Or at least, I think we have.
This was related to me by another engineer, based on experiences
from some time past. So this is not guaranteed. Most likely, though.
The following approach should help such real numa systems just as
much as it helps fake numa systems, or any combination thereof.
2) The effort to distinguish fake from real numa, using node_distance,
so that we could cache a fake numa node and optimize choosing
it over equivalent distance fake nodes, while continuing to
properly scan all real nodes in distance order, was going to
require a nasty blob of zonelist and node distance munging.
The following approach has no new dependency on node distances or
zone sorting.
See comment in the patch below for a description of what it actually does.
Technical details of note (or controversy):
- See the use of "zlc_active" and "did_zlc_setup" below, to delay
adding any work for this new mechanism until we've looked at the
first zone in zonelist. I figured the odds of the first zone
having the memory we needed were high enough that we should just
look there, first, then get fancy only if we need to keep looking.
- Some odd hackery was needed to add items to struct zonelist, while
not tripping up the custom zonelists built by the mm/mempolicy.c
code for MPOL_BIND. My usual wordy comments below explain this.
Search for "MPOL_BIND".
- Some per-node data in the struct zonelist is now modified frequently,
with no locking. Multiple CPU cores on a node could hit and mangle
this data. The theory is that this is just performance hint data,
and the memory allocator will work just fine despite any such mangling.
The fields at risk are the struct 'zonelist_cache' fields 'fullzones'
(a bitmask) and 'last_full_zap' (unsigned long jiffies). It should
all be self correcting after at most a one second delay.
- This still does a linear scan of the same lengths as before. All
I've optimized is making the scan faster, not algorithmically
shorter. It is now able to scan a compact array of 'unsigned
short' in the case of many full nodes, so one cache line should
cover quite a few nodes, rather than each node hitting another
one or two new and distinct cache lines.
- If both Andi and Nick don't find this too complicated, I will be
(pleasantly) flabbergasted.
- I removed the comment claiming we only use one cachline's worth of
zonelist. We seem, at least in the fake numa case, to have put the
lie to that claim.
- I pay no attention to the various watermarks and such in this performance
hint. A node could be marked full for one watermark, and then skipped
over when searching for a page using a different watermark. I think
that's actually quite ok, as it will tend to slightly increase the
spreading of memory over other nodes, away from a memory stressed node.
===============
Performance - some benchmark results and analysis:
This benchmark runs a memory hog program that uses multiple
threads to touch alot of memory as quickly as it can.
Multiple runs were made, touching 12, 38, 64 or 90 GBytes out of
the total 96 GBytes on the system, and using 1, 19, 37, or 55
threads (on a 56 CPU system.) System, user and real (elapsed)
timings were recorded for each run, shown in units of seconds,
in the table below.
Two kernels were tested - 2.6.18-mm3 and the same kernel with
this zonelist caching patch added. The table also shows the
percentage improvement the zonelist caching sys time is over
(lower than) the stock *-mm kernel.
number 2.6.18-mm3 zonelist-cache delta (< 0 good) percent
GBs N ------------ -------------- ---------------- systime
mem threads sys user real sys user real sys user real better
12 1 153 24 177 151 24 176 -2 0 -1 1%
12 19 99 22 8 99 22 8 0 0 0 0%
12 37 111 25 6 112 25 6 1 0 0 -0%
12 55 115 25 5 110 23 5 -5 -2 0 4%
38 1 502 74 576 497 73 570 -5 -1 -6 0%
38 19 426 78 48 373 76 39 -53 -2 -9 12%
38 37 544 83 36 547 82 36 3 -1 0 -0%
38 55 501 77 23 511 80 24 10 3 1 -1%
64 1 917 125 1042 890 124 1014 -27 -1 -28 2%
64 19 1118 138 119 965 141 103 -153 3 -16 13%
64 37 1202 151 94 1136 150 81 -66 -1 -13 5%
64 55 1118 141 61 1072 140 58 -46 -1 -3 4%
90 1 1342 177 1519 1275 174 1450 -67 -3 -69 4%
90 19 2392 199 192 2116 189 176 -276 -10 -16 11%
90 37 3313 238 175 2972 225 145 -341 -13 -30 10%
90 55 1948 210 104 1843 213 100 -105 3 -4 5%
Notes:
1) This test ran a memory hog program that started a specified number N of
threads, and had each thread allocate and touch 1/N'th of
the total memory to be used in the test run in a single loop,
writing a constant word to memory, one store every 4096 bytes.
Watching this test during some earlier trial runs, I would see
each of these threads sit down on one CPU and stay there, for
the remainder of the pass, a different CPU for each thread.
2) The 'real' column is not comparable to the 'sys' or 'user' columns.
The 'real' column is seconds wall clock time elapsed, from beginning
to end of that test pass. The 'sys' and 'user' columns are total
CPU seconds spent on that test pass. For a 19 thread test run,
for example, the sum of 'sys' and 'user' could be up to 19 times the
number of 'real' elapsed wall clock seconds.
3) Tests were run on a fresh, single-user boot, to minimize the amount
of memory already in use at the start of the test, and to minimize
the amount of background activity that might interfere.
4) Tests were done on a 56 CPU, 28 Node system with 96 GBytes of RAM.
5) Notice that the 'real' time gets large for the single thread runs, even
though the measured 'sys' and 'user' times are modest. I'm not sure what
that means - probably something to do with it being slow for one thread to
be accessing memory along ways away. Perhaps the fake numa system, running
ostensibly the same workload, would not show this substantial degradation
of 'real' time for one thread on many nodes -- lets hope not.
6) The high thread count passes (one thread per CPU - on 55 of 56 CPUs)
ran quite efficiently, as one might expect. Each pair of threads needed
to allocate and touch the memory on the node the two threads shared, a
pleasantly parallizable workload.
7) The intermediate thread count passes, when asking for alot of memory forcing
them to go to a few neighboring nodes, improved the most with this zonelist
caching patch.
Conclusions:
* This zonelist cache patch probably makes little difference one way or the
other for most workloads on real numa hardware, if those workloads avoid
heavy off node allocations.
* For memory intensive workloads requiring substantial off-node allocations
on real numa hardware, this patch improves both kernel and elapsed timings
up to ten per-cent.
* For fake numa systems, I'm optimistic, but will have to leave that up to
Rohit Seth to actually test (once I get him a 2.6.18 backport.)
Signed-off-by: Paul Jackson <pj@sgi.com>
Cc: Rohit Seth <rohitseth@google.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Cc: David Rientjes <rientjes@cs.washington.edu>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:31:48 +00:00
|
|
|
#endif /* CONFIG_NUMA */
|
|
|
|
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
/*
|
|
|
|
* The restriction on ZONE_DMA32 as being a suitable zone to use to avoid
|
|
|
|
* fragmentation is subtle. If the preferred zone was HIGHMEM then
|
|
|
|
* premature use of a lower zone may cause lowmem pressure problems that
|
|
|
|
* are worse than fragmentation. If the next zone is ZONE_DMA then it is
|
|
|
|
* probably too small. It only makes sense to spread allocations to avoid
|
|
|
|
* fragmentation between the Normal and DMA32 zones.
|
|
|
|
*/
|
|
|
|
static inline unsigned int
|
2018-12-28 08:35:48 +00:00
|
|
|
alloc_flags_nofragment(struct zone *zone, gfp_t gfp_mask)
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
{
|
2020-04-02 04:09:47 +00:00
|
|
|
unsigned int alloc_flags;
|
2018-12-28 08:35:48 +00:00
|
|
|
|
2020-04-02 04:09:47 +00:00
|
|
|
/*
|
|
|
|
* __GFP_KSWAPD_RECLAIM is assumed to be the same as ALLOC_KSWAPD
|
|
|
|
* to save a branch.
|
|
|
|
*/
|
|
|
|
alloc_flags = (__force int) (gfp_mask & __GFP_KSWAPD_RECLAIM);
|
2018-12-28 08:35:48 +00:00
|
|
|
|
|
|
|
#ifdef CONFIG_ZONE_DMA32
|
2019-04-26 05:23:58 +00:00
|
|
|
if (!zone)
|
|
|
|
return alloc_flags;
|
|
|
|
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
if (zone_idx(zone) != ZONE_NORMAL)
|
2019-04-26 05:24:01 +00:00
|
|
|
return alloc_flags;
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If ZONE_DMA32 exists, assume it is the one after ZONE_NORMAL and
|
|
|
|
* the pointer is within zone->zone_pgdat->node_zones[]. Also assume
|
|
|
|
* on UMA that if Normal is populated then so is DMA32.
|
|
|
|
*/
|
|
|
|
BUILD_BUG_ON(ZONE_NORMAL - ZONE_DMA32 != 1);
|
|
|
|
if (nr_online_nodes > 1 && !populated_zone(--zone))
|
2019-04-26 05:24:01 +00:00
|
|
|
return alloc_flags;
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
|
2019-04-26 05:24:01 +00:00
|
|
|
alloc_flags |= ALLOC_NOFRAGMENT;
|
2018-12-28 08:35:48 +00:00
|
|
|
#endif /* CONFIG_ZONE_DMA32 */
|
|
|
|
return alloc_flags;
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
}
|
|
|
|
|
2020-08-07 06:26:04 +00:00
|
|
|
static inline unsigned int current_alloc_flags(gfp_t gfp_mask,
|
|
|
|
unsigned int alloc_flags)
|
|
|
|
{
|
|
|
|
#ifdef CONFIG_CMA
|
|
|
|
unsigned int pflags = current->flags;
|
|
|
|
|
|
|
|
if (!(pflags & PF_MEMALLOC_NOCMA) &&
|
|
|
|
gfp_migratetype(gfp_mask) == MIGRATE_MOVABLE)
|
|
|
|
alloc_flags |= ALLOC_CMA;
|
|
|
|
|
|
|
|
#endif
|
|
|
|
return alloc_flags;
|
|
|
|
}
|
|
|
|
|
2005-11-14 00:06:43 +00:00
|
|
|
/*
|
2006-12-07 04:31:38 +00:00
|
|
|
* get_page_from_freelist goes through the zonelist trying to allocate
|
2005-11-14 00:06:43 +00:00
|
|
|
* a page.
|
|
|
|
*/
|
|
|
|
static struct page *
|
2015-02-11 23:25:41 +00:00
|
|
|
get_page_from_freelist(gfp_t gfp_mask, unsigned int order, int alloc_flags,
|
|
|
|
const struct alloc_context *ac)
|
2005-06-22 00:14:41 +00:00
|
|
|
{
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
struct zoneref *z;
|
2009-06-16 22:31:59 +00:00
|
|
|
struct zone *zone;
|
2016-07-28 22:46:53 +00:00
|
|
|
struct pglist_data *last_pgdat_dirty_limit = NULL;
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
bool no_fallback;
|
2016-07-28 22:46:53 +00:00
|
|
|
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
retry:
|
2005-11-14 00:06:43 +00:00
|
|
|
/*
|
[PATCH] memory page_alloc zonelist caching speedup
Optimize the critical zonelist scanning for free pages in the kernel memory
allocator by caching the zones that were found to be full recently, and
skipping them.
Remembers the zones in a zonelist that were short of free memory in the
last second. And it stashes a zone-to-node table in the zonelist struct,
to optimize that conversion (minimize its cache footprint.)
Recent changes:
This differs in a significant way from a similar patch that I
posted a week ago. Now, instead of having a nodemask_t of
recently full nodes, I have a bitmask of recently full zones.
This solves a problem that last weeks patch had, which on
systems with multiple zones per node (such as DMA zone) would
take seeing any of these zones full as meaning that all zones
on that node were full.
Also I changed names - from "zonelist faster" to "zonelist cache",
as that seemed to better convey what we're doing here - caching
some of the key zonelist state (for faster access.)
See below for some performance benchmark results. After all that
discussion with David on why I didn't need them, I went and got
some ;). I wanted to verify that I had not hurt the normal case
of memory allocation noticeably. At least for my one little
microbenchmark, I found (1) the normal case wasn't affected, and
(2) workloads that forced scanning across multiple nodes for
memory improved up to 10% fewer System CPU cycles and lower
elapsed clock time ('sys' and 'real'). Good. See details, below.
I didn't have the logic in get_page_from_freelist() for various
full nodes and zone reclaim failures correct. That should be
fixed up now - notice the new goto labels zonelist_scan,
this_zone_full, and try_next_zone, in get_page_from_freelist().
There are two reasons I persued this alternative, over some earlier
proposals that would have focused on optimizing the fake numa
emulation case by caching the last useful zone:
1) Contrary to what I said before, we (SGI, on large ia64 sn2 systems)
have seen real customer loads where the cost to scan the zonelist
was a problem, due to many nodes being full of memory before
we got to a node we could use. Or at least, I think we have.
This was related to me by another engineer, based on experiences
from some time past. So this is not guaranteed. Most likely, though.
The following approach should help such real numa systems just as
much as it helps fake numa systems, or any combination thereof.
2) The effort to distinguish fake from real numa, using node_distance,
so that we could cache a fake numa node and optimize choosing
it over equivalent distance fake nodes, while continuing to
properly scan all real nodes in distance order, was going to
require a nasty blob of zonelist and node distance munging.
The following approach has no new dependency on node distances or
zone sorting.
See comment in the patch below for a description of what it actually does.
Technical details of note (or controversy):
- See the use of "zlc_active" and "did_zlc_setup" below, to delay
adding any work for this new mechanism until we've looked at the
first zone in zonelist. I figured the odds of the first zone
having the memory we needed were high enough that we should just
look there, first, then get fancy only if we need to keep looking.
- Some odd hackery was needed to add items to struct zonelist, while
not tripping up the custom zonelists built by the mm/mempolicy.c
code for MPOL_BIND. My usual wordy comments below explain this.
Search for "MPOL_BIND".
- Some per-node data in the struct zonelist is now modified frequently,
with no locking. Multiple CPU cores on a node could hit and mangle
this data. The theory is that this is just performance hint data,
and the memory allocator will work just fine despite any such mangling.
The fields at risk are the struct 'zonelist_cache' fields 'fullzones'
(a bitmask) and 'last_full_zap' (unsigned long jiffies). It should
all be self correcting after at most a one second delay.
- This still does a linear scan of the same lengths as before. All
I've optimized is making the scan faster, not algorithmically
shorter. It is now able to scan a compact array of 'unsigned
short' in the case of many full nodes, so one cache line should
cover quite a few nodes, rather than each node hitting another
one or two new and distinct cache lines.
- If both Andi and Nick don't find this too complicated, I will be
(pleasantly) flabbergasted.
- I removed the comment claiming we only use one cachline's worth of
zonelist. We seem, at least in the fake numa case, to have put the
lie to that claim.
- I pay no attention to the various watermarks and such in this performance
hint. A node could be marked full for one watermark, and then skipped
over when searching for a page using a different watermark. I think
that's actually quite ok, as it will tend to slightly increase the
spreading of memory over other nodes, away from a memory stressed node.
===============
Performance - some benchmark results and analysis:
This benchmark runs a memory hog program that uses multiple
threads to touch alot of memory as quickly as it can.
Multiple runs were made, touching 12, 38, 64 or 90 GBytes out of
the total 96 GBytes on the system, and using 1, 19, 37, or 55
threads (on a 56 CPU system.) System, user and real (elapsed)
timings were recorded for each run, shown in units of seconds,
in the table below.
Two kernels were tested - 2.6.18-mm3 and the same kernel with
this zonelist caching patch added. The table also shows the
percentage improvement the zonelist caching sys time is over
(lower than) the stock *-mm kernel.
number 2.6.18-mm3 zonelist-cache delta (< 0 good) percent
GBs N ------------ -------------- ---------------- systime
mem threads sys user real sys user real sys user real better
12 1 153 24 177 151 24 176 -2 0 -1 1%
12 19 99 22 8 99 22 8 0 0 0 0%
12 37 111 25 6 112 25 6 1 0 0 -0%
12 55 115 25 5 110 23 5 -5 -2 0 4%
38 1 502 74 576 497 73 570 -5 -1 -6 0%
38 19 426 78 48 373 76 39 -53 -2 -9 12%
38 37 544 83 36 547 82 36 3 -1 0 -0%
38 55 501 77 23 511 80 24 10 3 1 -1%
64 1 917 125 1042 890 124 1014 -27 -1 -28 2%
64 19 1118 138 119 965 141 103 -153 3 -16 13%
64 37 1202 151 94 1136 150 81 -66 -1 -13 5%
64 55 1118 141 61 1072 140 58 -46 -1 -3 4%
90 1 1342 177 1519 1275 174 1450 -67 -3 -69 4%
90 19 2392 199 192 2116 189 176 -276 -10 -16 11%
90 37 3313 238 175 2972 225 145 -341 -13 -30 10%
90 55 1948 210 104 1843 213 100 -105 3 -4 5%
Notes:
1) This test ran a memory hog program that started a specified number N of
threads, and had each thread allocate and touch 1/N'th of
the total memory to be used in the test run in a single loop,
writing a constant word to memory, one store every 4096 bytes.
Watching this test during some earlier trial runs, I would see
each of these threads sit down on one CPU and stay there, for
the remainder of the pass, a different CPU for each thread.
2) The 'real' column is not comparable to the 'sys' or 'user' columns.
The 'real' column is seconds wall clock time elapsed, from beginning
to end of that test pass. The 'sys' and 'user' columns are total
CPU seconds spent on that test pass. For a 19 thread test run,
for example, the sum of 'sys' and 'user' could be up to 19 times the
number of 'real' elapsed wall clock seconds.
3) Tests were run on a fresh, single-user boot, to minimize the amount
of memory already in use at the start of the test, and to minimize
the amount of background activity that might interfere.
4) Tests were done on a 56 CPU, 28 Node system with 96 GBytes of RAM.
5) Notice that the 'real' time gets large for the single thread runs, even
though the measured 'sys' and 'user' times are modest. I'm not sure what
that means - probably something to do with it being slow for one thread to
be accessing memory along ways away. Perhaps the fake numa system, running
ostensibly the same workload, would not show this substantial degradation
of 'real' time for one thread on many nodes -- lets hope not.
6) The high thread count passes (one thread per CPU - on 55 of 56 CPUs)
ran quite efficiently, as one might expect. Each pair of threads needed
to allocate and touch the memory on the node the two threads shared, a
pleasantly parallizable workload.
7) The intermediate thread count passes, when asking for alot of memory forcing
them to go to a few neighboring nodes, improved the most with this zonelist
caching patch.
Conclusions:
* This zonelist cache patch probably makes little difference one way or the
other for most workloads on real numa hardware, if those workloads avoid
heavy off node allocations.
* For memory intensive workloads requiring substantial off-node allocations
on real numa hardware, this patch improves both kernel and elapsed timings
up to ten per-cent.
* For fake numa systems, I'm optimistic, but will have to leave that up to
Rohit Seth to actually test (once I get him a 2.6.18 backport.)
Signed-off-by: Paul Jackson <pj@sgi.com>
Cc: Rohit Seth <rohitseth@google.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Cc: David Rientjes <rientjes@cs.washington.edu>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:31:48 +00:00
|
|
|
* Scan zonelist, looking for a zone with enough free.
|
2014-10-20 11:50:30 +00:00
|
|
|
* See also __cpuset_node_allowed() comment in kernel/cpuset.c.
|
2005-11-14 00:06:43 +00:00
|
|
|
*/
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
no_fallback = alloc_flags & ALLOC_NOFRAGMENT;
|
|
|
|
z = ac->preferred_zoneref;
|
2020-10-13 23:55:57 +00:00
|
|
|
for_next_zone_zonelist_nodemask(zone, z, ac->highest_zoneidx,
|
|
|
|
ac->nodemask) {
|
2016-05-20 00:13:47 +00:00
|
|
|
struct page *page;
|
2013-09-11 21:20:46 +00:00
|
|
|
unsigned long mark;
|
|
|
|
|
2014-06-04 23:10:08 +00:00
|
|
|
if (cpusets_enabled() &&
|
|
|
|
(alloc_flags & ALLOC_CPUSET) &&
|
2016-05-20 00:14:30 +00:00
|
|
|
!__cpuset_zone_allowed(zone, gfp_mask))
|
mm: page allocator: initialise ZLC for first zone eligible for zone_reclaim
There have been a small number of complaints about significant stalls
while copying large amounts of data on NUMA machines reported on a
distribution bugzilla. In these cases, zone_reclaim was enabled by
default due to large NUMA distances. In general, the complaints have not
been about the workload itself unless it was a file server (in which case
the recommendation was disable zone_reclaim).
The stalls are mostly due to significant amounts of time spent scanning
the preferred zone for pages to free. After a failure, it might fallback
to another node (as zonelists are often node-ordered rather than
zone-ordered) but stall quickly again when the next allocation attempt
occurs. In bad cases, each page allocated results in a full scan of the
preferred zone.
Patch 1 checks the preferred zone for recent allocation failure
which is particularly important if zone_reclaim has failed
recently. This avoids rescanning the zone in the near future and
instead falling back to another node. This may hurt node locality
in some cases but a failure to zone_reclaim is more expensive than
a remote access.
Patch 2 clears the zlc information after direct reclaim.
Otherwise, zone_reclaim can mark zones full, direct reclaim can
reclaim enough pages but the zone is still not considered for
allocation.
This was tested on a 24-thread 2-node x86_64 machine. The tests were
focused on large amounts of IO. All tests were bound to the CPUs on
node-0 to avoid disturbances due to processes being scheduled on different
nodes. The kernels tested are
3.0-rc6-vanilla Vanilla 3.0-rc6
zlcfirst Patch 1 applied
zlcreconsider Patches 1+2 applied
FS-Mark
./fs_mark -d /tmp/fsmark-10813 -D 100 -N 5000 -n 208 -L 35 -t 24 -S0 -s 524288
fsmark-3.0-rc6 3.0-rc6 3.0-rc6
vanilla zlcfirs zlcreconsider
Files/s min 54.90 ( 0.00%) 49.80 (-10.24%) 49.10 (-11.81%)
Files/s mean 100.11 ( 0.00%) 135.17 (25.94%) 146.93 (31.87%)
Files/s stddev 57.51 ( 0.00%) 138.97 (58.62%) 158.69 (63.76%)
Files/s max 361.10 ( 0.00%) 834.40 (56.72%) 802.40 (55.00%)
Overhead min 76704.00 ( 0.00%) 76501.00 ( 0.27%) 77784.00 (-1.39%)
Overhead mean 1485356.51 ( 0.00%) 1035797.83 (43.40%) 1594680.26 (-6.86%)
Overhead stddev 1848122.53 ( 0.00%) 881489.88 (109.66%) 1772354.90 ( 4.27%)
Overhead max 7989060.00 ( 0.00%) 3369118.00 (137.13%) 10135324.00 (-21.18%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 501.49 493.91 499.93
Total Elapsed Time (seconds) 2451.57 2257.48 2215.92
MMTests Statistics: vmstat
Page Ins 46268 63840 66008
Page Outs 90821596 90671128 88043732
Swap Ins 0 0 0
Swap Outs 0 0 0
Direct pages scanned 13091697 8966863 8971790
Kswapd pages scanned 0 1830011 1831116
Kswapd pages reclaimed 0 1829068 1829930
Direct pages reclaimed 13037777 8956828 8648314
Kswapd efficiency 100% 99% 99%
Kswapd velocity 0.000 810.643 826.346
Direct efficiency 99% 99% 96%
Direct velocity 5340.128 3972.068 4048.788
Percentage direct scans 100% 83% 83%
Page writes by reclaim 0 3 0
Slabs scanned 796672 720640 720256
Direct inode steals 7422667 7160012 7088638
Kswapd inode steals 0 1736840 2021238
Test completes far faster with a large increase in the number of files
created per second. Standard deviation is high as a small number of
iterations were much higher than the mean. The number of pages scanned by
zone_reclaim is reduced and kswapd is used for more work.
LARGE DD
3.0-rc6 3.0-rc6 3.0-rc6
vanilla zlcfirst zlcreconsider
download tar 59 ( 0.00%) 59 ( 0.00%) 55 ( 7.27%)
dd source files 527 ( 0.00%) 296 (78.04%) 320 (64.69%)
delete source 36 ( 0.00%) 19 (89.47%) 20 (80.00%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 125.03 118.98 122.01
Total Elapsed Time (seconds) 624.56 375.02 398.06
MMTests Statistics: vmstat
Page Ins 3594216 439368 407032
Page Outs 23380832 23380488 23377444
Swap Ins 0 0 0
Swap Outs 0 436 287
Direct pages scanned 17482342 69315973 82864918
Kswapd pages scanned 0 519123 575425
Kswapd pages reclaimed 0 466501 522487
Direct pages reclaimed 5858054 2732949 2712547
Kswapd efficiency 100% 89% 90%
Kswapd velocity 0.000 1384.254 1445.574
Direct efficiency 33% 3% 3%
Direct velocity 27991.453 184832.737 208171.929
Percentage direct scans 100% 99% 99%
Page writes by reclaim 0 5082 13917
Slabs scanned 17280 29952 35328
Direct inode steals 115257 1431122 332201
Kswapd inode steals 0 0 979532
This test downloads a large tarfile and copies it with dd a number of
times - similar to the most recent bug report I've dealt with. Time to
completion is reduced. The number of pages scanned directly is still
disturbingly high with a low efficiency but this is likely due to the
number of dirty pages encountered. The figures could probably be improved
with more work around how kswapd is used and how dirty pages are handled
but that is separate work and this result is significant on its own.
Streaming Mapped Writer
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 124.47 111.67 112.64
Total Elapsed Time (seconds) 2138.14 1816.30 1867.56
MMTests Statistics: vmstat
Page Ins 90760 89124 89516
Page Outs 121028340 120199524 120736696
Swap Ins 0 86 55
Swap Outs 0 0 0
Direct pages scanned 114989363 96461439 96330619
Kswapd pages scanned 56430948 56965763 57075875
Kswapd pages reclaimed 27743219 27752044 27766606
Direct pages reclaimed 49777 46884 36655
Kswapd efficiency 49% 48% 48%
Kswapd velocity 26392.541 31363.631 30561.736
Direct efficiency 0% 0% 0%
Direct velocity 53780.091 53108.759 51581.004
Percentage direct scans 67% 62% 62%
Page writes by reclaim 385 122 1513
Slabs scanned 43008 39040 42112
Direct inode steals 0 10 8
Kswapd inode steals 733 534 477
This test just creates a large file mapping and writes to it linearly.
Time to completion is again reduced.
The gains are mostly down to two things. In many cases, there is less
scanning as zone_reclaim simply gives up faster due to recent failures.
The second reason is that memory is used more efficiently. Instead of
scanning the preferred zone every time, the allocator falls back to
another zone and uses it instead improving overall memory utilisation.
This patch: initialise ZLC for first zone eligible for zone_reclaim.
The zonelist cache (ZLC) is used among other things to record if
zone_reclaim() failed for a particular zone recently. The intention is to
avoid a high cost scanning extremely long zonelists or scanning within the
zone uselessly.
Currently the zonelist cache is setup only after the first zone has been
considered and zone_reclaim() has been called. The objective was to avoid
a costly setup but zone_reclaim is itself quite expensive. If it is
failing regularly such as the first eligible zone having mostly mapped
pages, the cost in scanning and allocation stalls is far higher than the
ZLC initialisation step.
This patch initialises ZLC before the first eligible zone calls
zone_reclaim(). Once initialised, it is checked whether the zone failed
zone_reclaim recently. If it has, the zone is skipped. As the first zone
is now being checked, additional care has to be taken about zones marked
full. A zone can be marked "full" because it should not have enough
unmapped pages for zone_reclaim but this is excessive as direct reclaim or
kswapd may succeed where zone_reclaim fails. Only mark zones "full" after
zone_reclaim fails if it failed to reclaim enough pages after scanning.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-07-26 00:12:29 +00:00
|
|
|
continue;
|
mm: try to distribute dirty pages fairly across zones
The maximum number of dirty pages that exist in the system at any time is
determined by a number of pages considered dirtyable and a user-configured
percentage of those, or an absolute number in bytes.
This number of dirtyable pages is the sum of memory provided by all the
zones in the system minus their lowmem reserves and high watermarks, so
that the system can retain a healthy number of free pages without having
to reclaim dirty pages.
But there is a flaw in that we have a zoned page allocator which does not
care about the global state but rather the state of individual memory
zones. And right now there is nothing that prevents one zone from filling
up with dirty pages while other zones are spared, which frequently leads
to situations where kswapd, in order to restore the watermark of free
pages, does indeed have to write pages from that zone's LRU list. This
can interfere so badly with IO from the flusher threads that major
filesystems (btrfs, xfs, ext4) mostly ignore write requests from reclaim
already, taking away the VM's only possibility to keep such a zone
balanced, aside from hoping the flushers will soon clean pages from that
zone.
Enter per-zone dirty limits. They are to a zone's dirtyable memory what
the global limit is to the global amount of dirtyable memory, and try to
make sure that no single zone receives more than its fair share of the
globally allowed dirty pages in the first place. As the number of pages
considered dirtyable excludes the zones' lowmem reserves and high
watermarks, the maximum number of dirty pages in a zone is such that the
zone can always be balanced without requiring page cleaning.
As this is a placement decision in the page allocator and pages are
dirtied only after the allocation, this patch allows allocators to pass
__GFP_WRITE when they know in advance that the page will be written to and
become dirty soon. The page allocator will then attempt to allocate from
the first zone of the zonelist - which on NUMA is determined by the task's
NUMA memory policy - that has not exceeded its dirty limit.
At first glance, it would appear that the diversion to lower zones can
increase pressure on them, but this is not the case. With a full high
zone, allocations will be diverted to lower zones eventually, so it is
more of a shift in timing of the lower zone allocations. Workloads that
previously could fit their dirty pages completely in the higher zone may
be forced to allocate from lower zones, but the amount of pages that
"spill over" are limited themselves by the lower zones' dirty constraints,
and thus unlikely to become a problem.
For now, the problem of unfair dirty page distribution remains for NUMA
configurations where the zones allowed for allocation are in sum not big
enough to trigger the global dirty limits, wake up the flusher threads and
remedy the situation. Because of this, an allocation that could not
succeed on any of the considered zones is allowed to ignore the dirty
limits before going into direct reclaim or even failing the allocation,
until a future patch changes the global dirty throttling and flusher
thread activation so that they take individual zone states into account.
Test results
15M DMA + 3246M DMA32 + 504 Normal = 3765M memory
40% dirty ratio
16G USB thumb drive
10 runs of dd if=/dev/zero of=disk/zeroes bs=32k count=$((10 << 15))
seconds nr_vmscan_write
(stddev) min| median| max
xfs
vanilla: 549.747( 3.492) 0.000| 0.000| 0.000
patched: 550.996( 3.802) 0.000| 0.000| 0.000
fuse-ntfs
vanilla: 1183.094(53.178) 54349.000| 59341.000| 65163.000
patched: 558.049(17.914) 0.000| 0.000| 43.000
btrfs
vanilla: 573.679(14.015) 156657.000| 460178.000| 606926.000
patched: 563.365(11.368) 0.000| 0.000| 1362.000
ext4
vanilla: 561.197(15.782) 0.000|2725438.000|4143837.000
patched: 568.806(17.496) 0.000| 0.000| 0.000
Signed-off-by: Johannes Weiner <jweiner@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Tested-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-10 23:07:49 +00:00
|
|
|
/*
|
|
|
|
* When allocating a page cache page for writing, we
|
2016-07-28 22:46:11 +00:00
|
|
|
* want to get it from a node that is within its dirty
|
|
|
|
* limit, such that no single node holds more than its
|
mm: try to distribute dirty pages fairly across zones
The maximum number of dirty pages that exist in the system at any time is
determined by a number of pages considered dirtyable and a user-configured
percentage of those, or an absolute number in bytes.
This number of dirtyable pages is the sum of memory provided by all the
zones in the system minus their lowmem reserves and high watermarks, so
that the system can retain a healthy number of free pages without having
to reclaim dirty pages.
But there is a flaw in that we have a zoned page allocator which does not
care about the global state but rather the state of individual memory
zones. And right now there is nothing that prevents one zone from filling
up with dirty pages while other zones are spared, which frequently leads
to situations where kswapd, in order to restore the watermark of free
pages, does indeed have to write pages from that zone's LRU list. This
can interfere so badly with IO from the flusher threads that major
filesystems (btrfs, xfs, ext4) mostly ignore write requests from reclaim
already, taking away the VM's only possibility to keep such a zone
balanced, aside from hoping the flushers will soon clean pages from that
zone.
Enter per-zone dirty limits. They are to a zone's dirtyable memory what
the global limit is to the global amount of dirtyable memory, and try to
make sure that no single zone receives more than its fair share of the
globally allowed dirty pages in the first place. As the number of pages
considered dirtyable excludes the zones' lowmem reserves and high
watermarks, the maximum number of dirty pages in a zone is such that the
zone can always be balanced without requiring page cleaning.
As this is a placement decision in the page allocator and pages are
dirtied only after the allocation, this patch allows allocators to pass
__GFP_WRITE when they know in advance that the page will be written to and
become dirty soon. The page allocator will then attempt to allocate from
the first zone of the zonelist - which on NUMA is determined by the task's
NUMA memory policy - that has not exceeded its dirty limit.
At first glance, it would appear that the diversion to lower zones can
increase pressure on them, but this is not the case. With a full high
zone, allocations will be diverted to lower zones eventually, so it is
more of a shift in timing of the lower zone allocations. Workloads that
previously could fit their dirty pages completely in the higher zone may
be forced to allocate from lower zones, but the amount of pages that
"spill over" are limited themselves by the lower zones' dirty constraints,
and thus unlikely to become a problem.
For now, the problem of unfair dirty page distribution remains for NUMA
configurations where the zones allowed for allocation are in sum not big
enough to trigger the global dirty limits, wake up the flusher threads and
remedy the situation. Because of this, an allocation that could not
succeed on any of the considered zones is allowed to ignore the dirty
limits before going into direct reclaim or even failing the allocation,
until a future patch changes the global dirty throttling and flusher
thread activation so that they take individual zone states into account.
Test results
15M DMA + 3246M DMA32 + 504 Normal = 3765M memory
40% dirty ratio
16G USB thumb drive
10 runs of dd if=/dev/zero of=disk/zeroes bs=32k count=$((10 << 15))
seconds nr_vmscan_write
(stddev) min| median| max
xfs
vanilla: 549.747( 3.492) 0.000| 0.000| 0.000
patched: 550.996( 3.802) 0.000| 0.000| 0.000
fuse-ntfs
vanilla: 1183.094(53.178) 54349.000| 59341.000| 65163.000
patched: 558.049(17.914) 0.000| 0.000| 43.000
btrfs
vanilla: 573.679(14.015) 156657.000| 460178.000| 606926.000
patched: 563.365(11.368) 0.000| 0.000| 1362.000
ext4
vanilla: 561.197(15.782) 0.000|2725438.000|4143837.000
patched: 568.806(17.496) 0.000| 0.000| 0.000
Signed-off-by: Johannes Weiner <jweiner@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Tested-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-10 23:07:49 +00:00
|
|
|
* proportional share of globally allowed dirty pages.
|
2016-07-28 22:46:11 +00:00
|
|
|
* The dirty limits take into account the node's
|
mm: try to distribute dirty pages fairly across zones
The maximum number of dirty pages that exist in the system at any time is
determined by a number of pages considered dirtyable and a user-configured
percentage of those, or an absolute number in bytes.
This number of dirtyable pages is the sum of memory provided by all the
zones in the system minus their lowmem reserves and high watermarks, so
that the system can retain a healthy number of free pages without having
to reclaim dirty pages.
But there is a flaw in that we have a zoned page allocator which does not
care about the global state but rather the state of individual memory
zones. And right now there is nothing that prevents one zone from filling
up with dirty pages while other zones are spared, which frequently leads
to situations where kswapd, in order to restore the watermark of free
pages, does indeed have to write pages from that zone's LRU list. This
can interfere so badly with IO from the flusher threads that major
filesystems (btrfs, xfs, ext4) mostly ignore write requests from reclaim
already, taking away the VM's only possibility to keep such a zone
balanced, aside from hoping the flushers will soon clean pages from that
zone.
Enter per-zone dirty limits. They are to a zone's dirtyable memory what
the global limit is to the global amount of dirtyable memory, and try to
make sure that no single zone receives more than its fair share of the
globally allowed dirty pages in the first place. As the number of pages
considered dirtyable excludes the zones' lowmem reserves and high
watermarks, the maximum number of dirty pages in a zone is such that the
zone can always be balanced without requiring page cleaning.
As this is a placement decision in the page allocator and pages are
dirtied only after the allocation, this patch allows allocators to pass
__GFP_WRITE when they know in advance that the page will be written to and
become dirty soon. The page allocator will then attempt to allocate from
the first zone of the zonelist - which on NUMA is determined by the task's
NUMA memory policy - that has not exceeded its dirty limit.
At first glance, it would appear that the diversion to lower zones can
increase pressure on them, but this is not the case. With a full high
zone, allocations will be diverted to lower zones eventually, so it is
more of a shift in timing of the lower zone allocations. Workloads that
previously could fit their dirty pages completely in the higher zone may
be forced to allocate from lower zones, but the amount of pages that
"spill over" are limited themselves by the lower zones' dirty constraints,
and thus unlikely to become a problem.
For now, the problem of unfair dirty page distribution remains for NUMA
configurations where the zones allowed for allocation are in sum not big
enough to trigger the global dirty limits, wake up the flusher threads and
remedy the situation. Because of this, an allocation that could not
succeed on any of the considered zones is allowed to ignore the dirty
limits before going into direct reclaim or even failing the allocation,
until a future patch changes the global dirty throttling and flusher
thread activation so that they take individual zone states into account.
Test results
15M DMA + 3246M DMA32 + 504 Normal = 3765M memory
40% dirty ratio
16G USB thumb drive
10 runs of dd if=/dev/zero of=disk/zeroes bs=32k count=$((10 << 15))
seconds nr_vmscan_write
(stddev) min| median| max
xfs
vanilla: 549.747( 3.492) 0.000| 0.000| 0.000
patched: 550.996( 3.802) 0.000| 0.000| 0.000
fuse-ntfs
vanilla: 1183.094(53.178) 54349.000| 59341.000| 65163.000
patched: 558.049(17.914) 0.000| 0.000| 43.000
btrfs
vanilla: 573.679(14.015) 156657.000| 460178.000| 606926.000
patched: 563.365(11.368) 0.000| 0.000| 1362.000
ext4
vanilla: 561.197(15.782) 0.000|2725438.000|4143837.000
patched: 568.806(17.496) 0.000| 0.000| 0.000
Signed-off-by: Johannes Weiner <jweiner@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Tested-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-10 23:07:49 +00:00
|
|
|
* lowmem reserves and high watermark so that kswapd
|
|
|
|
* should be able to balance it without having to
|
|
|
|
* write pages from its LRU list.
|
|
|
|
*
|
|
|
|
* XXX: For now, allow allocations to potentially
|
2016-07-28 22:46:11 +00:00
|
|
|
* exceed the per-node dirty limit in the slowpath
|
2015-11-07 00:28:12 +00:00
|
|
|
* (spread_dirty_pages unset) before going into reclaim,
|
mm: try to distribute dirty pages fairly across zones
The maximum number of dirty pages that exist in the system at any time is
determined by a number of pages considered dirtyable and a user-configured
percentage of those, or an absolute number in bytes.
This number of dirtyable pages is the sum of memory provided by all the
zones in the system minus their lowmem reserves and high watermarks, so
that the system can retain a healthy number of free pages without having
to reclaim dirty pages.
But there is a flaw in that we have a zoned page allocator which does not
care about the global state but rather the state of individual memory
zones. And right now there is nothing that prevents one zone from filling
up with dirty pages while other zones are spared, which frequently leads
to situations where kswapd, in order to restore the watermark of free
pages, does indeed have to write pages from that zone's LRU list. This
can interfere so badly with IO from the flusher threads that major
filesystems (btrfs, xfs, ext4) mostly ignore write requests from reclaim
already, taking away the VM's only possibility to keep such a zone
balanced, aside from hoping the flushers will soon clean pages from that
zone.
Enter per-zone dirty limits. They are to a zone's dirtyable memory what
the global limit is to the global amount of dirtyable memory, and try to
make sure that no single zone receives more than its fair share of the
globally allowed dirty pages in the first place. As the number of pages
considered dirtyable excludes the zones' lowmem reserves and high
watermarks, the maximum number of dirty pages in a zone is such that the
zone can always be balanced without requiring page cleaning.
As this is a placement decision in the page allocator and pages are
dirtied only after the allocation, this patch allows allocators to pass
__GFP_WRITE when they know in advance that the page will be written to and
become dirty soon. The page allocator will then attempt to allocate from
the first zone of the zonelist - which on NUMA is determined by the task's
NUMA memory policy - that has not exceeded its dirty limit.
At first glance, it would appear that the diversion to lower zones can
increase pressure on them, but this is not the case. With a full high
zone, allocations will be diverted to lower zones eventually, so it is
more of a shift in timing of the lower zone allocations. Workloads that
previously could fit their dirty pages completely in the higher zone may
be forced to allocate from lower zones, but the amount of pages that
"spill over" are limited themselves by the lower zones' dirty constraints,
and thus unlikely to become a problem.
For now, the problem of unfair dirty page distribution remains for NUMA
configurations where the zones allowed for allocation are in sum not big
enough to trigger the global dirty limits, wake up the flusher threads and
remedy the situation. Because of this, an allocation that could not
succeed on any of the considered zones is allowed to ignore the dirty
limits before going into direct reclaim or even failing the allocation,
until a future patch changes the global dirty throttling and flusher
thread activation so that they take individual zone states into account.
Test results
15M DMA + 3246M DMA32 + 504 Normal = 3765M memory
40% dirty ratio
16G USB thumb drive
10 runs of dd if=/dev/zero of=disk/zeroes bs=32k count=$((10 << 15))
seconds nr_vmscan_write
(stddev) min| median| max
xfs
vanilla: 549.747( 3.492) 0.000| 0.000| 0.000
patched: 550.996( 3.802) 0.000| 0.000| 0.000
fuse-ntfs
vanilla: 1183.094(53.178) 54349.000| 59341.000| 65163.000
patched: 558.049(17.914) 0.000| 0.000| 43.000
btrfs
vanilla: 573.679(14.015) 156657.000| 460178.000| 606926.000
patched: 563.365(11.368) 0.000| 0.000| 1362.000
ext4
vanilla: 561.197(15.782) 0.000|2725438.000|4143837.000
patched: 568.806(17.496) 0.000| 0.000| 0.000
Signed-off-by: Johannes Weiner <jweiner@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Tested-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-10 23:07:49 +00:00
|
|
|
* which is important when on a NUMA setup the allowed
|
2016-07-28 22:46:11 +00:00
|
|
|
* nodes are together not big enough to reach the
|
mm: try to distribute dirty pages fairly across zones
The maximum number of dirty pages that exist in the system at any time is
determined by a number of pages considered dirtyable and a user-configured
percentage of those, or an absolute number in bytes.
This number of dirtyable pages is the sum of memory provided by all the
zones in the system minus their lowmem reserves and high watermarks, so
that the system can retain a healthy number of free pages without having
to reclaim dirty pages.
But there is a flaw in that we have a zoned page allocator which does not
care about the global state but rather the state of individual memory
zones. And right now there is nothing that prevents one zone from filling
up with dirty pages while other zones are spared, which frequently leads
to situations where kswapd, in order to restore the watermark of free
pages, does indeed have to write pages from that zone's LRU list. This
can interfere so badly with IO from the flusher threads that major
filesystems (btrfs, xfs, ext4) mostly ignore write requests from reclaim
already, taking away the VM's only possibility to keep such a zone
balanced, aside from hoping the flushers will soon clean pages from that
zone.
Enter per-zone dirty limits. They are to a zone's dirtyable memory what
the global limit is to the global amount of dirtyable memory, and try to
make sure that no single zone receives more than its fair share of the
globally allowed dirty pages in the first place. As the number of pages
considered dirtyable excludes the zones' lowmem reserves and high
watermarks, the maximum number of dirty pages in a zone is such that the
zone can always be balanced without requiring page cleaning.
As this is a placement decision in the page allocator and pages are
dirtied only after the allocation, this patch allows allocators to pass
__GFP_WRITE when they know in advance that the page will be written to and
become dirty soon. The page allocator will then attempt to allocate from
the first zone of the zonelist - which on NUMA is determined by the task's
NUMA memory policy - that has not exceeded its dirty limit.
At first glance, it would appear that the diversion to lower zones can
increase pressure on them, but this is not the case. With a full high
zone, allocations will be diverted to lower zones eventually, so it is
more of a shift in timing of the lower zone allocations. Workloads that
previously could fit their dirty pages completely in the higher zone may
be forced to allocate from lower zones, but the amount of pages that
"spill over" are limited themselves by the lower zones' dirty constraints,
and thus unlikely to become a problem.
For now, the problem of unfair dirty page distribution remains for NUMA
configurations where the zones allowed for allocation are in sum not big
enough to trigger the global dirty limits, wake up the flusher threads and
remedy the situation. Because of this, an allocation that could not
succeed on any of the considered zones is allowed to ignore the dirty
limits before going into direct reclaim or even failing the allocation,
until a future patch changes the global dirty throttling and flusher
thread activation so that they take individual zone states into account.
Test results
15M DMA + 3246M DMA32 + 504 Normal = 3765M memory
40% dirty ratio
16G USB thumb drive
10 runs of dd if=/dev/zero of=disk/zeroes bs=32k count=$((10 << 15))
seconds nr_vmscan_write
(stddev) min| median| max
xfs
vanilla: 549.747( 3.492) 0.000| 0.000| 0.000
patched: 550.996( 3.802) 0.000| 0.000| 0.000
fuse-ntfs
vanilla: 1183.094(53.178) 54349.000| 59341.000| 65163.000
patched: 558.049(17.914) 0.000| 0.000| 43.000
btrfs
vanilla: 573.679(14.015) 156657.000| 460178.000| 606926.000
patched: 563.365(11.368) 0.000| 0.000| 1362.000
ext4
vanilla: 561.197(15.782) 0.000|2725438.000|4143837.000
patched: 568.806(17.496) 0.000| 0.000| 0.000
Signed-off-by: Johannes Weiner <jweiner@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Tested-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-10 23:07:49 +00:00
|
|
|
* global limit. The proper fix for these situations
|
2016-07-28 22:46:11 +00:00
|
|
|
* will require awareness of nodes in the
|
mm: try to distribute dirty pages fairly across zones
The maximum number of dirty pages that exist in the system at any time is
determined by a number of pages considered dirtyable and a user-configured
percentage of those, or an absolute number in bytes.
This number of dirtyable pages is the sum of memory provided by all the
zones in the system minus their lowmem reserves and high watermarks, so
that the system can retain a healthy number of free pages without having
to reclaim dirty pages.
But there is a flaw in that we have a zoned page allocator which does not
care about the global state but rather the state of individual memory
zones. And right now there is nothing that prevents one zone from filling
up with dirty pages while other zones are spared, which frequently leads
to situations where kswapd, in order to restore the watermark of free
pages, does indeed have to write pages from that zone's LRU list. This
can interfere so badly with IO from the flusher threads that major
filesystems (btrfs, xfs, ext4) mostly ignore write requests from reclaim
already, taking away the VM's only possibility to keep such a zone
balanced, aside from hoping the flushers will soon clean pages from that
zone.
Enter per-zone dirty limits. They are to a zone's dirtyable memory what
the global limit is to the global amount of dirtyable memory, and try to
make sure that no single zone receives more than its fair share of the
globally allowed dirty pages in the first place. As the number of pages
considered dirtyable excludes the zones' lowmem reserves and high
watermarks, the maximum number of dirty pages in a zone is such that the
zone can always be balanced without requiring page cleaning.
As this is a placement decision in the page allocator and pages are
dirtied only after the allocation, this patch allows allocators to pass
__GFP_WRITE when they know in advance that the page will be written to and
become dirty soon. The page allocator will then attempt to allocate from
the first zone of the zonelist - which on NUMA is determined by the task's
NUMA memory policy - that has not exceeded its dirty limit.
At first glance, it would appear that the diversion to lower zones can
increase pressure on them, but this is not the case. With a full high
zone, allocations will be diverted to lower zones eventually, so it is
more of a shift in timing of the lower zone allocations. Workloads that
previously could fit their dirty pages completely in the higher zone may
be forced to allocate from lower zones, but the amount of pages that
"spill over" are limited themselves by the lower zones' dirty constraints,
and thus unlikely to become a problem.
For now, the problem of unfair dirty page distribution remains for NUMA
configurations where the zones allowed for allocation are in sum not big
enough to trigger the global dirty limits, wake up the flusher threads and
remedy the situation. Because of this, an allocation that could not
succeed on any of the considered zones is allowed to ignore the dirty
limits before going into direct reclaim or even failing the allocation,
until a future patch changes the global dirty throttling and flusher
thread activation so that they take individual zone states into account.
Test results
15M DMA + 3246M DMA32 + 504 Normal = 3765M memory
40% dirty ratio
16G USB thumb drive
10 runs of dd if=/dev/zero of=disk/zeroes bs=32k count=$((10 << 15))
seconds nr_vmscan_write
(stddev) min| median| max
xfs
vanilla: 549.747( 3.492) 0.000| 0.000| 0.000
patched: 550.996( 3.802) 0.000| 0.000| 0.000
fuse-ntfs
vanilla: 1183.094(53.178) 54349.000| 59341.000| 65163.000
patched: 558.049(17.914) 0.000| 0.000| 43.000
btrfs
vanilla: 573.679(14.015) 156657.000| 460178.000| 606926.000
patched: 563.365(11.368) 0.000| 0.000| 1362.000
ext4
vanilla: 561.197(15.782) 0.000|2725438.000|4143837.000
patched: 568.806(17.496) 0.000| 0.000| 0.000
Signed-off-by: Johannes Weiner <jweiner@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Tested-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-01-10 23:07:49 +00:00
|
|
|
* dirty-throttling and the flusher threads.
|
|
|
|
*/
|
2016-07-28 22:46:53 +00:00
|
|
|
if (ac->spread_dirty_pages) {
|
|
|
|
if (last_pgdat_dirty_limit == zone->zone_pgdat)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (!node_dirty_ok(zone->zone_pgdat)) {
|
|
|
|
last_pgdat_dirty_limit = zone->zone_pgdat;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
}
|
2005-11-14 00:06:43 +00:00
|
|
|
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
if (no_fallback && nr_online_nodes > 1 &&
|
|
|
|
zone != ac->preferred_zoneref->zone) {
|
|
|
|
int local_nid;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If moving to a remote node, retry but allow
|
|
|
|
* fragmenting fallbacks. Locality is more important
|
|
|
|
* than fragmentation avoidance.
|
|
|
|
*/
|
|
|
|
local_nid = zone_to_nid(ac->preferred_zoneref->zone);
|
|
|
|
if (zone_to_nid(zone) != local_nid) {
|
|
|
|
alloc_flags &= ~ALLOC_NOFRAGMENT;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-12-28 08:35:44 +00:00
|
|
|
mark = wmark_pages(zone, alloc_flags & ALLOC_WMARK_MASK);
|
2016-05-20 00:14:07 +00:00
|
|
|
if (!zone_watermark_fast(zone, order, mark,
|
mm, page_alloc: skip ->waternark_boost for atomic order-0 allocations
When boosting is enabled, it is observed that rate of atomic order-0
allocation failures are high due to the fact that free levels in the
system are checked with ->watermark_boost offset. This is not a problem
for sleepable allocations but for atomic allocations which looks like
regression.
This problem is seen frequently on system setup of Android kernel running
on Snapdragon hardware with 4GB RAM size. When no extfrag event occurred
in the system, ->watermark_boost factor is zero, thus the watermark
configurations in the system are:
_watermark = (
[WMARK_MIN] = 1272, --> ~5MB
[WMARK_LOW] = 9067, --> ~36MB
[WMARK_HIGH] = 9385), --> ~38MB
watermark_boost = 0
After launching some memory hungry applications in Android which can cause
extfrag events in the system to an extent that ->watermark_boost can be
set to max i.e. default boost factor makes it to 150% of high watermark.
_watermark = (
[WMARK_MIN] = 1272, --> ~5MB
[WMARK_LOW] = 9067, --> ~36MB
[WMARK_HIGH] = 9385), --> ~38MB
watermark_boost = 14077, -->~57MB
With default system configuration, for an atomic order-0 allocation to
succeed, having free memory of ~2MB will suffice. But boosting makes the
min_wmark to ~61MB thus for an atomic order-0 allocation to be successful
system should have minimum of ~23MB of free memory(from calculations of
zone_watermark_ok(), min = 3/4(min/2)). But failures are observed despite
system is having ~20MB of free memory. In the testing, this is
reproducible as early as first 300secs since boot and with furtherlowram
configurations(<2GB) it is observed as early as first 150secs since boot.
These failures can be avoided by excluding the ->watermark_boost in
watermark caluculations for atomic order-0 allocations.
[akpm@linux-foundation.org: fix comment grammar, reflow comment]
[charante@codeaurora.org: fix suggested by Mel Gorman]
Link: http://lkml.kernel.org/r/31556793-57b1-1c21-1a9d-22674d9bd938@codeaurora.org
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Link: http://lkml.kernel.org/r/1589882284-21010-1-git-send-email-charante@codeaurora.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-07 06:25:24 +00:00
|
|
|
ac->highest_zoneidx, alloc_flags,
|
|
|
|
gfp_mask)) {
|
2009-06-16 22:33:22 +00:00
|
|
|
int ret;
|
|
|
|
|
2018-04-05 23:22:31 +00:00
|
|
|
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
|
|
|
|
/*
|
|
|
|
* Watermark failed for this zone, but see if we can
|
|
|
|
* grow this zone if it contains deferred pages.
|
|
|
|
*/
|
|
|
|
if (static_branch_unlikely(&deferred_pages)) {
|
|
|
|
if (_deferred_grow_zone(zone, order))
|
|
|
|
goto try_this_zone;
|
|
|
|
}
|
|
|
|
#endif
|
2014-06-04 23:10:14 +00:00
|
|
|
/* Checked here to keep the fast path fast */
|
|
|
|
BUILD_BUG_ON(ALLOC_NO_WATERMARKS < NR_WMARK);
|
|
|
|
if (alloc_flags & ALLOC_NO_WATERMARKS)
|
|
|
|
goto try_this_zone;
|
|
|
|
|
2016-07-28 22:46:32 +00:00
|
|
|
if (node_reclaim_mode == 0 ||
|
2016-05-20 00:14:10 +00:00
|
|
|
!zone_allows_reclaim(ac->preferred_zoneref->zone, zone))
|
mm: page allocator: initialise ZLC for first zone eligible for zone_reclaim
There have been a small number of complaints about significant stalls
while copying large amounts of data on NUMA machines reported on a
distribution bugzilla. In these cases, zone_reclaim was enabled by
default due to large NUMA distances. In general, the complaints have not
been about the workload itself unless it was a file server (in which case
the recommendation was disable zone_reclaim).
The stalls are mostly due to significant amounts of time spent scanning
the preferred zone for pages to free. After a failure, it might fallback
to another node (as zonelists are often node-ordered rather than
zone-ordered) but stall quickly again when the next allocation attempt
occurs. In bad cases, each page allocated results in a full scan of the
preferred zone.
Patch 1 checks the preferred zone for recent allocation failure
which is particularly important if zone_reclaim has failed
recently. This avoids rescanning the zone in the near future and
instead falling back to another node. This may hurt node locality
in some cases but a failure to zone_reclaim is more expensive than
a remote access.
Patch 2 clears the zlc information after direct reclaim.
Otherwise, zone_reclaim can mark zones full, direct reclaim can
reclaim enough pages but the zone is still not considered for
allocation.
This was tested on a 24-thread 2-node x86_64 machine. The tests were
focused on large amounts of IO. All tests were bound to the CPUs on
node-0 to avoid disturbances due to processes being scheduled on different
nodes. The kernels tested are
3.0-rc6-vanilla Vanilla 3.0-rc6
zlcfirst Patch 1 applied
zlcreconsider Patches 1+2 applied
FS-Mark
./fs_mark -d /tmp/fsmark-10813 -D 100 -N 5000 -n 208 -L 35 -t 24 -S0 -s 524288
fsmark-3.0-rc6 3.0-rc6 3.0-rc6
vanilla zlcfirs zlcreconsider
Files/s min 54.90 ( 0.00%) 49.80 (-10.24%) 49.10 (-11.81%)
Files/s mean 100.11 ( 0.00%) 135.17 (25.94%) 146.93 (31.87%)
Files/s stddev 57.51 ( 0.00%) 138.97 (58.62%) 158.69 (63.76%)
Files/s max 361.10 ( 0.00%) 834.40 (56.72%) 802.40 (55.00%)
Overhead min 76704.00 ( 0.00%) 76501.00 ( 0.27%) 77784.00 (-1.39%)
Overhead mean 1485356.51 ( 0.00%) 1035797.83 (43.40%) 1594680.26 (-6.86%)
Overhead stddev 1848122.53 ( 0.00%) 881489.88 (109.66%) 1772354.90 ( 4.27%)
Overhead max 7989060.00 ( 0.00%) 3369118.00 (137.13%) 10135324.00 (-21.18%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 501.49 493.91 499.93
Total Elapsed Time (seconds) 2451.57 2257.48 2215.92
MMTests Statistics: vmstat
Page Ins 46268 63840 66008
Page Outs 90821596 90671128 88043732
Swap Ins 0 0 0
Swap Outs 0 0 0
Direct pages scanned 13091697 8966863 8971790
Kswapd pages scanned 0 1830011 1831116
Kswapd pages reclaimed 0 1829068 1829930
Direct pages reclaimed 13037777 8956828 8648314
Kswapd efficiency 100% 99% 99%
Kswapd velocity 0.000 810.643 826.346
Direct efficiency 99% 99% 96%
Direct velocity 5340.128 3972.068 4048.788
Percentage direct scans 100% 83% 83%
Page writes by reclaim 0 3 0
Slabs scanned 796672 720640 720256
Direct inode steals 7422667 7160012 7088638
Kswapd inode steals 0 1736840 2021238
Test completes far faster with a large increase in the number of files
created per second. Standard deviation is high as a small number of
iterations were much higher than the mean. The number of pages scanned by
zone_reclaim is reduced and kswapd is used for more work.
LARGE DD
3.0-rc6 3.0-rc6 3.0-rc6
vanilla zlcfirst zlcreconsider
download tar 59 ( 0.00%) 59 ( 0.00%) 55 ( 7.27%)
dd source files 527 ( 0.00%) 296 (78.04%) 320 (64.69%)
delete source 36 ( 0.00%) 19 (89.47%) 20 (80.00%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 125.03 118.98 122.01
Total Elapsed Time (seconds) 624.56 375.02 398.06
MMTests Statistics: vmstat
Page Ins 3594216 439368 407032
Page Outs 23380832 23380488 23377444
Swap Ins 0 0 0
Swap Outs 0 436 287
Direct pages scanned 17482342 69315973 82864918
Kswapd pages scanned 0 519123 575425
Kswapd pages reclaimed 0 466501 522487
Direct pages reclaimed 5858054 2732949 2712547
Kswapd efficiency 100% 89% 90%
Kswapd velocity 0.000 1384.254 1445.574
Direct efficiency 33% 3% 3%
Direct velocity 27991.453 184832.737 208171.929
Percentage direct scans 100% 99% 99%
Page writes by reclaim 0 5082 13917
Slabs scanned 17280 29952 35328
Direct inode steals 115257 1431122 332201
Kswapd inode steals 0 0 979532
This test downloads a large tarfile and copies it with dd a number of
times - similar to the most recent bug report I've dealt with. Time to
completion is reduced. The number of pages scanned directly is still
disturbingly high with a low efficiency but this is likely due to the
number of dirty pages encountered. The figures could probably be improved
with more work around how kswapd is used and how dirty pages are handled
but that is separate work and this result is significant on its own.
Streaming Mapped Writer
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 124.47 111.67 112.64
Total Elapsed Time (seconds) 2138.14 1816.30 1867.56
MMTests Statistics: vmstat
Page Ins 90760 89124 89516
Page Outs 121028340 120199524 120736696
Swap Ins 0 86 55
Swap Outs 0 0 0
Direct pages scanned 114989363 96461439 96330619
Kswapd pages scanned 56430948 56965763 57075875
Kswapd pages reclaimed 27743219 27752044 27766606
Direct pages reclaimed 49777 46884 36655
Kswapd efficiency 49% 48% 48%
Kswapd velocity 26392.541 31363.631 30561.736
Direct efficiency 0% 0% 0%
Direct velocity 53780.091 53108.759 51581.004
Percentage direct scans 67% 62% 62%
Page writes by reclaim 385 122 1513
Slabs scanned 43008 39040 42112
Direct inode steals 0 10 8
Kswapd inode steals 733 534 477
This test just creates a large file mapping and writes to it linearly.
Time to completion is again reduced.
The gains are mostly down to two things. In many cases, there is less
scanning as zone_reclaim simply gives up faster due to recent failures.
The second reason is that memory is used more efficiently. Instead of
scanning the preferred zone every time, the allocator falls back to
another zone and uses it instead improving overall memory utilisation.
This patch: initialise ZLC for first zone eligible for zone_reclaim.
The zonelist cache (ZLC) is used among other things to record if
zone_reclaim() failed for a particular zone recently. The intention is to
avoid a high cost scanning extremely long zonelists or scanning within the
zone uselessly.
Currently the zonelist cache is setup only after the first zone has been
considered and zone_reclaim() has been called. The objective was to avoid
a costly setup but zone_reclaim is itself quite expensive. If it is
failing regularly such as the first eligible zone having mostly mapped
pages, the cost in scanning and allocation stalls is far higher than the
ZLC initialisation step.
This patch initialises ZLC before the first eligible zone calls
zone_reclaim(). Once initialised, it is checked whether the zone failed
zone_reclaim recently. If it has, the zone is skipped. As the first zone
is now being checked, additional care has to be taken about zones marked
full. A zone can be marked "full" because it should not have enough
unmapped pages for zone_reclaim but this is excessive as direct reclaim or
kswapd may succeed where zone_reclaim fails. Only mark zones "full" after
zone_reclaim fails if it failed to reclaim enough pages after scanning.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-07-26 00:12:29 +00:00
|
|
|
continue;
|
|
|
|
|
2016-07-28 22:46:32 +00:00
|
|
|
ret = node_reclaim(zone->zone_pgdat, gfp_mask, order);
|
2009-06-16 22:33:22 +00:00
|
|
|
switch (ret) {
|
2016-07-28 22:46:32 +00:00
|
|
|
case NODE_RECLAIM_NOSCAN:
|
2009-06-16 22:33:22 +00:00
|
|
|
/* did not scan */
|
mm: page allocator: initialise ZLC for first zone eligible for zone_reclaim
There have been a small number of complaints about significant stalls
while copying large amounts of data on NUMA machines reported on a
distribution bugzilla. In these cases, zone_reclaim was enabled by
default due to large NUMA distances. In general, the complaints have not
been about the workload itself unless it was a file server (in which case
the recommendation was disable zone_reclaim).
The stalls are mostly due to significant amounts of time spent scanning
the preferred zone for pages to free. After a failure, it might fallback
to another node (as zonelists are often node-ordered rather than
zone-ordered) but stall quickly again when the next allocation attempt
occurs. In bad cases, each page allocated results in a full scan of the
preferred zone.
Patch 1 checks the preferred zone for recent allocation failure
which is particularly important if zone_reclaim has failed
recently. This avoids rescanning the zone in the near future and
instead falling back to another node. This may hurt node locality
in some cases but a failure to zone_reclaim is more expensive than
a remote access.
Patch 2 clears the zlc information after direct reclaim.
Otherwise, zone_reclaim can mark zones full, direct reclaim can
reclaim enough pages but the zone is still not considered for
allocation.
This was tested on a 24-thread 2-node x86_64 machine. The tests were
focused on large amounts of IO. All tests were bound to the CPUs on
node-0 to avoid disturbances due to processes being scheduled on different
nodes. The kernels tested are
3.0-rc6-vanilla Vanilla 3.0-rc6
zlcfirst Patch 1 applied
zlcreconsider Patches 1+2 applied
FS-Mark
./fs_mark -d /tmp/fsmark-10813 -D 100 -N 5000 -n 208 -L 35 -t 24 -S0 -s 524288
fsmark-3.0-rc6 3.0-rc6 3.0-rc6
vanilla zlcfirs zlcreconsider
Files/s min 54.90 ( 0.00%) 49.80 (-10.24%) 49.10 (-11.81%)
Files/s mean 100.11 ( 0.00%) 135.17 (25.94%) 146.93 (31.87%)
Files/s stddev 57.51 ( 0.00%) 138.97 (58.62%) 158.69 (63.76%)
Files/s max 361.10 ( 0.00%) 834.40 (56.72%) 802.40 (55.00%)
Overhead min 76704.00 ( 0.00%) 76501.00 ( 0.27%) 77784.00 (-1.39%)
Overhead mean 1485356.51 ( 0.00%) 1035797.83 (43.40%) 1594680.26 (-6.86%)
Overhead stddev 1848122.53 ( 0.00%) 881489.88 (109.66%) 1772354.90 ( 4.27%)
Overhead max 7989060.00 ( 0.00%) 3369118.00 (137.13%) 10135324.00 (-21.18%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 501.49 493.91 499.93
Total Elapsed Time (seconds) 2451.57 2257.48 2215.92
MMTests Statistics: vmstat
Page Ins 46268 63840 66008
Page Outs 90821596 90671128 88043732
Swap Ins 0 0 0
Swap Outs 0 0 0
Direct pages scanned 13091697 8966863 8971790
Kswapd pages scanned 0 1830011 1831116
Kswapd pages reclaimed 0 1829068 1829930
Direct pages reclaimed 13037777 8956828 8648314
Kswapd efficiency 100% 99% 99%
Kswapd velocity 0.000 810.643 826.346
Direct efficiency 99% 99% 96%
Direct velocity 5340.128 3972.068 4048.788
Percentage direct scans 100% 83% 83%
Page writes by reclaim 0 3 0
Slabs scanned 796672 720640 720256
Direct inode steals 7422667 7160012 7088638
Kswapd inode steals 0 1736840 2021238
Test completes far faster with a large increase in the number of files
created per second. Standard deviation is high as a small number of
iterations were much higher than the mean. The number of pages scanned by
zone_reclaim is reduced and kswapd is used for more work.
LARGE DD
3.0-rc6 3.0-rc6 3.0-rc6
vanilla zlcfirst zlcreconsider
download tar 59 ( 0.00%) 59 ( 0.00%) 55 ( 7.27%)
dd source files 527 ( 0.00%) 296 (78.04%) 320 (64.69%)
delete source 36 ( 0.00%) 19 (89.47%) 20 (80.00%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 125.03 118.98 122.01
Total Elapsed Time (seconds) 624.56 375.02 398.06
MMTests Statistics: vmstat
Page Ins 3594216 439368 407032
Page Outs 23380832 23380488 23377444
Swap Ins 0 0 0
Swap Outs 0 436 287
Direct pages scanned 17482342 69315973 82864918
Kswapd pages scanned 0 519123 575425
Kswapd pages reclaimed 0 466501 522487
Direct pages reclaimed 5858054 2732949 2712547
Kswapd efficiency 100% 89% 90%
Kswapd velocity 0.000 1384.254 1445.574
Direct efficiency 33% 3% 3%
Direct velocity 27991.453 184832.737 208171.929
Percentage direct scans 100% 99% 99%
Page writes by reclaim 0 5082 13917
Slabs scanned 17280 29952 35328
Direct inode steals 115257 1431122 332201
Kswapd inode steals 0 0 979532
This test downloads a large tarfile and copies it with dd a number of
times - similar to the most recent bug report I've dealt with. Time to
completion is reduced. The number of pages scanned directly is still
disturbingly high with a low efficiency but this is likely due to the
number of dirty pages encountered. The figures could probably be improved
with more work around how kswapd is used and how dirty pages are handled
but that is separate work and this result is significant on its own.
Streaming Mapped Writer
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 124.47 111.67 112.64
Total Elapsed Time (seconds) 2138.14 1816.30 1867.56
MMTests Statistics: vmstat
Page Ins 90760 89124 89516
Page Outs 121028340 120199524 120736696
Swap Ins 0 86 55
Swap Outs 0 0 0
Direct pages scanned 114989363 96461439 96330619
Kswapd pages scanned 56430948 56965763 57075875
Kswapd pages reclaimed 27743219 27752044 27766606
Direct pages reclaimed 49777 46884 36655
Kswapd efficiency 49% 48% 48%
Kswapd velocity 26392.541 31363.631 30561.736
Direct efficiency 0% 0% 0%
Direct velocity 53780.091 53108.759 51581.004
Percentage direct scans 67% 62% 62%
Page writes by reclaim 385 122 1513
Slabs scanned 43008 39040 42112
Direct inode steals 0 10 8
Kswapd inode steals 733 534 477
This test just creates a large file mapping and writes to it linearly.
Time to completion is again reduced.
The gains are mostly down to two things. In many cases, there is less
scanning as zone_reclaim simply gives up faster due to recent failures.
The second reason is that memory is used more efficiently. Instead of
scanning the preferred zone every time, the allocator falls back to
another zone and uses it instead improving overall memory utilisation.
This patch: initialise ZLC for first zone eligible for zone_reclaim.
The zonelist cache (ZLC) is used among other things to record if
zone_reclaim() failed for a particular zone recently. The intention is to
avoid a high cost scanning extremely long zonelists or scanning within the
zone uselessly.
Currently the zonelist cache is setup only after the first zone has been
considered and zone_reclaim() has been called. The objective was to avoid
a costly setup but zone_reclaim is itself quite expensive. If it is
failing regularly such as the first eligible zone having mostly mapped
pages, the cost in scanning and allocation stalls is far higher than the
ZLC initialisation step.
This patch initialises ZLC before the first eligible zone calls
zone_reclaim(). Once initialised, it is checked whether the zone failed
zone_reclaim recently. If it has, the zone is skipped. As the first zone
is now being checked, additional care has to be taken about zones marked
full. A zone can be marked "full" because it should not have enough
unmapped pages for zone_reclaim but this is excessive as direct reclaim or
kswapd may succeed where zone_reclaim fails. Only mark zones "full" after
zone_reclaim fails if it failed to reclaim enough pages after scanning.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-07-26 00:12:29 +00:00
|
|
|
continue;
|
2016-07-28 22:46:32 +00:00
|
|
|
case NODE_RECLAIM_FULL:
|
2009-06-16 22:33:22 +00:00
|
|
|
/* scanned but unreclaimable */
|
mm: page allocator: initialise ZLC for first zone eligible for zone_reclaim
There have been a small number of complaints about significant stalls
while copying large amounts of data on NUMA machines reported on a
distribution bugzilla. In these cases, zone_reclaim was enabled by
default due to large NUMA distances. In general, the complaints have not
been about the workload itself unless it was a file server (in which case
the recommendation was disable zone_reclaim).
The stalls are mostly due to significant amounts of time spent scanning
the preferred zone for pages to free. After a failure, it might fallback
to another node (as zonelists are often node-ordered rather than
zone-ordered) but stall quickly again when the next allocation attempt
occurs. In bad cases, each page allocated results in a full scan of the
preferred zone.
Patch 1 checks the preferred zone for recent allocation failure
which is particularly important if zone_reclaim has failed
recently. This avoids rescanning the zone in the near future and
instead falling back to another node. This may hurt node locality
in some cases but a failure to zone_reclaim is more expensive than
a remote access.
Patch 2 clears the zlc information after direct reclaim.
Otherwise, zone_reclaim can mark zones full, direct reclaim can
reclaim enough pages but the zone is still not considered for
allocation.
This was tested on a 24-thread 2-node x86_64 machine. The tests were
focused on large amounts of IO. All tests were bound to the CPUs on
node-0 to avoid disturbances due to processes being scheduled on different
nodes. The kernels tested are
3.0-rc6-vanilla Vanilla 3.0-rc6
zlcfirst Patch 1 applied
zlcreconsider Patches 1+2 applied
FS-Mark
./fs_mark -d /tmp/fsmark-10813 -D 100 -N 5000 -n 208 -L 35 -t 24 -S0 -s 524288
fsmark-3.0-rc6 3.0-rc6 3.0-rc6
vanilla zlcfirs zlcreconsider
Files/s min 54.90 ( 0.00%) 49.80 (-10.24%) 49.10 (-11.81%)
Files/s mean 100.11 ( 0.00%) 135.17 (25.94%) 146.93 (31.87%)
Files/s stddev 57.51 ( 0.00%) 138.97 (58.62%) 158.69 (63.76%)
Files/s max 361.10 ( 0.00%) 834.40 (56.72%) 802.40 (55.00%)
Overhead min 76704.00 ( 0.00%) 76501.00 ( 0.27%) 77784.00 (-1.39%)
Overhead mean 1485356.51 ( 0.00%) 1035797.83 (43.40%) 1594680.26 (-6.86%)
Overhead stddev 1848122.53 ( 0.00%) 881489.88 (109.66%) 1772354.90 ( 4.27%)
Overhead max 7989060.00 ( 0.00%) 3369118.00 (137.13%) 10135324.00 (-21.18%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 501.49 493.91 499.93
Total Elapsed Time (seconds) 2451.57 2257.48 2215.92
MMTests Statistics: vmstat
Page Ins 46268 63840 66008
Page Outs 90821596 90671128 88043732
Swap Ins 0 0 0
Swap Outs 0 0 0
Direct pages scanned 13091697 8966863 8971790
Kswapd pages scanned 0 1830011 1831116
Kswapd pages reclaimed 0 1829068 1829930
Direct pages reclaimed 13037777 8956828 8648314
Kswapd efficiency 100% 99% 99%
Kswapd velocity 0.000 810.643 826.346
Direct efficiency 99% 99% 96%
Direct velocity 5340.128 3972.068 4048.788
Percentage direct scans 100% 83% 83%
Page writes by reclaim 0 3 0
Slabs scanned 796672 720640 720256
Direct inode steals 7422667 7160012 7088638
Kswapd inode steals 0 1736840 2021238
Test completes far faster with a large increase in the number of files
created per second. Standard deviation is high as a small number of
iterations were much higher than the mean. The number of pages scanned by
zone_reclaim is reduced and kswapd is used for more work.
LARGE DD
3.0-rc6 3.0-rc6 3.0-rc6
vanilla zlcfirst zlcreconsider
download tar 59 ( 0.00%) 59 ( 0.00%) 55 ( 7.27%)
dd source files 527 ( 0.00%) 296 (78.04%) 320 (64.69%)
delete source 36 ( 0.00%) 19 (89.47%) 20 (80.00%)
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 125.03 118.98 122.01
Total Elapsed Time (seconds) 624.56 375.02 398.06
MMTests Statistics: vmstat
Page Ins 3594216 439368 407032
Page Outs 23380832 23380488 23377444
Swap Ins 0 0 0
Swap Outs 0 436 287
Direct pages scanned 17482342 69315973 82864918
Kswapd pages scanned 0 519123 575425
Kswapd pages reclaimed 0 466501 522487
Direct pages reclaimed 5858054 2732949 2712547
Kswapd efficiency 100% 89% 90%
Kswapd velocity 0.000 1384.254 1445.574
Direct efficiency 33% 3% 3%
Direct velocity 27991.453 184832.737 208171.929
Percentage direct scans 100% 99% 99%
Page writes by reclaim 0 5082 13917
Slabs scanned 17280 29952 35328
Direct inode steals 115257 1431122 332201
Kswapd inode steals 0 0 979532
This test downloads a large tarfile and copies it with dd a number of
times - similar to the most recent bug report I've dealt with. Time to
completion is reduced. The number of pages scanned directly is still
disturbingly high with a low efficiency but this is likely due to the
number of dirty pages encountered. The figures could probably be improved
with more work around how kswapd is used and how dirty pages are handled
but that is separate work and this result is significant on its own.
Streaming Mapped Writer
MMTests Statistics: duration
User/Sys Time Running Test (seconds) 124.47 111.67 112.64
Total Elapsed Time (seconds) 2138.14 1816.30 1867.56
MMTests Statistics: vmstat
Page Ins 90760 89124 89516
Page Outs 121028340 120199524 120736696
Swap Ins 0 86 55
Swap Outs 0 0 0
Direct pages scanned 114989363 96461439 96330619
Kswapd pages scanned 56430948 56965763 57075875
Kswapd pages reclaimed 27743219 27752044 27766606
Direct pages reclaimed 49777 46884 36655
Kswapd efficiency 49% 48% 48%
Kswapd velocity 26392.541 31363.631 30561.736
Direct efficiency 0% 0% 0%
Direct velocity 53780.091 53108.759 51581.004
Percentage direct scans 67% 62% 62%
Page writes by reclaim 385 122 1513
Slabs scanned 43008 39040 42112
Direct inode steals 0 10 8
Kswapd inode steals 733 534 477
This test just creates a large file mapping and writes to it linearly.
Time to completion is again reduced.
The gains are mostly down to two things. In many cases, there is less
scanning as zone_reclaim simply gives up faster due to recent failures.
The second reason is that memory is used more efficiently. Instead of
scanning the preferred zone every time, the allocator falls back to
another zone and uses it instead improving overall memory utilisation.
This patch: initialise ZLC for first zone eligible for zone_reclaim.
The zonelist cache (ZLC) is used among other things to record if
zone_reclaim() failed for a particular zone recently. The intention is to
avoid a high cost scanning extremely long zonelists or scanning within the
zone uselessly.
Currently the zonelist cache is setup only after the first zone has been
considered and zone_reclaim() has been called. The objective was to avoid
a costly setup but zone_reclaim is itself quite expensive. If it is
failing regularly such as the first eligible zone having mostly mapped
pages, the cost in scanning and allocation stalls is far higher than the
ZLC initialisation step.
This patch initialises ZLC before the first eligible zone calls
zone_reclaim(). Once initialised, it is checked whether the zone failed
zone_reclaim recently. If it has, the zone is skipped. As the first zone
is now being checked, additional care has to be taken about zones marked
full. A zone can be marked "full" because it should not have enough
unmapped pages for zone_reclaim but this is excessive as direct reclaim or
kswapd may succeed where zone_reclaim fails. Only mark zones "full" after
zone_reclaim fails if it failed to reclaim enough pages after scanning.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-07-26 00:12:29 +00:00
|
|
|
continue;
|
2009-06-16 22:33:22 +00:00
|
|
|
default:
|
|
|
|
/* did we reclaim enough */
|
2013-04-29 22:07:57 +00:00
|
|
|
if (zone_watermark_ok(zone, order, mark,
|
2020-06-03 22:59:01 +00:00
|
|
|
ac->highest_zoneidx, alloc_flags))
|
2013-04-29 22:07:57 +00:00
|
|
|
goto try_this_zone;
|
|
|
|
|
|
|
|
continue;
|
2006-12-07 04:31:38 +00:00
|
|
|
}
|
2005-11-14 00:06:43 +00:00
|
|
|
}
|
|
|
|
|
2009-06-16 22:33:22 +00:00
|
|
|
try_this_zone:
|
2017-02-24 22:56:26 +00:00
|
|
|
page = rmqueue(ac->preferred_zoneref->zone, zone, order,
|
2015-11-07 00:28:37 +00:00
|
|
|
gfp_mask, alloc_flags, ac->migratetype);
|
mm: set page->pfmemalloc in prep_new_page()
The possibility of replacing the numerous parameters of alloc_pages*
functions with a single structure has been discussed when Minchan proposed
to expand the x86 kernel stack [1]. This series implements the change,
along with few more cleanups/microoptimizations.
The series is based on next-20150108 and I used gcc 4.8.3 20140627 on
openSUSE 13.2 for compiling. Config includess NUMA and COMPACTION.
The core change is the introduction of a new struct alloc_context, which looks
like this:
struct alloc_context {
struct zonelist *zonelist;
nodemask_t *nodemask;
struct zone *preferred_zone;
int classzone_idx;
int migratetype;
enum zone_type high_zoneidx;
};
All the contents is mostly constant, except that __alloc_pages_slowpath()
changes preferred_zone, classzone_idx and potentially zonelist. But
that's not a problem in case control returns to retry_cpuset: in
__alloc_pages_nodemask(), those will be reset to initial values again
(although it's a bit subtle). On the other hand, gfp_flags and alloc_info
mutate so much that it doesn't make sense to put them into alloc_context.
Still, the result is one parameter instead of up to 7. This is all in
Patch 2.
Patch 3 is a step to expand alloc_context usage out of page_alloc.c
itself. The function try_to_compact_pages() can also much benefit from
the parameter reduction, but it means the struct definition has to be
moved to a shared header.
Patch 1 should IMHO be included even if the rest is deemed not useful
enough. It improves maintainability and also has some code/stack
reduction. Patch 4 is OTOH a tiny optimization.
Overall bloat-o-meter results:
add/remove: 0/0 grow/shrink: 0/4 up/down: 0/-460 (-460)
function old new delta
nr_free_zone_pages 129 115 -14
__alloc_pages_direct_compact 329 256 -73
get_page_from_freelist 2670 2576 -94
__alloc_pages_nodemask 2564 2285 -279
try_to_compact_pages 582 579 -3
Overall stack sizes per ./scripts/checkstack.pl:
old new delta
get_page_from_freelist: 184 184 0
__alloc_pages_nodemask 248 200 -48
__alloc_pages_direct_c 40 - -40
try_to_compact_pages 72 72 0
-88
[1] http://marc.info/?l=linux-mm&m=140142462528257&w=2
This patch (of 4):
prep_new_page() sets almost everything in the struct page of the page
being allocated, except page->pfmemalloc. This is not obvious and has at
least once led to a bug where page->pfmemalloc was forgotten to be set
correctly, see commit 8fb74b9fb2b1 ("mm: compaction: partially revert
capture of suitable high-order page").
This patch moves the pfmemalloc setting to prep_new_page(), which means it
needs to gain alloc_flags parameter. The call to prep_new_page is moved
from buffered_rmqueue() to get_page_from_freelist(), which also leads to
simpler code. An obsolete comment for buffered_rmqueue() is replaced.
In addition to better maintainability there is a small reduction of code
and stack usage for get_page_from_freelist(), which inlines the other
functions involved.
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-145 (-145)
function old new delta
get_page_from_freelist 2670 2525 -145
Stack usage is reduced from 184 to 168 bytes.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:25:38 +00:00
|
|
|
if (page) {
|
2016-05-20 00:14:35 +00:00
|
|
|
prep_new_page(page, order, gfp_mask, alloc_flags);
|
2015-11-07 00:28:37 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is a high-order atomic allocation then check
|
|
|
|
* if the pageblock should be reserved for the future
|
|
|
|
*/
|
|
|
|
if (unlikely(order && (alloc_flags & ALLOC_HARDER)))
|
|
|
|
reserve_highatomic_pageblock(page, zone, order);
|
|
|
|
|
mm: set page->pfmemalloc in prep_new_page()
The possibility of replacing the numerous parameters of alloc_pages*
functions with a single structure has been discussed when Minchan proposed
to expand the x86 kernel stack [1]. This series implements the change,
along with few more cleanups/microoptimizations.
The series is based on next-20150108 and I used gcc 4.8.3 20140627 on
openSUSE 13.2 for compiling. Config includess NUMA and COMPACTION.
The core change is the introduction of a new struct alloc_context, which looks
like this:
struct alloc_context {
struct zonelist *zonelist;
nodemask_t *nodemask;
struct zone *preferred_zone;
int classzone_idx;
int migratetype;
enum zone_type high_zoneidx;
};
All the contents is mostly constant, except that __alloc_pages_slowpath()
changes preferred_zone, classzone_idx and potentially zonelist. But
that's not a problem in case control returns to retry_cpuset: in
__alloc_pages_nodemask(), those will be reset to initial values again
(although it's a bit subtle). On the other hand, gfp_flags and alloc_info
mutate so much that it doesn't make sense to put them into alloc_context.
Still, the result is one parameter instead of up to 7. This is all in
Patch 2.
Patch 3 is a step to expand alloc_context usage out of page_alloc.c
itself. The function try_to_compact_pages() can also much benefit from
the parameter reduction, but it means the struct definition has to be
moved to a shared header.
Patch 1 should IMHO be included even if the rest is deemed not useful
enough. It improves maintainability and also has some code/stack
reduction. Patch 4 is OTOH a tiny optimization.
Overall bloat-o-meter results:
add/remove: 0/0 grow/shrink: 0/4 up/down: 0/-460 (-460)
function old new delta
nr_free_zone_pages 129 115 -14
__alloc_pages_direct_compact 329 256 -73
get_page_from_freelist 2670 2576 -94
__alloc_pages_nodemask 2564 2285 -279
try_to_compact_pages 582 579 -3
Overall stack sizes per ./scripts/checkstack.pl:
old new delta
get_page_from_freelist: 184 184 0
__alloc_pages_nodemask 248 200 -48
__alloc_pages_direct_c 40 - -40
try_to_compact_pages 72 72 0
-88
[1] http://marc.info/?l=linux-mm&m=140142462528257&w=2
This patch (of 4):
prep_new_page() sets almost everything in the struct page of the page
being allocated, except page->pfmemalloc. This is not obvious and has at
least once led to a bug where page->pfmemalloc was forgotten to be set
correctly, see commit 8fb74b9fb2b1 ("mm: compaction: partially revert
capture of suitable high-order page").
This patch moves the pfmemalloc setting to prep_new_page(), which means it
needs to gain alloc_flags parameter. The call to prep_new_page is moved
from buffered_rmqueue() to get_page_from_freelist(), which also leads to
simpler code. An obsolete comment for buffered_rmqueue() is replaced.
In addition to better maintainability there is a small reduction of code
and stack usage for get_page_from_freelist(), which inlines the other
functions involved.
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-145 (-145)
function old new delta
get_page_from_freelist 2670 2525 -145
Stack usage is reduced from 184 to 168 bytes.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:25:38 +00:00
|
|
|
return page;
|
2018-04-05 23:22:31 +00:00
|
|
|
} else {
|
|
|
|
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
|
|
|
|
/* Try again if zone has deferred pages */
|
|
|
|
if (static_branch_unlikely(&deferred_pages)) {
|
|
|
|
if (_deferred_grow_zone(zone, order))
|
|
|
|
goto try_this_zone;
|
|
|
|
}
|
|
|
|
#endif
|
mm: set page->pfmemalloc in prep_new_page()
The possibility of replacing the numerous parameters of alloc_pages*
functions with a single structure has been discussed when Minchan proposed
to expand the x86 kernel stack [1]. This series implements the change,
along with few more cleanups/microoptimizations.
The series is based on next-20150108 and I used gcc 4.8.3 20140627 on
openSUSE 13.2 for compiling. Config includess NUMA and COMPACTION.
The core change is the introduction of a new struct alloc_context, which looks
like this:
struct alloc_context {
struct zonelist *zonelist;
nodemask_t *nodemask;
struct zone *preferred_zone;
int classzone_idx;
int migratetype;
enum zone_type high_zoneidx;
};
All the contents is mostly constant, except that __alloc_pages_slowpath()
changes preferred_zone, classzone_idx and potentially zonelist. But
that's not a problem in case control returns to retry_cpuset: in
__alloc_pages_nodemask(), those will be reset to initial values again
(although it's a bit subtle). On the other hand, gfp_flags and alloc_info
mutate so much that it doesn't make sense to put them into alloc_context.
Still, the result is one parameter instead of up to 7. This is all in
Patch 2.
Patch 3 is a step to expand alloc_context usage out of page_alloc.c
itself. The function try_to_compact_pages() can also much benefit from
the parameter reduction, but it means the struct definition has to be
moved to a shared header.
Patch 1 should IMHO be included even if the rest is deemed not useful
enough. It improves maintainability and also has some code/stack
reduction. Patch 4 is OTOH a tiny optimization.
Overall bloat-o-meter results:
add/remove: 0/0 grow/shrink: 0/4 up/down: 0/-460 (-460)
function old new delta
nr_free_zone_pages 129 115 -14
__alloc_pages_direct_compact 329 256 -73
get_page_from_freelist 2670 2576 -94
__alloc_pages_nodemask 2564 2285 -279
try_to_compact_pages 582 579 -3
Overall stack sizes per ./scripts/checkstack.pl:
old new delta
get_page_from_freelist: 184 184 0
__alloc_pages_nodemask 248 200 -48
__alloc_pages_direct_c 40 - -40
try_to_compact_pages 72 72 0
-88
[1] http://marc.info/?l=linux-mm&m=140142462528257&w=2
This patch (of 4):
prep_new_page() sets almost everything in the struct page of the page
being allocated, except page->pfmemalloc. This is not obvious and has at
least once led to a bug where page->pfmemalloc was forgotten to be set
correctly, see commit 8fb74b9fb2b1 ("mm: compaction: partially revert
capture of suitable high-order page").
This patch moves the pfmemalloc setting to prep_new_page(), which means it
needs to gain alloc_flags parameter. The call to prep_new_page is moved
from buffered_rmqueue() to get_page_from_freelist(), which also leads to
simpler code. An obsolete comment for buffered_rmqueue() is replaced.
In addition to better maintainability there is a small reduction of code
and stack usage for get_page_from_freelist(), which inlines the other
functions involved.
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-145 (-145)
function old new delta
get_page_from_freelist 2670 2525 -145
Stack usage is reduced from 184 to 168 bytes.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:25:38 +00:00
|
|
|
}
|
2008-04-28 09:12:16 +00:00
|
|
|
}
|
[PATCH] memory page_alloc zonelist caching speedup
Optimize the critical zonelist scanning for free pages in the kernel memory
allocator by caching the zones that were found to be full recently, and
skipping them.
Remembers the zones in a zonelist that were short of free memory in the
last second. And it stashes a zone-to-node table in the zonelist struct,
to optimize that conversion (minimize its cache footprint.)
Recent changes:
This differs in a significant way from a similar patch that I
posted a week ago. Now, instead of having a nodemask_t of
recently full nodes, I have a bitmask of recently full zones.
This solves a problem that last weeks patch had, which on
systems with multiple zones per node (such as DMA zone) would
take seeing any of these zones full as meaning that all zones
on that node were full.
Also I changed names - from "zonelist faster" to "zonelist cache",
as that seemed to better convey what we're doing here - caching
some of the key zonelist state (for faster access.)
See below for some performance benchmark results. After all that
discussion with David on why I didn't need them, I went and got
some ;). I wanted to verify that I had not hurt the normal case
of memory allocation noticeably. At least for my one little
microbenchmark, I found (1) the normal case wasn't affected, and
(2) workloads that forced scanning across multiple nodes for
memory improved up to 10% fewer System CPU cycles and lower
elapsed clock time ('sys' and 'real'). Good. See details, below.
I didn't have the logic in get_page_from_freelist() for various
full nodes and zone reclaim failures correct. That should be
fixed up now - notice the new goto labels zonelist_scan,
this_zone_full, and try_next_zone, in get_page_from_freelist().
There are two reasons I persued this alternative, over some earlier
proposals that would have focused on optimizing the fake numa
emulation case by caching the last useful zone:
1) Contrary to what I said before, we (SGI, on large ia64 sn2 systems)
have seen real customer loads where the cost to scan the zonelist
was a problem, due to many nodes being full of memory before
we got to a node we could use. Or at least, I think we have.
This was related to me by another engineer, based on experiences
from some time past. So this is not guaranteed. Most likely, though.
The following approach should help such real numa systems just as
much as it helps fake numa systems, or any combination thereof.
2) The effort to distinguish fake from real numa, using node_distance,
so that we could cache a fake numa node and optimize choosing
it over equivalent distance fake nodes, while continuing to
properly scan all real nodes in distance order, was going to
require a nasty blob of zonelist and node distance munging.
The following approach has no new dependency on node distances or
zone sorting.
See comment in the patch below for a description of what it actually does.
Technical details of note (or controversy):
- See the use of "zlc_active" and "did_zlc_setup" below, to delay
adding any work for this new mechanism until we've looked at the
first zone in zonelist. I figured the odds of the first zone
having the memory we needed were high enough that we should just
look there, first, then get fancy only if we need to keep looking.
- Some odd hackery was needed to add items to struct zonelist, while
not tripping up the custom zonelists built by the mm/mempolicy.c
code for MPOL_BIND. My usual wordy comments below explain this.
Search for "MPOL_BIND".
- Some per-node data in the struct zonelist is now modified frequently,
with no locking. Multiple CPU cores on a node could hit and mangle
this data. The theory is that this is just performance hint data,
and the memory allocator will work just fine despite any such mangling.
The fields at risk are the struct 'zonelist_cache' fields 'fullzones'
(a bitmask) and 'last_full_zap' (unsigned long jiffies). It should
all be self correcting after at most a one second delay.
- This still does a linear scan of the same lengths as before. All
I've optimized is making the scan faster, not algorithmically
shorter. It is now able to scan a compact array of 'unsigned
short' in the case of many full nodes, so one cache line should
cover quite a few nodes, rather than each node hitting another
one or two new and distinct cache lines.
- If both Andi and Nick don't find this too complicated, I will be
(pleasantly) flabbergasted.
- I removed the comment claiming we only use one cachline's worth of
zonelist. We seem, at least in the fake numa case, to have put the
lie to that claim.
- I pay no attention to the various watermarks and such in this performance
hint. A node could be marked full for one watermark, and then skipped
over when searching for a page using a different watermark. I think
that's actually quite ok, as it will tend to slightly increase the
spreading of memory over other nodes, away from a memory stressed node.
===============
Performance - some benchmark results and analysis:
This benchmark runs a memory hog program that uses multiple
threads to touch alot of memory as quickly as it can.
Multiple runs were made, touching 12, 38, 64 or 90 GBytes out of
the total 96 GBytes on the system, and using 1, 19, 37, or 55
threads (on a 56 CPU system.) System, user and real (elapsed)
timings were recorded for each run, shown in units of seconds,
in the table below.
Two kernels were tested - 2.6.18-mm3 and the same kernel with
this zonelist caching patch added. The table also shows the
percentage improvement the zonelist caching sys time is over
(lower than) the stock *-mm kernel.
number 2.6.18-mm3 zonelist-cache delta (< 0 good) percent
GBs N ------------ -------------- ---------------- systime
mem threads sys user real sys user real sys user real better
12 1 153 24 177 151 24 176 -2 0 -1 1%
12 19 99 22 8 99 22 8 0 0 0 0%
12 37 111 25 6 112 25 6 1 0 0 -0%
12 55 115 25 5 110 23 5 -5 -2 0 4%
38 1 502 74 576 497 73 570 -5 -1 -6 0%
38 19 426 78 48 373 76 39 -53 -2 -9 12%
38 37 544 83 36 547 82 36 3 -1 0 -0%
38 55 501 77 23 511 80 24 10 3 1 -1%
64 1 917 125 1042 890 124 1014 -27 -1 -28 2%
64 19 1118 138 119 965 141 103 -153 3 -16 13%
64 37 1202 151 94 1136 150 81 -66 -1 -13 5%
64 55 1118 141 61 1072 140 58 -46 -1 -3 4%
90 1 1342 177 1519 1275 174 1450 -67 -3 -69 4%
90 19 2392 199 192 2116 189 176 -276 -10 -16 11%
90 37 3313 238 175 2972 225 145 -341 -13 -30 10%
90 55 1948 210 104 1843 213 100 -105 3 -4 5%
Notes:
1) This test ran a memory hog program that started a specified number N of
threads, and had each thread allocate and touch 1/N'th of
the total memory to be used in the test run in a single loop,
writing a constant word to memory, one store every 4096 bytes.
Watching this test during some earlier trial runs, I would see
each of these threads sit down on one CPU and stay there, for
the remainder of the pass, a different CPU for each thread.
2) The 'real' column is not comparable to the 'sys' or 'user' columns.
The 'real' column is seconds wall clock time elapsed, from beginning
to end of that test pass. The 'sys' and 'user' columns are total
CPU seconds spent on that test pass. For a 19 thread test run,
for example, the sum of 'sys' and 'user' could be up to 19 times the
number of 'real' elapsed wall clock seconds.
3) Tests were run on a fresh, single-user boot, to minimize the amount
of memory already in use at the start of the test, and to minimize
the amount of background activity that might interfere.
4) Tests were done on a 56 CPU, 28 Node system with 96 GBytes of RAM.
5) Notice that the 'real' time gets large for the single thread runs, even
though the measured 'sys' and 'user' times are modest. I'm not sure what
that means - probably something to do with it being slow for one thread to
be accessing memory along ways away. Perhaps the fake numa system, running
ostensibly the same workload, would not show this substantial degradation
of 'real' time for one thread on many nodes -- lets hope not.
6) The high thread count passes (one thread per CPU - on 55 of 56 CPUs)
ran quite efficiently, as one might expect. Each pair of threads needed
to allocate and touch the memory on the node the two threads shared, a
pleasantly parallizable workload.
7) The intermediate thread count passes, when asking for alot of memory forcing
them to go to a few neighboring nodes, improved the most with this zonelist
caching patch.
Conclusions:
* This zonelist cache patch probably makes little difference one way or the
other for most workloads on real numa hardware, if those workloads avoid
heavy off node allocations.
* For memory intensive workloads requiring substantial off-node allocations
on real numa hardware, this patch improves both kernel and elapsed timings
up to ten per-cent.
* For fake numa systems, I'm optimistic, but will have to leave that up to
Rohit Seth to actually test (once I get him a 2.6.18 backport.)
Signed-off-by: Paul Jackson <pj@sgi.com>
Cc: Rohit Seth <rohitseth@google.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Cc: David Rientjes <rientjes@cs.washington.edu>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:31:48 +00:00
|
|
|
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
/*
|
|
|
|
* It's possible on a UMA machine to get through all zones that are
|
|
|
|
* fragmented. If avoiding fragmentation, reset and try again.
|
|
|
|
*/
|
|
|
|
if (no_fallback) {
|
|
|
|
alloc_flags &= ~ALLOC_NOFRAGMENT;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
|
2014-08-06 23:07:22 +00:00
|
|
|
return NULL;
|
2005-06-22 00:14:41 +00:00
|
|
|
}
|
|
|
|
|
2017-02-22 23:46:16 +00:00
|
|
|
static void warn_alloc_show_mem(gfp_t gfp_mask, nodemask_t *nodemask)
|
2011-05-25 00:12:16 +00:00
|
|
|
{
|
|
|
|
unsigned int filter = SHOW_MEM_FILTER_NODES;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This documents exceptions given to allocations in certain
|
|
|
|
* contexts that are allowed to allocate outside current's set
|
|
|
|
* of allowed nodes.
|
|
|
|
*/
|
|
|
|
if (!(gfp_mask & __GFP_NOMEMALLOC))
|
2017-09-06 23:24:50 +00:00
|
|
|
if (tsk_is_oom_victim(current) ||
|
2011-05-25 00:12:16 +00:00
|
|
|
(current->flags & (PF_MEMALLOC | PF_EXITING)))
|
|
|
|
filter &= ~SHOW_MEM_FILTER_NODES;
|
2015-11-07 00:28:21 +00:00
|
|
|
if (in_interrupt() || !(gfp_mask & __GFP_DIRECT_RECLAIM))
|
2011-05-25 00:12:16 +00:00
|
|
|
filter &= ~SHOW_MEM_FILTER_NODES;
|
|
|
|
|
2017-02-22 23:46:16 +00:00
|
|
|
show_mem(filter, nodemask);
|
2017-02-22 23:41:45 +00:00
|
|
|
}
|
|
|
|
|
2017-02-22 23:46:10 +00:00
|
|
|
void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...)
|
2017-02-22 23:41:45 +00:00
|
|
|
{
|
|
|
|
struct va_format vaf;
|
|
|
|
va_list args;
|
2019-11-06 05:16:51 +00:00
|
|
|
static DEFINE_RATELIMIT_STATE(nopage_rs, 10*HZ, 1);
|
2017-02-22 23:41:45 +00:00
|
|
|
|
2017-05-03 21:55:34 +00:00
|
|
|
if ((gfp_mask & __GFP_NOWARN) || !__ratelimit(&nopage_rs))
|
2017-02-22 23:41:45 +00:00
|
|
|
return;
|
|
|
|
|
2016-10-08 00:01:55 +00:00
|
|
|
va_start(args, fmt);
|
|
|
|
vaf.fmt = fmt;
|
|
|
|
vaf.va = &args;
|
mm, oom: reorganize the oom report in dump_header
OOM report contains several sections. The first one is the allocation
context that has triggered the OOM. Then we have cpuset context followed
by the stack trace of the OOM path. The tird one is the OOM memory
information. Followed by the current memory state of all system tasks.
At last, we will show oom eligible tasks and the information about the
chosen oom victim.
One thing that makes parsing more awkward than necessary is that we do not
have a single and easily parsable line about the oom context. This patch
is reorganizing the oom report to
1) who invoked oom and what was the allocation request
[ 515.902945] tuned invoked oom-killer: gfp_mask=0x6200ca(GFP_HIGHUSER_MOVABLE), order=0, oom_score_adj=0
2) OOM stack trace
[ 515.904273] CPU: 24 PID: 1809 Comm: tuned Not tainted 4.20.0-rc3+ #3
[ 515.905518] Hardware name: Inspur SA5212M4/YZMB-00370-107, BIOS 4.1.10 11/14/2016
[ 515.906821] Call Trace:
[ 515.908062] dump_stack+0x5a/0x73
[ 515.909311] dump_header+0x55/0x28c
[ 515.914260] oom_kill_process+0x2d8/0x300
[ 515.916708] out_of_memory+0x145/0x4a0
[ 515.917932] __alloc_pages_slowpath+0x7d2/0xa16
[ 515.919157] __alloc_pages_nodemask+0x277/0x290
[ 515.920367] filemap_fault+0x3d0/0x6c0
[ 515.921529] ? filemap_map_pages+0x2b8/0x420
[ 515.922709] ext4_filemap_fault+0x2c/0x40 [ext4]
[ 515.923884] __do_fault+0x20/0x80
[ 515.925032] __handle_mm_fault+0xbc0/0xe80
[ 515.926195] handle_mm_fault+0xfa/0x210
[ 515.927357] __do_page_fault+0x233/0x4c0
[ 515.928506] do_page_fault+0x32/0x140
[ 515.929646] ? page_fault+0x8/0x30
[ 515.930770] page_fault+0x1e/0x30
3) OOM memory information
[ 515.958093] Mem-Info:
[ 515.959647] active_anon:26501758 inactive_anon:1179809 isolated_anon:0
active_file:4402672 inactive_file:483963 isolated_file:1344
unevictable:0 dirty:4886753 writeback:0 unstable:0
slab_reclaimable:148442 slab_unreclaimable:18741
mapped:1347 shmem:1347 pagetables:58669 bounce:0
free:88663 free_pcp:0 free_cma:0
...
4) current memory state of all system tasks
[ 516.079544] [ 744] 0 744 9211 1345 114688 82 0 systemd-journal
[ 516.082034] [ 787] 0 787 31764 0 143360 92 0 lvmetad
[ 516.084465] [ 792] 0 792 10930 1 110592 208 -1000 systemd-udevd
[ 516.086865] [ 1199] 0 1199 13866 0 131072 112 -1000 auditd
[ 516.089190] [ 1222] 0 1222 31990 1 110592 157 0 smartd
[ 516.091477] [ 1225] 0 1225 4864 85 81920 43 0 irqbalance
[ 516.093712] [ 1226] 0 1226 52612 0 258048 426 0 abrtd
[ 516.112128] [ 1280] 0 1280 109774 55 299008 400 0 NetworkManager
[ 516.113998] [ 1295] 0 1295 28817 37 69632 24 0 ksmtuned
[ 516.144596] [ 10718] 0 10718 2622484 1721372 15998976 267219 0 panic
[ 516.145792] [ 10719] 0 10719 2622484 1164767 9818112 53576 0 panic
[ 516.146977] [ 10720] 0 10720 2622484 1174361 9904128 53709 0 panic
[ 516.148163] [ 10721] 0 10721 2622484 1209070 10194944 54824 0 panic
[ 516.149329] [ 10722] 0 10722 2622484 1745799 14774272 91138 0 panic
5) oom context (contrains and the chosen victim).
oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0-1,task=panic,pid=10737,uid=0
An admin can easily get the full oom context at a single line which
makes parsing much easier.
Link: http://lkml.kernel.org/r/1542799799-36184-1-git-send-email-ufo19890607@gmail.com
Signed-off-by: yuzhoujian <yuzhoujian@didichuxing.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: Yang Shi <yang.s@alibaba-inc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:36:07 +00:00
|
|
|
pr_warn("%s: %pV, mode:%#x(%pGg), nodemask=%*pbl",
|
2017-11-16 01:39:14 +00:00
|
|
|
current->comm, &vaf, gfp_mask, &gfp_mask,
|
|
|
|
nodemask_pr_args(nodemask));
|
2016-10-08 00:01:55 +00:00
|
|
|
va_end(args);
|
2011-11-01 00:08:35 +00:00
|
|
|
|
2017-02-22 23:46:10 +00:00
|
|
|
cpuset_print_current_mems_allowed();
|
mm, oom: reorganize the oom report in dump_header
OOM report contains several sections. The first one is the allocation
context that has triggered the OOM. Then we have cpuset context followed
by the stack trace of the OOM path. The tird one is the OOM memory
information. Followed by the current memory state of all system tasks.
At last, we will show oom eligible tasks and the information about the
chosen oom victim.
One thing that makes parsing more awkward than necessary is that we do not
have a single and easily parsable line about the oom context. This patch
is reorganizing the oom report to
1) who invoked oom and what was the allocation request
[ 515.902945] tuned invoked oom-killer: gfp_mask=0x6200ca(GFP_HIGHUSER_MOVABLE), order=0, oom_score_adj=0
2) OOM stack trace
[ 515.904273] CPU: 24 PID: 1809 Comm: tuned Not tainted 4.20.0-rc3+ #3
[ 515.905518] Hardware name: Inspur SA5212M4/YZMB-00370-107, BIOS 4.1.10 11/14/2016
[ 515.906821] Call Trace:
[ 515.908062] dump_stack+0x5a/0x73
[ 515.909311] dump_header+0x55/0x28c
[ 515.914260] oom_kill_process+0x2d8/0x300
[ 515.916708] out_of_memory+0x145/0x4a0
[ 515.917932] __alloc_pages_slowpath+0x7d2/0xa16
[ 515.919157] __alloc_pages_nodemask+0x277/0x290
[ 515.920367] filemap_fault+0x3d0/0x6c0
[ 515.921529] ? filemap_map_pages+0x2b8/0x420
[ 515.922709] ext4_filemap_fault+0x2c/0x40 [ext4]
[ 515.923884] __do_fault+0x20/0x80
[ 515.925032] __handle_mm_fault+0xbc0/0xe80
[ 515.926195] handle_mm_fault+0xfa/0x210
[ 515.927357] __do_page_fault+0x233/0x4c0
[ 515.928506] do_page_fault+0x32/0x140
[ 515.929646] ? page_fault+0x8/0x30
[ 515.930770] page_fault+0x1e/0x30
3) OOM memory information
[ 515.958093] Mem-Info:
[ 515.959647] active_anon:26501758 inactive_anon:1179809 isolated_anon:0
active_file:4402672 inactive_file:483963 isolated_file:1344
unevictable:0 dirty:4886753 writeback:0 unstable:0
slab_reclaimable:148442 slab_unreclaimable:18741
mapped:1347 shmem:1347 pagetables:58669 bounce:0
free:88663 free_pcp:0 free_cma:0
...
4) current memory state of all system tasks
[ 516.079544] [ 744] 0 744 9211 1345 114688 82 0 systemd-journal
[ 516.082034] [ 787] 0 787 31764 0 143360 92 0 lvmetad
[ 516.084465] [ 792] 0 792 10930 1 110592 208 -1000 systemd-udevd
[ 516.086865] [ 1199] 0 1199 13866 0 131072 112 -1000 auditd
[ 516.089190] [ 1222] 0 1222 31990 1 110592 157 0 smartd
[ 516.091477] [ 1225] 0 1225 4864 85 81920 43 0 irqbalance
[ 516.093712] [ 1226] 0 1226 52612 0 258048 426 0 abrtd
[ 516.112128] [ 1280] 0 1280 109774 55 299008 400 0 NetworkManager
[ 516.113998] [ 1295] 0 1295 28817 37 69632 24 0 ksmtuned
[ 516.144596] [ 10718] 0 10718 2622484 1721372 15998976 267219 0 panic
[ 516.145792] [ 10719] 0 10719 2622484 1164767 9818112 53576 0 panic
[ 516.146977] [ 10720] 0 10720 2622484 1174361 9904128 53709 0 panic
[ 516.148163] [ 10721] 0 10721 2622484 1209070 10194944 54824 0 panic
[ 516.149329] [ 10722] 0 10722 2622484 1745799 14774272 91138 0 panic
5) oom context (contrains and the chosen victim).
oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0-1,task=panic,pid=10737,uid=0
An admin can easily get the full oom context at a single line which
makes parsing much easier.
Link: http://lkml.kernel.org/r/1542799799-36184-1-git-send-email-ufo19890607@gmail.com
Signed-off-by: yuzhoujian <yuzhoujian@didichuxing.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: Yang Shi <yang.s@alibaba-inc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:36:07 +00:00
|
|
|
pr_cont("\n");
|
2011-05-25 00:12:16 +00:00
|
|
|
dump_stack();
|
2017-02-22 23:46:28 +00:00
|
|
|
warn_alloc_show_mem(gfp_mask, nodemask);
|
2011-05-25 00:12:16 +00:00
|
|
|
}
|
|
|
|
|
2017-02-22 23:46:25 +00:00
|
|
|
static inline struct page *
|
|
|
|
__alloc_pages_cpuset_fallback(gfp_t gfp_mask, unsigned int order,
|
|
|
|
unsigned int alloc_flags,
|
|
|
|
const struct alloc_context *ac)
|
|
|
|
{
|
|
|
|
struct page *page;
|
|
|
|
|
|
|
|
page = get_page_from_freelist(gfp_mask, order,
|
|
|
|
alloc_flags|ALLOC_CPUSET, ac);
|
|
|
|
/*
|
|
|
|
* fallback to ignore cpuset restriction if our nodes
|
|
|
|
* are depleted
|
|
|
|
*/
|
|
|
|
if (!page)
|
|
|
|
page = get_page_from_freelist(gfp_mask, order,
|
|
|
|
alloc_flags, ac);
|
|
|
|
|
|
|
|
return page;
|
|
|
|
}
|
|
|
|
|
2009-06-16 22:31:57 +00:00
|
|
|
static inline struct page *
|
|
|
|
__alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
|
2015-02-11 23:25:41 +00:00
|
|
|
const struct alloc_context *ac, unsigned long *did_some_progress)
|
2009-06-16 22:31:57 +00:00
|
|
|
{
|
2015-09-08 22:00:36 +00:00
|
|
|
struct oom_control oc = {
|
|
|
|
.zonelist = ac->zonelist,
|
|
|
|
.nodemask = ac->nodemask,
|
2016-07-26 22:22:33 +00:00
|
|
|
.memcg = NULL,
|
2015-09-08 22:00:36 +00:00
|
|
|
.gfp_mask = gfp_mask,
|
|
|
|
.order = order,
|
|
|
|
};
|
2009-06-16 22:31:57 +00:00
|
|
|
struct page *page;
|
|
|
|
|
2015-01-26 20:58:32 +00:00
|
|
|
*did_some_progress = 0;
|
|
|
|
|
|
|
|
/*
|
2015-06-24 23:57:19 +00:00
|
|
|
* Acquire the oom lock. If that fails, somebody else is
|
|
|
|
* making progress for us.
|
2015-01-26 20:58:32 +00:00
|
|
|
*/
|
2015-06-24 23:57:19 +00:00
|
|
|
if (!mutex_trylock(&oom_lock)) {
|
2015-01-26 20:58:32 +00:00
|
|
|
*did_some_progress = 1;
|
2009-06-16 22:31:57 +00:00
|
|
|
schedule_timeout_uninterruptible(1);
|
2005-04-16 22:20:36 +00:00
|
|
|
return NULL;
|
|
|
|
}
|
2005-11-17 20:35:02 +00:00
|
|
|
|
2009-06-16 22:31:57 +00:00
|
|
|
/*
|
|
|
|
* Go through the zonelist yet one more time, keep very high watermark
|
|
|
|
* here, this is only to catch a parallel oom killing, we must fail if
|
2017-08-31 23:15:20 +00:00
|
|
|
* we're still under heavy pressure. But make sure that this reclaim
|
|
|
|
* attempt shall not depend on __GFP_DIRECT_RECLAIM && !__GFP_NORETRY
|
|
|
|
* allocation which will never fail due to oom_lock already held.
|
2009-06-16 22:31:57 +00:00
|
|
|
*/
|
2017-08-31 23:15:20 +00:00
|
|
|
page = get_page_from_freelist((gfp_mask | __GFP_HARDWALL) &
|
|
|
|
~__GFP_DIRECT_RECLAIM, order,
|
|
|
|
ALLOC_WMARK_HIGH|ALLOC_CPUSET, ac);
|
2005-11-14 00:06:43 +00:00
|
|
|
if (page)
|
2009-06-16 22:31:57 +00:00
|
|
|
goto out;
|
|
|
|
|
2017-02-22 23:46:22 +00:00
|
|
|
/* Coredumps can quickly deplete all memory reserves */
|
|
|
|
if (current->flags & PF_DUMPCORE)
|
|
|
|
goto out;
|
|
|
|
/* The OOM killer will not help higher order allocs */
|
|
|
|
if (order > PAGE_ALLOC_COSTLY_ORDER)
|
|
|
|
goto out;
|
2017-07-12 21:36:45 +00:00
|
|
|
/*
|
|
|
|
* We have already exhausted all our reclaim opportunities without any
|
|
|
|
* success so it is time to admit defeat. We will skip the OOM killer
|
|
|
|
* because it is very likely that the caller has a more reasonable
|
|
|
|
* fallback than shooting a random task.
|
2020-10-13 23:55:45 +00:00
|
|
|
*
|
|
|
|
* The OOM killer may not free memory on a specific node.
|
2017-07-12 21:36:45 +00:00
|
|
|
*/
|
2020-10-13 23:55:45 +00:00
|
|
|
if (gfp_mask & (__GFP_RETRY_MAYFAIL | __GFP_THISNODE))
|
2017-07-12 21:36:45 +00:00
|
|
|
goto out;
|
2017-02-22 23:46:22 +00:00
|
|
|
/* The OOM killer does not needlessly kill tasks for lowmem */
|
2020-06-03 22:59:01 +00:00
|
|
|
if (ac->highest_zoneidx < ZONE_NORMAL)
|
2017-02-22 23:46:22 +00:00
|
|
|
goto out;
|
|
|
|
if (pm_suspended_storage())
|
|
|
|
goto out;
|
|
|
|
/*
|
|
|
|
* XXX: GFP_NOFS allocations should rather fail than rely on
|
|
|
|
* other request to make a forward progress.
|
|
|
|
* We are in an unfortunate situation where out_of_memory cannot
|
|
|
|
* do much for this context but let's try it to at least get
|
|
|
|
* access to memory reserved if the current task is killed (see
|
|
|
|
* out_of_memory). Once filesystems are ready to handle allocation
|
|
|
|
* failures more gracefully we should just bail out here.
|
|
|
|
*/
|
|
|
|
|
2018-02-01 00:20:07 +00:00
|
|
|
/* Exhausted what can be done so it's blame time */
|
2016-01-14 23:20:36 +00:00
|
|
|
if (out_of_memory(&oc) || WARN_ON_ONCE(gfp_mask & __GFP_NOFAIL)) {
|
2015-02-11 23:26:24 +00:00
|
|
|
*did_some_progress = 1;
|
2016-01-14 23:20:36 +00:00
|
|
|
|
2017-02-22 23:46:25 +00:00
|
|
|
/*
|
|
|
|
* Help non-failing allocations by giving them access to memory
|
|
|
|
* reserves
|
|
|
|
*/
|
|
|
|
if (gfp_mask & __GFP_NOFAIL)
|
|
|
|
page = __alloc_pages_cpuset_fallback(gfp_mask, order,
|
2016-01-14 23:20:36 +00:00
|
|
|
ALLOC_NO_WATERMARKS, ac);
|
|
|
|
}
|
2009-06-16 22:31:57 +00:00
|
|
|
out:
|
2015-06-24 23:57:19 +00:00
|
|
|
mutex_unlock(&oom_lock);
|
2009-06-16 22:31:57 +00:00
|
|
|
return page;
|
|
|
|
}
|
|
|
|
|
2016-05-20 23:57:06 +00:00
|
|
|
/*
|
|
|
|
* Maximum number of compaction retries wit a progress before OOM
|
|
|
|
* killer is consider as the only way to move forward.
|
|
|
|
*/
|
|
|
|
#define MAX_COMPACT_RETRIES 16
|
|
|
|
|
2010-05-24 21:32:30 +00:00
|
|
|
#ifdef CONFIG_COMPACTION
|
|
|
|
/* Try memory compaction for high-order allocations before reclaim */
|
|
|
|
static struct page *
|
|
|
|
__alloc_pages_direct_compact(gfp_t gfp_mask, unsigned int order,
|
2016-05-20 00:13:38 +00:00
|
|
|
unsigned int alloc_flags, const struct alloc_context *ac,
|
2016-07-28 22:49:28 +00:00
|
|
|
enum compact_priority prio, enum compact_result *compact_result)
|
2010-05-24 21:32:30 +00:00
|
|
|
{
|
2019-03-05 23:45:41 +00:00
|
|
|
struct page *page = NULL;
|
psi: pressure stall information for CPU, memory, and IO
When systems are overcommitted and resources become contended, it's hard
to tell exactly the impact this has on workload productivity, or how close
the system is to lockups and OOM kills. In particular, when machines work
multiple jobs concurrently, the impact of overcommit in terms of latency
and throughput on the individual job can be enormous.
In order to maximize hardware utilization without sacrificing individual
job health or risk complete machine lockups, this patch implements a way
to quantify resource pressure in the system.
A kernel built with CONFIG_PSI=y creates files in /proc/pressure/ that
expose the percentage of time the system is stalled on CPU, memory, or IO,
respectively. Stall states are aggregate versions of the per-task delay
accounting delays:
cpu: some tasks are runnable but not executing on a CPU
memory: tasks are reclaiming, or waiting for swapin or thrashing cache
io: tasks are waiting for io completions
These percentages of walltime can be thought of as pressure percentages,
and they give a general sense of system health and productivity loss
incurred by resource overcommit. They can also indicate when the system
is approaching lockup scenarios and OOMs.
To do this, psi keeps track of the task states associated with each CPU
and samples the time they spend in stall states. Every 2 seconds, the
samples are averaged across CPUs - weighted by the CPUs' non-idle time to
eliminate artifacts from unused CPUs - and translated into percentages of
walltime. A running average of those percentages is maintained over 10s,
1m, and 5m periods (similar to the loadaverage).
[hannes@cmpxchg.org: doc fixlet, per Randy]
Link: http://lkml.kernel.org/r/20180828205625.GA14030@cmpxchg.org
[hannes@cmpxchg.org: code optimization]
Link: http://lkml.kernel.org/r/20180907175015.GA8479@cmpxchg.org
[hannes@cmpxchg.org: rename psi_clock() to psi_update_work(), per Peter]
Link: http://lkml.kernel.org/r/20180907145404.GB11088@cmpxchg.org
[hannes@cmpxchg.org: fix build]
Link: http://lkml.kernel.org/r/20180913014222.GA2370@cmpxchg.org
Link: http://lkml.kernel.org/r/20180828172258.3185-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Daniel Drake <drake@endlessm.com>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:06:27 +00:00
|
|
|
unsigned long pflags;
|
2017-05-08 22:59:50 +00:00
|
|
|
unsigned int noreclaim_flag;
|
mm, compaction: defer each zone individually instead of preferred zone
When direct sync compaction is often unsuccessful, it may become deferred
for some time to avoid further useless attempts, both sync and async.
Successful high-order allocations un-defer compaction, while further
unsuccessful compaction attempts prolong the compaction deferred period.
Currently the checking and setting deferred status is performed only on
the preferred zone of the allocation that invoked direct compaction. But
compaction itself is attempted on all eligible zones in the zonelist, so
the behavior is suboptimal and may lead both to scenarios where 1)
compaction is attempted uselessly, or 2) where it's not attempted despite
good chances of succeeding, as shown on the examples below:
1) A direct compaction with Normal preferred zone failed and set
deferred compaction for the Normal zone. Another unrelated direct
compaction with DMA32 as preferred zone will attempt to compact DMA32
zone even though the first compaction attempt also included DMA32 zone.
In another scenario, compaction with Normal preferred zone failed to
compact Normal zone, but succeeded in the DMA32 zone, so it will not
defer compaction. In the next attempt, it will try Normal zone which
will fail again, instead of skipping Normal zone and trying DMA32
directly.
2) Kswapd will balance DMA32 zone and reset defer status based on
watermarks looking good. A direct compaction with preferred Normal
zone will skip compaction of all zones including DMA32 because Normal
was still deferred. The allocation might have succeeded in DMA32, but
won't.
This patch makes compaction deferring work on individual zone basis
instead of preferred zone. For each zone, it checks compaction_deferred()
to decide if the zone should be skipped. If watermarks fail after
compacting the zone, defer_compaction() is called. The zone where
watermarks passed can still be deferred when the allocation attempt is
unsuccessful. When allocation is successful, compaction_defer_reset() is
called for the zone containing the allocated page. This approach should
approximate calling defer_compaction() only on zones where compaction was
attempted and did not yield allocated page. There might be corner cases
but that is inevitable as long as the decision to stop compacting dues not
guarantee that a page will be allocated.
Due to a new COMPACT_DEFERRED return value, some functions relying
implicitly on COMPACT_SKIPPED = 0 had to be updated, with comments made
more accurate. The did_some_progress output parameter of
__alloc_pages_direct_compact() is removed completely, as the caller
actually does not use it after compaction sets it - it is only considered
when direct reclaim sets it.
During testing on a two-node machine with a single very small Normal zone
on node 1, this patch has improved success rates in stress-highalloc
mmtests benchmark. The success here were previously made worse by commit
3a025760fc15 ("mm: page_alloc: spill to remote nodes before waking
kswapd") as kswapd was no longer resetting often enough the deferred
compaction for the Normal zone, and DMA32 zones on both nodes were thus
not considered for compaction. On different machine, success rates were
improved with __GFP_NO_KSWAPD allocations.
[akpm@linux-foundation.org: fix CONFIG_COMPACTION=n build]
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-09 22:27:02 +00:00
|
|
|
|
|
|
|
if (!order)
|
2012-01-13 01:19:41 +00:00
|
|
|
return NULL;
|
|
|
|
|
psi: pressure stall information for CPU, memory, and IO
When systems are overcommitted and resources become contended, it's hard
to tell exactly the impact this has on workload productivity, or how close
the system is to lockups and OOM kills. In particular, when machines work
multiple jobs concurrently, the impact of overcommit in terms of latency
and throughput on the individual job can be enormous.
In order to maximize hardware utilization without sacrificing individual
job health or risk complete machine lockups, this patch implements a way
to quantify resource pressure in the system.
A kernel built with CONFIG_PSI=y creates files in /proc/pressure/ that
expose the percentage of time the system is stalled on CPU, memory, or IO,
respectively. Stall states are aggregate versions of the per-task delay
accounting delays:
cpu: some tasks are runnable but not executing on a CPU
memory: tasks are reclaiming, or waiting for swapin or thrashing cache
io: tasks are waiting for io completions
These percentages of walltime can be thought of as pressure percentages,
and they give a general sense of system health and productivity loss
incurred by resource overcommit. They can also indicate when the system
is approaching lockup scenarios and OOMs.
To do this, psi keeps track of the task states associated with each CPU
and samples the time they spend in stall states. Every 2 seconds, the
samples are averaged across CPUs - weighted by the CPUs' non-idle time to
eliminate artifacts from unused CPUs - and translated into percentages of
walltime. A running average of those percentages is maintained over 10s,
1m, and 5m periods (similar to the loadaverage).
[hannes@cmpxchg.org: doc fixlet, per Randy]
Link: http://lkml.kernel.org/r/20180828205625.GA14030@cmpxchg.org
[hannes@cmpxchg.org: code optimization]
Link: http://lkml.kernel.org/r/20180907175015.GA8479@cmpxchg.org
[hannes@cmpxchg.org: rename psi_clock() to psi_update_work(), per Peter]
Link: http://lkml.kernel.org/r/20180907145404.GB11088@cmpxchg.org
[hannes@cmpxchg.org: fix build]
Link: http://lkml.kernel.org/r/20180913014222.GA2370@cmpxchg.org
Link: http://lkml.kernel.org/r/20180828172258.3185-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Daniel Drake <drake@endlessm.com>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:06:27 +00:00
|
|
|
psi_memstall_enter(&pflags);
|
2017-05-08 22:59:50 +00:00
|
|
|
noreclaim_flag = memalloc_noreclaim_save();
|
psi: pressure stall information for CPU, memory, and IO
When systems are overcommitted and resources become contended, it's hard
to tell exactly the impact this has on workload productivity, or how close
the system is to lockups and OOM kills. In particular, when machines work
multiple jobs concurrently, the impact of overcommit in terms of latency
and throughput on the individual job can be enormous.
In order to maximize hardware utilization without sacrificing individual
job health or risk complete machine lockups, this patch implements a way
to quantify resource pressure in the system.
A kernel built with CONFIG_PSI=y creates files in /proc/pressure/ that
expose the percentage of time the system is stalled on CPU, memory, or IO,
respectively. Stall states are aggregate versions of the per-task delay
accounting delays:
cpu: some tasks are runnable but not executing on a CPU
memory: tasks are reclaiming, or waiting for swapin or thrashing cache
io: tasks are waiting for io completions
These percentages of walltime can be thought of as pressure percentages,
and they give a general sense of system health and productivity loss
incurred by resource overcommit. They can also indicate when the system
is approaching lockup scenarios and OOMs.
To do this, psi keeps track of the task states associated with each CPU
and samples the time they spend in stall states. Every 2 seconds, the
samples are averaged across CPUs - weighted by the CPUs' non-idle time to
eliminate artifacts from unused CPUs - and translated into percentages of
walltime. A running average of those percentages is maintained over 10s,
1m, and 5m periods (similar to the loadaverage).
[hannes@cmpxchg.org: doc fixlet, per Randy]
Link: http://lkml.kernel.org/r/20180828205625.GA14030@cmpxchg.org
[hannes@cmpxchg.org: code optimization]
Link: http://lkml.kernel.org/r/20180907175015.GA8479@cmpxchg.org
[hannes@cmpxchg.org: rename psi_clock() to psi_update_work(), per Peter]
Link: http://lkml.kernel.org/r/20180907145404.GB11088@cmpxchg.org
[hannes@cmpxchg.org: fix build]
Link: http://lkml.kernel.org/r/20180913014222.GA2370@cmpxchg.org
Link: http://lkml.kernel.org/r/20180828172258.3185-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Daniel Drake <drake@endlessm.com>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:06:27 +00:00
|
|
|
|
2016-05-20 23:56:53 +00:00
|
|
|
*compact_result = try_to_compact_pages(gfp_mask, order, alloc_flags, ac,
|
2019-03-05 23:45:41 +00:00
|
|
|
prio, &page);
|
psi: pressure stall information for CPU, memory, and IO
When systems are overcommitted and resources become contended, it's hard
to tell exactly the impact this has on workload productivity, or how close
the system is to lockups and OOM kills. In particular, when machines work
multiple jobs concurrently, the impact of overcommit in terms of latency
and throughput on the individual job can be enormous.
In order to maximize hardware utilization without sacrificing individual
job health or risk complete machine lockups, this patch implements a way
to quantify resource pressure in the system.
A kernel built with CONFIG_PSI=y creates files in /proc/pressure/ that
expose the percentage of time the system is stalled on CPU, memory, or IO,
respectively. Stall states are aggregate versions of the per-task delay
accounting delays:
cpu: some tasks are runnable but not executing on a CPU
memory: tasks are reclaiming, or waiting for swapin or thrashing cache
io: tasks are waiting for io completions
These percentages of walltime can be thought of as pressure percentages,
and they give a general sense of system health and productivity loss
incurred by resource overcommit. They can also indicate when the system
is approaching lockup scenarios and OOMs.
To do this, psi keeps track of the task states associated with each CPU
and samples the time they spend in stall states. Every 2 seconds, the
samples are averaged across CPUs - weighted by the CPUs' non-idle time to
eliminate artifacts from unused CPUs - and translated into percentages of
walltime. A running average of those percentages is maintained over 10s,
1m, and 5m periods (similar to the loadaverage).
[hannes@cmpxchg.org: doc fixlet, per Randy]
Link: http://lkml.kernel.org/r/20180828205625.GA14030@cmpxchg.org
[hannes@cmpxchg.org: code optimization]
Link: http://lkml.kernel.org/r/20180907175015.GA8479@cmpxchg.org
[hannes@cmpxchg.org: rename psi_clock() to psi_update_work(), per Peter]
Link: http://lkml.kernel.org/r/20180907145404.GB11088@cmpxchg.org
[hannes@cmpxchg.org: fix build]
Link: http://lkml.kernel.org/r/20180913014222.GA2370@cmpxchg.org
Link: http://lkml.kernel.org/r/20180828172258.3185-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Daniel Drake <drake@endlessm.com>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:06:27 +00:00
|
|
|
|
2017-05-08 22:59:50 +00:00
|
|
|
memalloc_noreclaim_restore(noreclaim_flag);
|
psi: pressure stall information for CPU, memory, and IO
When systems are overcommitted and resources become contended, it's hard
to tell exactly the impact this has on workload productivity, or how close
the system is to lockups and OOM kills. In particular, when machines work
multiple jobs concurrently, the impact of overcommit in terms of latency
and throughput on the individual job can be enormous.
In order to maximize hardware utilization without sacrificing individual
job health or risk complete machine lockups, this patch implements a way
to quantify resource pressure in the system.
A kernel built with CONFIG_PSI=y creates files in /proc/pressure/ that
expose the percentage of time the system is stalled on CPU, memory, or IO,
respectively. Stall states are aggregate versions of the per-task delay
accounting delays:
cpu: some tasks are runnable but not executing on a CPU
memory: tasks are reclaiming, or waiting for swapin or thrashing cache
io: tasks are waiting for io completions
These percentages of walltime can be thought of as pressure percentages,
and they give a general sense of system health and productivity loss
incurred by resource overcommit. They can also indicate when the system
is approaching lockup scenarios and OOMs.
To do this, psi keeps track of the task states associated with each CPU
and samples the time they spend in stall states. Every 2 seconds, the
samples are averaged across CPUs - weighted by the CPUs' non-idle time to
eliminate artifacts from unused CPUs - and translated into percentages of
walltime. A running average of those percentages is maintained over 10s,
1m, and 5m periods (similar to the loadaverage).
[hannes@cmpxchg.org: doc fixlet, per Randy]
Link: http://lkml.kernel.org/r/20180828205625.GA14030@cmpxchg.org
[hannes@cmpxchg.org: code optimization]
Link: http://lkml.kernel.org/r/20180907175015.GA8479@cmpxchg.org
[hannes@cmpxchg.org: rename psi_clock() to psi_update_work(), per Peter]
Link: http://lkml.kernel.org/r/20180907145404.GB11088@cmpxchg.org
[hannes@cmpxchg.org: fix build]
Link: http://lkml.kernel.org/r/20180913014222.GA2370@cmpxchg.org
Link: http://lkml.kernel.org/r/20180828172258.3185-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Daniel Drake <drake@endlessm.com>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:06:27 +00:00
|
|
|
psi_memstall_leave(&pflags);
|
2010-05-24 21:32:30 +00:00
|
|
|
|
2014-10-09 22:27:04 +00:00
|
|
|
/*
|
|
|
|
* At least in one zone compaction wasn't deferred or skipped, so let's
|
|
|
|
* count a compaction stall
|
|
|
|
*/
|
|
|
|
count_vm_event(COMPACTSTALL);
|
2013-01-11 22:32:16 +00:00
|
|
|
|
2019-03-05 23:45:41 +00:00
|
|
|
/* Prep a captured page if available */
|
|
|
|
if (page)
|
|
|
|
prep_new_page(page, order, gfp_mask, alloc_flags);
|
|
|
|
|
|
|
|
/* Try get a page from the freelist if available */
|
|
|
|
if (!page)
|
|
|
|
page = get_page_from_freelist(gfp_mask, order, alloc_flags, ac);
|
mm, compaction: defer each zone individually instead of preferred zone
When direct sync compaction is often unsuccessful, it may become deferred
for some time to avoid further useless attempts, both sync and async.
Successful high-order allocations un-defer compaction, while further
unsuccessful compaction attempts prolong the compaction deferred period.
Currently the checking and setting deferred status is performed only on
the preferred zone of the allocation that invoked direct compaction. But
compaction itself is attempted on all eligible zones in the zonelist, so
the behavior is suboptimal and may lead both to scenarios where 1)
compaction is attempted uselessly, or 2) where it's not attempted despite
good chances of succeeding, as shown on the examples below:
1) A direct compaction with Normal preferred zone failed and set
deferred compaction for the Normal zone. Another unrelated direct
compaction with DMA32 as preferred zone will attempt to compact DMA32
zone even though the first compaction attempt also included DMA32 zone.
In another scenario, compaction with Normal preferred zone failed to
compact Normal zone, but succeeded in the DMA32 zone, so it will not
defer compaction. In the next attempt, it will try Normal zone which
will fail again, instead of skipping Normal zone and trying DMA32
directly.
2) Kswapd will balance DMA32 zone and reset defer status based on
watermarks looking good. A direct compaction with preferred Normal
zone will skip compaction of all zones including DMA32 because Normal
was still deferred. The allocation might have succeeded in DMA32, but
won't.
This patch makes compaction deferring work on individual zone basis
instead of preferred zone. For each zone, it checks compaction_deferred()
to decide if the zone should be skipped. If watermarks fail after
compacting the zone, defer_compaction() is called. The zone where
watermarks passed can still be deferred when the allocation attempt is
unsuccessful. When allocation is successful, compaction_defer_reset() is
called for the zone containing the allocated page. This approach should
approximate calling defer_compaction() only on zones where compaction was
attempted and did not yield allocated page. There might be corner cases
but that is inevitable as long as the decision to stop compacting dues not
guarantee that a page will be allocated.
Due to a new COMPACT_DEFERRED return value, some functions relying
implicitly on COMPACT_SKIPPED = 0 had to be updated, with comments made
more accurate. The did_some_progress output parameter of
__alloc_pages_direct_compact() is removed completely, as the caller
actually does not use it after compaction sets it - it is only considered
when direct reclaim sets it.
During testing on a two-node machine with a single very small Normal zone
on node 1, this patch has improved success rates in stress-highalloc
mmtests benchmark. The success here were previously made worse by commit
3a025760fc15 ("mm: page_alloc: spill to remote nodes before waking
kswapd") as kswapd was no longer resetting often enough the deferred
compaction for the Normal zone, and DMA32 zones on both nodes were thus
not considered for compaction. On different machine, success rates were
improved with __GFP_NO_KSWAPD allocations.
[akpm@linux-foundation.org: fix CONFIG_COMPACTION=n build]
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-09 22:27:02 +00:00
|
|
|
|
2014-10-09 22:27:04 +00:00
|
|
|
if (page) {
|
|
|
|
struct zone *zone = page_zone(page);
|
mm, compaction: defer each zone individually instead of preferred zone
When direct sync compaction is often unsuccessful, it may become deferred
for some time to avoid further useless attempts, both sync and async.
Successful high-order allocations un-defer compaction, while further
unsuccessful compaction attempts prolong the compaction deferred period.
Currently the checking and setting deferred status is performed only on
the preferred zone of the allocation that invoked direct compaction. But
compaction itself is attempted on all eligible zones in the zonelist, so
the behavior is suboptimal and may lead both to scenarios where 1)
compaction is attempted uselessly, or 2) where it's not attempted despite
good chances of succeeding, as shown on the examples below:
1) A direct compaction with Normal preferred zone failed and set
deferred compaction for the Normal zone. Another unrelated direct
compaction with DMA32 as preferred zone will attempt to compact DMA32
zone even though the first compaction attempt also included DMA32 zone.
In another scenario, compaction with Normal preferred zone failed to
compact Normal zone, but succeeded in the DMA32 zone, so it will not
defer compaction. In the next attempt, it will try Normal zone which
will fail again, instead of skipping Normal zone and trying DMA32
directly.
2) Kswapd will balance DMA32 zone and reset defer status based on
watermarks looking good. A direct compaction with preferred Normal
zone will skip compaction of all zones including DMA32 because Normal
was still deferred. The allocation might have succeeded in DMA32, but
won't.
This patch makes compaction deferring work on individual zone basis
instead of preferred zone. For each zone, it checks compaction_deferred()
to decide if the zone should be skipped. If watermarks fail after
compacting the zone, defer_compaction() is called. The zone where
watermarks passed can still be deferred when the allocation attempt is
unsuccessful. When allocation is successful, compaction_defer_reset() is
called for the zone containing the allocated page. This approach should
approximate calling defer_compaction() only on zones where compaction was
attempted and did not yield allocated page. There might be corner cases
but that is inevitable as long as the decision to stop compacting dues not
guarantee that a page will be allocated.
Due to a new COMPACT_DEFERRED return value, some functions relying
implicitly on COMPACT_SKIPPED = 0 had to be updated, with comments made
more accurate. The did_some_progress output parameter of
__alloc_pages_direct_compact() is removed completely, as the caller
actually does not use it after compaction sets it - it is only considered
when direct reclaim sets it.
During testing on a two-node machine with a single very small Normal zone
on node 1, this patch has improved success rates in stress-highalloc
mmtests benchmark. The success here were previously made worse by commit
3a025760fc15 ("mm: page_alloc: spill to remote nodes before waking
kswapd") as kswapd was no longer resetting often enough the deferred
compaction for the Normal zone, and DMA32 zones on both nodes were thus
not considered for compaction. On different machine, success rates were
improved with __GFP_NO_KSWAPD allocations.
[akpm@linux-foundation.org: fix CONFIG_COMPACTION=n build]
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-09 22:27:02 +00:00
|
|
|
|
2014-10-09 22:27:04 +00:00
|
|
|
zone->compact_blockskip_flush = false;
|
|
|
|
compaction_defer_reset(zone, order, true);
|
|
|
|
count_vm_event(COMPACTSUCCESS);
|
|
|
|
return page;
|
|
|
|
}
|
2010-05-24 21:32:30 +00:00
|
|
|
|
2014-10-09 22:27:04 +00:00
|
|
|
/*
|
|
|
|
* It's bad if compaction run occurs and fails. The most likely reason
|
|
|
|
* is that pages exist, but not enough to satisfy watermarks.
|
|
|
|
*/
|
|
|
|
count_vm_event(COMPACTFAIL);
|
2012-01-13 01:19:41 +00:00
|
|
|
|
2014-10-09 22:27:04 +00:00
|
|
|
cond_resched();
|
2010-05-24 21:32:30 +00:00
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
2016-05-20 23:57:06 +00:00
|
|
|
|
2016-10-08 00:00:28 +00:00
|
|
|
static inline bool
|
|
|
|
should_compact_retry(struct alloc_context *ac, int order, int alloc_flags,
|
|
|
|
enum compact_result compact_result,
|
|
|
|
enum compact_priority *compact_priority,
|
2016-10-08 00:00:31 +00:00
|
|
|
int *compaction_retries)
|
2016-10-08 00:00:28 +00:00
|
|
|
{
|
|
|
|
int max_retries = MAX_COMPACT_RETRIES;
|
2016-10-08 00:00:34 +00:00
|
|
|
int min_priority;
|
2017-02-22 23:42:03 +00:00
|
|
|
bool ret = false;
|
|
|
|
int retries = *compaction_retries;
|
|
|
|
enum compact_priority priority = *compact_priority;
|
2016-10-08 00:00:28 +00:00
|
|
|
|
|
|
|
if (!order)
|
|
|
|
return false;
|
|
|
|
|
2016-10-08 00:00:31 +00:00
|
|
|
if (compaction_made_progress(compact_result))
|
|
|
|
(*compaction_retries)++;
|
|
|
|
|
2016-10-08 00:00:28 +00:00
|
|
|
/*
|
|
|
|
* compaction considers all the zone as desperately out of memory
|
|
|
|
* so it doesn't really make much sense to retry except when the
|
|
|
|
* failure could be caused by insufficient priority
|
|
|
|
*/
|
2016-10-08 00:00:31 +00:00
|
|
|
if (compaction_failed(compact_result))
|
|
|
|
goto check_priority;
|
2016-10-08 00:00:28 +00:00
|
|
|
|
mm, compaction: raise compaction priority after it withdrawns
Mike Kravetz reports that "hugetlb allocations could stall for minutes or
hours when should_compact_retry() would return true more often then it
should. Specifically, this was in the case where compact_result was
COMPACT_DEFERRED and COMPACT_PARTIAL_SKIPPED and no progress was being
made."
The problem is that the compaction_withdrawn() test in
should_compact_retry() includes compaction outcomes that are only possible
on low compaction priority, and results in a retry without increasing the
priority. This may result in furter reclaim, and more incomplete
compaction attempts.
With this patch, compaction priority is raised when possible, or
should_compact_retry() returns false.
The COMPACT_SKIPPED result doesn't really fit together with the other
outcomes in compaction_withdrawn(), as that's a result caused by
insufficient order-0 pages, not due to low compaction priority. With this
patch, it is moved to a new compaction_needs_reclaim() function, and for
that outcome we keep the current logic of retrying if it looks like
reclaim will be able to help.
Link: http://lkml.kernel.org/r/20190806014744.15446-4-mike.kravetz@oracle.com
Reported-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Tested-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-09-23 22:37:32 +00:00
|
|
|
/*
|
|
|
|
* compaction was skipped because there are not enough order-0 pages
|
|
|
|
* to work with, so we retry only if it looks like reclaim can help.
|
|
|
|
*/
|
|
|
|
if (compaction_needs_reclaim(compact_result)) {
|
|
|
|
ret = compaction_zonelist_suitable(ac, order, alloc_flags);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2016-10-08 00:00:28 +00:00
|
|
|
/*
|
|
|
|
* make sure the compaction wasn't deferred or didn't bail out early
|
|
|
|
* due to locks contention before we declare that we should give up.
|
mm, compaction: raise compaction priority after it withdrawns
Mike Kravetz reports that "hugetlb allocations could stall for minutes or
hours when should_compact_retry() would return true more often then it
should. Specifically, this was in the case where compact_result was
COMPACT_DEFERRED and COMPACT_PARTIAL_SKIPPED and no progress was being
made."
The problem is that the compaction_withdrawn() test in
should_compact_retry() includes compaction outcomes that are only possible
on low compaction priority, and results in a retry without increasing the
priority. This may result in furter reclaim, and more incomplete
compaction attempts.
With this patch, compaction priority is raised when possible, or
should_compact_retry() returns false.
The COMPACT_SKIPPED result doesn't really fit together with the other
outcomes in compaction_withdrawn(), as that's a result caused by
insufficient order-0 pages, not due to low compaction priority. With this
patch, it is moved to a new compaction_needs_reclaim() function, and for
that outcome we keep the current logic of retrying if it looks like
reclaim will be able to help.
Link: http://lkml.kernel.org/r/20190806014744.15446-4-mike.kravetz@oracle.com
Reported-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Tested-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-09-23 22:37:32 +00:00
|
|
|
* But the next retry should use a higher priority if allowed, so
|
|
|
|
* we don't just keep bailing out endlessly.
|
2016-10-08 00:00:28 +00:00
|
|
|
*/
|
2017-02-22 23:42:03 +00:00
|
|
|
if (compaction_withdrawn(compact_result)) {
|
mm, compaction: raise compaction priority after it withdrawns
Mike Kravetz reports that "hugetlb allocations could stall for minutes or
hours when should_compact_retry() would return true more often then it
should. Specifically, this was in the case where compact_result was
COMPACT_DEFERRED and COMPACT_PARTIAL_SKIPPED and no progress was being
made."
The problem is that the compaction_withdrawn() test in
should_compact_retry() includes compaction outcomes that are only possible
on low compaction priority, and results in a retry without increasing the
priority. This may result in furter reclaim, and more incomplete
compaction attempts.
With this patch, compaction priority is raised when possible, or
should_compact_retry() returns false.
The COMPACT_SKIPPED result doesn't really fit together with the other
outcomes in compaction_withdrawn(), as that's a result caused by
insufficient order-0 pages, not due to low compaction priority. With this
patch, it is moved to a new compaction_needs_reclaim() function, and for
that outcome we keep the current logic of retrying if it looks like
reclaim will be able to help.
Link: http://lkml.kernel.org/r/20190806014744.15446-4-mike.kravetz@oracle.com
Reported-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Tested-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-09-23 22:37:32 +00:00
|
|
|
goto check_priority;
|
2017-02-22 23:42:03 +00:00
|
|
|
}
|
2016-10-08 00:00:28 +00:00
|
|
|
|
|
|
|
/*
|
2017-07-12 21:36:45 +00:00
|
|
|
* !costly requests are much more important than __GFP_RETRY_MAYFAIL
|
2016-10-08 00:00:28 +00:00
|
|
|
* costly ones because they are de facto nofail and invoke OOM
|
|
|
|
* killer to move on while costly can fail and users are ready
|
|
|
|
* to cope with that. 1/4 retries is rather arbitrary but we
|
|
|
|
* would need much more detailed feedback from compaction to
|
|
|
|
* make a better decision.
|
|
|
|
*/
|
|
|
|
if (order > PAGE_ALLOC_COSTLY_ORDER)
|
|
|
|
max_retries /= 4;
|
2017-02-22 23:42:03 +00:00
|
|
|
if (*compaction_retries <= max_retries) {
|
|
|
|
ret = true;
|
|
|
|
goto out;
|
|
|
|
}
|
2016-10-08 00:00:28 +00:00
|
|
|
|
2016-10-08 00:00:31 +00:00
|
|
|
/*
|
|
|
|
* Make sure there are attempts at the highest priority if we exhausted
|
|
|
|
* all retries or failed at the lower priorities.
|
|
|
|
*/
|
|
|
|
check_priority:
|
2016-10-08 00:00:34 +00:00
|
|
|
min_priority = (order > PAGE_ALLOC_COSTLY_ORDER) ?
|
|
|
|
MIN_COMPACT_COSTLY_PRIORITY : MIN_COMPACT_PRIORITY;
|
2017-02-22 23:42:03 +00:00
|
|
|
|
2016-10-08 00:00:34 +00:00
|
|
|
if (*compact_priority > min_priority) {
|
2016-10-08 00:00:31 +00:00
|
|
|
(*compact_priority)--;
|
|
|
|
*compaction_retries = 0;
|
2017-02-22 23:42:03 +00:00
|
|
|
ret = true;
|
2016-10-08 00:00:31 +00:00
|
|
|
}
|
2017-02-22 23:42:03 +00:00
|
|
|
out:
|
|
|
|
trace_compact_retry(order, priority, compact_result, retries, max_retries, ret);
|
|
|
|
return ret;
|
2016-10-08 00:00:28 +00:00
|
|
|
}
|
2010-05-24 21:32:30 +00:00
|
|
|
#else
|
|
|
|
static inline struct page *
|
|
|
|
__alloc_pages_direct_compact(gfp_t gfp_mask, unsigned int order,
|
2016-05-20 00:13:38 +00:00
|
|
|
unsigned int alloc_flags, const struct alloc_context *ac,
|
2016-07-28 22:49:28 +00:00
|
|
|
enum compact_priority prio, enum compact_result *compact_result)
|
2010-05-24 21:32:30 +00:00
|
|
|
{
|
2016-05-20 23:57:06 +00:00
|
|
|
*compact_result = COMPACT_SKIPPED;
|
2010-05-24 21:32:30 +00:00
|
|
|
return NULL;
|
|
|
|
}
|
2016-05-20 23:57:06 +00:00
|
|
|
|
|
|
|
static inline bool
|
2016-05-20 23:57:12 +00:00
|
|
|
should_compact_retry(struct alloc_context *ac, unsigned int order, int alloc_flags,
|
|
|
|
enum compact_result compact_result,
|
2016-07-28 22:49:28 +00:00
|
|
|
enum compact_priority *compact_priority,
|
2016-10-08 00:00:31 +00:00
|
|
|
int *compaction_retries)
|
2016-05-20 23:57:06 +00:00
|
|
|
{
|
2016-05-20 23:57:15 +00:00
|
|
|
struct zone *zone;
|
|
|
|
struct zoneref *z;
|
|
|
|
|
|
|
|
if (!order || order > PAGE_ALLOC_COSTLY_ORDER)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* There are setups with compaction disabled which would prefer to loop
|
|
|
|
* inside the allocator rather than hit the oom killer prematurely.
|
|
|
|
* Let's give them a good hope and keep retrying while the order-0
|
|
|
|
* watermarks are OK.
|
|
|
|
*/
|
2020-06-03 22:59:01 +00:00
|
|
|
for_each_zone_zonelist_nodemask(zone, z, ac->zonelist,
|
|
|
|
ac->highest_zoneidx, ac->nodemask) {
|
2016-05-20 23:57:15 +00:00
|
|
|
if (zone_watermark_ok(zone, 0, min_wmark_pages(zone),
|
2020-06-03 22:59:01 +00:00
|
|
|
ac->highest_zoneidx, alloc_flags))
|
2016-05-20 23:57:15 +00:00
|
|
|
return true;
|
|
|
|
}
|
2016-05-20 23:57:06 +00:00
|
|
|
return false;
|
|
|
|
}
|
2016-10-08 00:00:28 +00:00
|
|
|
#endif /* CONFIG_COMPACTION */
|
2010-05-24 21:32:30 +00:00
|
|
|
|
2017-03-03 09:13:38 +00:00
|
|
|
#ifdef CONFIG_LOCKDEP
|
lockdep: fix fs_reclaim annotation
While revisiting my Btrfs swapfile series [1], I introduced a situation
in which reclaim would lock i_rwsem, and even though the swapon() path
clearly made GFP_KERNEL allocations while holding i_rwsem, I got no
complaints from lockdep. It turns out that the rework of the fs_reclaim
annotation was broken: if the current task has PF_MEMALLOC set, we don't
acquire the dummy fs_reclaim lock, but when reclaiming we always check
this _after_ we've just set the PF_MEMALLOC flag. In most cases, we can
fix this by moving the fs_reclaim_{acquire,release}() outside of the
memalloc_noreclaim_{save,restore}(), althought kswapd is slightly
different. After applying this, I got the expected lockdep splats.
1: https://lwn.net/Articles/625412/
Link: http://lkml.kernel.org/r/9f8aa70652a98e98d7c4de0fc96a4addcee13efe.1523778026.git.osandov@fb.com
Fixes: d92a8cfcb37e ("locking/lockdep: Rework FS_RECLAIM annotation")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-06-08 00:07:02 +00:00
|
|
|
static struct lockdep_map __fs_reclaim_map =
|
2017-03-03 09:13:38 +00:00
|
|
|
STATIC_LOCKDEP_MAP_INIT("fs_reclaim", &__fs_reclaim_map);
|
|
|
|
|
2020-12-15 03:08:30 +00:00
|
|
|
static bool __need_reclaim(gfp_t gfp_mask)
|
2017-03-03 09:13:38 +00:00
|
|
|
{
|
|
|
|
/* no reclaim without waiting on it */
|
|
|
|
if (!(gfp_mask & __GFP_DIRECT_RECLAIM))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
/* this guy won't enter reclaim */
|
2018-03-22 23:17:10 +00:00
|
|
|
if (current->flags & PF_MEMALLOC)
|
2017-03-03 09:13:38 +00:00
|
|
|
return false;
|
|
|
|
|
|
|
|
if (gfp_mask & __GFP_NOLOCKDEP)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
lockdep: fix fs_reclaim annotation
While revisiting my Btrfs swapfile series [1], I introduced a situation
in which reclaim would lock i_rwsem, and even though the swapon() path
clearly made GFP_KERNEL allocations while holding i_rwsem, I got no
complaints from lockdep. It turns out that the rework of the fs_reclaim
annotation was broken: if the current task has PF_MEMALLOC set, we don't
acquire the dummy fs_reclaim lock, but when reclaiming we always check
this _after_ we've just set the PF_MEMALLOC flag. In most cases, we can
fix this by moving the fs_reclaim_{acquire,release}() outside of the
memalloc_noreclaim_{save,restore}(), althought kswapd is slightly
different. After applying this, I got the expected lockdep splats.
1: https://lwn.net/Articles/625412/
Link: http://lkml.kernel.org/r/9f8aa70652a98e98d7c4de0fc96a4addcee13efe.1523778026.git.osandov@fb.com
Fixes: d92a8cfcb37e ("locking/lockdep: Rework FS_RECLAIM annotation")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-06-08 00:07:02 +00:00
|
|
|
void __fs_reclaim_acquire(void)
|
|
|
|
{
|
|
|
|
lock_map_acquire(&__fs_reclaim_map);
|
|
|
|
}
|
|
|
|
|
|
|
|
void __fs_reclaim_release(void)
|
|
|
|
{
|
|
|
|
lock_map_release(&__fs_reclaim_map);
|
|
|
|
}
|
|
|
|
|
2017-03-03 09:13:38 +00:00
|
|
|
void fs_reclaim_acquire(gfp_t gfp_mask)
|
|
|
|
{
|
2020-12-15 03:08:30 +00:00
|
|
|
gfp_mask = current_gfp_context(gfp_mask);
|
|
|
|
|
|
|
|
if (__need_reclaim(gfp_mask)) {
|
|
|
|
if (gfp_mask & __GFP_FS)
|
|
|
|
__fs_reclaim_acquire();
|
|
|
|
|
|
|
|
#ifdef CONFIG_MMU_NOTIFIER
|
|
|
|
lock_map_acquire(&__mmu_notifier_invalidate_range_start_map);
|
|
|
|
lock_map_release(&__mmu_notifier_invalidate_range_start_map);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
}
|
2017-03-03 09:13:38 +00:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(fs_reclaim_acquire);
|
|
|
|
|
|
|
|
void fs_reclaim_release(gfp_t gfp_mask)
|
|
|
|
{
|
2020-12-15 03:08:30 +00:00
|
|
|
gfp_mask = current_gfp_context(gfp_mask);
|
|
|
|
|
|
|
|
if (__need_reclaim(gfp_mask)) {
|
|
|
|
if (gfp_mask & __GFP_FS)
|
|
|
|
__fs_reclaim_release();
|
|
|
|
}
|
2017-03-03 09:13:38 +00:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(fs_reclaim_release);
|
|
|
|
#endif
|
|
|
|
|
2012-01-25 11:09:52 +00:00
|
|
|
/* Perform direct synchronous page reclaim */
|
2020-10-13 23:55:54 +00:00
|
|
|
static unsigned long
|
2015-02-11 23:25:41 +00:00
|
|
|
__perform_reclaim(gfp_t gfp_mask, unsigned int order,
|
|
|
|
const struct alloc_context *ac)
|
2009-06-16 22:31:57 +00:00
|
|
|
{
|
2017-05-08 22:59:50 +00:00
|
|
|
unsigned int noreclaim_flag;
|
2020-10-13 23:55:54 +00:00
|
|
|
unsigned long pflags, progress;
|
2009-06-16 22:31:57 +00:00
|
|
|
|
|
|
|
cond_resched();
|
|
|
|
|
|
|
|
/* We now go into synchronous reclaim */
|
|
|
|
cpuset_memory_pressure_bump();
|
psi: pressure stall information for CPU, memory, and IO
When systems are overcommitted and resources become contended, it's hard
to tell exactly the impact this has on workload productivity, or how close
the system is to lockups and OOM kills. In particular, when machines work
multiple jobs concurrently, the impact of overcommit in terms of latency
and throughput on the individual job can be enormous.
In order to maximize hardware utilization without sacrificing individual
job health or risk complete machine lockups, this patch implements a way
to quantify resource pressure in the system.
A kernel built with CONFIG_PSI=y creates files in /proc/pressure/ that
expose the percentage of time the system is stalled on CPU, memory, or IO,
respectively. Stall states are aggregate versions of the per-task delay
accounting delays:
cpu: some tasks are runnable but not executing on a CPU
memory: tasks are reclaiming, or waiting for swapin or thrashing cache
io: tasks are waiting for io completions
These percentages of walltime can be thought of as pressure percentages,
and they give a general sense of system health and productivity loss
incurred by resource overcommit. They can also indicate when the system
is approaching lockup scenarios and OOMs.
To do this, psi keeps track of the task states associated with each CPU
and samples the time they spend in stall states. Every 2 seconds, the
samples are averaged across CPUs - weighted by the CPUs' non-idle time to
eliminate artifacts from unused CPUs - and translated into percentages of
walltime. A running average of those percentages is maintained over 10s,
1m, and 5m periods (similar to the loadaverage).
[hannes@cmpxchg.org: doc fixlet, per Randy]
Link: http://lkml.kernel.org/r/20180828205625.GA14030@cmpxchg.org
[hannes@cmpxchg.org: code optimization]
Link: http://lkml.kernel.org/r/20180907175015.GA8479@cmpxchg.org
[hannes@cmpxchg.org: rename psi_clock() to psi_update_work(), per Peter]
Link: http://lkml.kernel.org/r/20180907145404.GB11088@cmpxchg.org
[hannes@cmpxchg.org: fix build]
Link: http://lkml.kernel.org/r/20180913014222.GA2370@cmpxchg.org
Link: http://lkml.kernel.org/r/20180828172258.3185-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Daniel Drake <drake@endlessm.com>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:06:27 +00:00
|
|
|
psi_memstall_enter(&pflags);
|
2017-03-03 09:13:38 +00:00
|
|
|
fs_reclaim_acquire(gfp_mask);
|
lockdep: fix fs_reclaim annotation
While revisiting my Btrfs swapfile series [1], I introduced a situation
in which reclaim would lock i_rwsem, and even though the swapon() path
clearly made GFP_KERNEL allocations while holding i_rwsem, I got no
complaints from lockdep. It turns out that the rework of the fs_reclaim
annotation was broken: if the current task has PF_MEMALLOC set, we don't
acquire the dummy fs_reclaim lock, but when reclaiming we always check
this _after_ we've just set the PF_MEMALLOC flag. In most cases, we can
fix this by moving the fs_reclaim_{acquire,release}() outside of the
memalloc_noreclaim_{save,restore}(), althought kswapd is slightly
different. After applying this, I got the expected lockdep splats.
1: https://lwn.net/Articles/625412/
Link: http://lkml.kernel.org/r/9f8aa70652a98e98d7c4de0fc96a4addcee13efe.1523778026.git.osandov@fb.com
Fixes: d92a8cfcb37e ("locking/lockdep: Rework FS_RECLAIM annotation")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-06-08 00:07:02 +00:00
|
|
|
noreclaim_flag = memalloc_noreclaim_save();
|
2009-06-16 22:31:57 +00:00
|
|
|
|
2015-02-11 23:25:41 +00:00
|
|
|
progress = try_to_free_pages(ac->zonelist, order, gfp_mask,
|
|
|
|
ac->nodemask);
|
2009-06-16 22:31:57 +00:00
|
|
|
|
2017-05-08 22:59:50 +00:00
|
|
|
memalloc_noreclaim_restore(noreclaim_flag);
|
lockdep: fix fs_reclaim annotation
While revisiting my Btrfs swapfile series [1], I introduced a situation
in which reclaim would lock i_rwsem, and even though the swapon() path
clearly made GFP_KERNEL allocations while holding i_rwsem, I got no
complaints from lockdep. It turns out that the rework of the fs_reclaim
annotation was broken: if the current task has PF_MEMALLOC set, we don't
acquire the dummy fs_reclaim lock, but when reclaiming we always check
this _after_ we've just set the PF_MEMALLOC flag. In most cases, we can
fix this by moving the fs_reclaim_{acquire,release}() outside of the
memalloc_noreclaim_{save,restore}(), althought kswapd is slightly
different. After applying this, I got the expected lockdep splats.
1: https://lwn.net/Articles/625412/
Link: http://lkml.kernel.org/r/9f8aa70652a98e98d7c4de0fc96a4addcee13efe.1523778026.git.osandov@fb.com
Fixes: d92a8cfcb37e ("locking/lockdep: Rework FS_RECLAIM annotation")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-06-08 00:07:02 +00:00
|
|
|
fs_reclaim_release(gfp_mask);
|
psi: pressure stall information for CPU, memory, and IO
When systems are overcommitted and resources become contended, it's hard
to tell exactly the impact this has on workload productivity, or how close
the system is to lockups and OOM kills. In particular, when machines work
multiple jobs concurrently, the impact of overcommit in terms of latency
and throughput on the individual job can be enormous.
In order to maximize hardware utilization without sacrificing individual
job health or risk complete machine lockups, this patch implements a way
to quantify resource pressure in the system.
A kernel built with CONFIG_PSI=y creates files in /proc/pressure/ that
expose the percentage of time the system is stalled on CPU, memory, or IO,
respectively. Stall states are aggregate versions of the per-task delay
accounting delays:
cpu: some tasks are runnable but not executing on a CPU
memory: tasks are reclaiming, or waiting for swapin or thrashing cache
io: tasks are waiting for io completions
These percentages of walltime can be thought of as pressure percentages,
and they give a general sense of system health and productivity loss
incurred by resource overcommit. They can also indicate when the system
is approaching lockup scenarios and OOMs.
To do this, psi keeps track of the task states associated with each CPU
and samples the time they spend in stall states. Every 2 seconds, the
samples are averaged across CPUs - weighted by the CPUs' non-idle time to
eliminate artifacts from unused CPUs - and translated into percentages of
walltime. A running average of those percentages is maintained over 10s,
1m, and 5m periods (similar to the loadaverage).
[hannes@cmpxchg.org: doc fixlet, per Randy]
Link: http://lkml.kernel.org/r/20180828205625.GA14030@cmpxchg.org
[hannes@cmpxchg.org: code optimization]
Link: http://lkml.kernel.org/r/20180907175015.GA8479@cmpxchg.org
[hannes@cmpxchg.org: rename psi_clock() to psi_update_work(), per Peter]
Link: http://lkml.kernel.org/r/20180907145404.GB11088@cmpxchg.org
[hannes@cmpxchg.org: fix build]
Link: http://lkml.kernel.org/r/20180913014222.GA2370@cmpxchg.org
Link: http://lkml.kernel.org/r/20180828172258.3185-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Daniel Drake <drake@endlessm.com>
Tested-by: Suren Baghdasaryan <surenb@google.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <jweiner@fb.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Enderborg <peter.enderborg@sony.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:06:27 +00:00
|
|
|
psi_memstall_leave(&pflags);
|
2009-06-16 22:31:57 +00:00
|
|
|
|
|
|
|
cond_resched();
|
|
|
|
|
2012-01-25 11:09:52 +00:00
|
|
|
return progress;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* The really slow allocator path where we enter direct reclaim */
|
|
|
|
static inline struct page *
|
|
|
|
__alloc_pages_direct_reclaim(gfp_t gfp_mask, unsigned int order,
|
2016-05-20 00:13:38 +00:00
|
|
|
unsigned int alloc_flags, const struct alloc_context *ac,
|
2015-02-11 23:25:41 +00:00
|
|
|
unsigned long *did_some_progress)
|
2012-01-25 11:09:52 +00:00
|
|
|
{
|
|
|
|
struct page *page = NULL;
|
|
|
|
bool drained = false;
|
|
|
|
|
2015-02-11 23:25:41 +00:00
|
|
|
*did_some_progress = __perform_reclaim(gfp_mask, order, ac);
|
2010-09-09 23:38:18 +00:00
|
|
|
if (unlikely(!(*did_some_progress)))
|
|
|
|
return NULL;
|
2009-06-16 22:31:57 +00:00
|
|
|
|
2010-09-09 23:38:18 +00:00
|
|
|
retry:
|
2016-07-28 22:49:13 +00:00
|
|
|
page = get_page_from_freelist(gfp_mask, order, alloc_flags, ac);
|
2010-09-09 23:38:18 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If an allocation failed after direct reclaim, it could be because
|
2015-11-07 00:28:37 +00:00
|
|
|
* pages are pinned on the per-cpu lists or in high alloc reserves.
|
2020-08-12 01:33:14 +00:00
|
|
|
* Shrink them and try again
|
2010-09-09 23:38:18 +00:00
|
|
|
*/
|
|
|
|
if (!page && !drained) {
|
2016-12-13 00:42:14 +00:00
|
|
|
unreserve_highatomic_pageblock(ac, false);
|
2014-12-10 23:43:01 +00:00
|
|
|
drain_all_pages(NULL);
|
2010-09-09 23:38:18 +00:00
|
|
|
drained = true;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
|
2009-06-16 22:31:57 +00:00
|
|
|
return page;
|
|
|
|
}
|
|
|
|
|
2018-04-05 23:25:16 +00:00
|
|
|
static void wake_all_kswapds(unsigned int order, gfp_t gfp_mask,
|
|
|
|
const struct alloc_context *ac)
|
mm: page_alloc: spill to remote nodes before waking kswapd
On NUMA systems, a node may start thrashing cache or even swap anonymous
pages while there are still free pages on remote nodes.
This is a result of commits 81c0a2bb515f ("mm: page_alloc: fair zone
allocator policy") and fff4068cba48 ("mm: page_alloc: revert NUMA aspect
of fair allocation policy").
Before those changes, the allocator would first try all allowed zones,
including those on remote nodes, before waking any kswapds. But now,
the allocator fastpath doubles as the fairness pass, which in turn can
only consider the local node to prevent remote spilling based on
exhausted fairness batches alone. Remote nodes are only considered in
the slowpath, after the kswapds are woken up. But if remote nodes still
have free memory, kswapd should not be woken to rebalance the local node
or it may thrash cash or swap prematurely.
Fix this by adding one more unfair pass over the zonelist that is
allowed to spill to remote nodes after the local fairness pass fails but
before entering the slowpath and waking the kswapds.
This also gets rid of the GFP_THISNODE exemption from the fairness
protocol because the unfair pass is no longer tied to kswapd, which
GFP_THISNODE is not allowed to wake up.
However, because remote spills can be more frequent now - we prefer them
over local kswapd reclaim - the allocation batches on remote nodes could
underflow more heavily. When resetting the batches, use
atomic_long_read() directly instead of zone_page_state() to calculate the
delta as the latter filters negative counter values.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: <stable@kernel.org> [3.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:48 +00:00
|
|
|
{
|
|
|
|
struct zoneref *z;
|
|
|
|
struct zone *zone;
|
2016-07-28 22:46:26 +00:00
|
|
|
pg_data_t *last_pgdat = NULL;
|
2020-06-03 22:59:01 +00:00
|
|
|
enum zone_type highest_zoneidx = ac->highest_zoneidx;
|
mm: page_alloc: spill to remote nodes before waking kswapd
On NUMA systems, a node may start thrashing cache or even swap anonymous
pages while there are still free pages on remote nodes.
This is a result of commits 81c0a2bb515f ("mm: page_alloc: fair zone
allocator policy") and fff4068cba48 ("mm: page_alloc: revert NUMA aspect
of fair allocation policy").
Before those changes, the allocator would first try all allowed zones,
including those on remote nodes, before waking any kswapds. But now,
the allocator fastpath doubles as the fairness pass, which in turn can
only consider the local node to prevent remote spilling based on
exhausted fairness batches alone. Remote nodes are only considered in
the slowpath, after the kswapds are woken up. But if remote nodes still
have free memory, kswapd should not be woken to rebalance the local node
or it may thrash cash or swap prematurely.
Fix this by adding one more unfair pass over the zonelist that is
allowed to spill to remote nodes after the local fairness pass fails but
before entering the slowpath and waking the kswapds.
This also gets rid of the GFP_THISNODE exemption from the fairness
protocol because the unfair pass is no longer tied to kswapd, which
GFP_THISNODE is not allowed to wake up.
However, because remote spills can be more frequent now - we prefer them
over local kswapd reclaim - the allocation batches on remote nodes could
underflow more heavily. When resetting the batches, use
atomic_long_read() directly instead of zone_page_state() to calculate the
delta as the latter filters negative counter values.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: <stable@kernel.org> [3.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:48 +00:00
|
|
|
|
2020-06-03 22:59:01 +00:00
|
|
|
for_each_zone_zonelist_nodemask(zone, z, ac->zonelist, highest_zoneidx,
|
2018-04-05 23:25:16 +00:00
|
|
|
ac->nodemask) {
|
2016-07-28 22:46:26 +00:00
|
|
|
if (last_pgdat != zone->zone_pgdat)
|
2020-06-03 22:59:01 +00:00
|
|
|
wakeup_kswapd(zone, gfp_mask, order, highest_zoneidx);
|
2016-07-28 22:46:26 +00:00
|
|
|
last_pgdat = zone->zone_pgdat;
|
|
|
|
}
|
mm: page_alloc: spill to remote nodes before waking kswapd
On NUMA systems, a node may start thrashing cache or even swap anonymous
pages while there are still free pages on remote nodes.
This is a result of commits 81c0a2bb515f ("mm: page_alloc: fair zone
allocator policy") and fff4068cba48 ("mm: page_alloc: revert NUMA aspect
of fair allocation policy").
Before those changes, the allocator would first try all allowed zones,
including those on remote nodes, before waking any kswapds. But now,
the allocator fastpath doubles as the fairness pass, which in turn can
only consider the local node to prevent remote spilling based on
exhausted fairness batches alone. Remote nodes are only considered in
the slowpath, after the kswapds are woken up. But if remote nodes still
have free memory, kswapd should not be woken to rebalance the local node
or it may thrash cash or swap prematurely.
Fix this by adding one more unfair pass over the zonelist that is
allowed to spill to remote nodes after the local fairness pass fails but
before entering the slowpath and waking the kswapds.
This also gets rid of the GFP_THISNODE exemption from the fairness
protocol because the unfair pass is no longer tied to kswapd, which
GFP_THISNODE is not allowed to wake up.
However, because remote spills can be more frequent now - we prefer them
over local kswapd reclaim - the allocation batches on remote nodes could
underflow more heavily. When resetting the batches, use
atomic_long_read() directly instead of zone_page_state() to calculate the
delta as the latter filters negative counter values.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: <stable@kernel.org> [3.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:48 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:13:38 +00:00
|
|
|
static inline unsigned int
|
2009-06-16 22:32:02 +00:00
|
|
|
gfp_to_alloc_flags(gfp_t gfp_mask)
|
|
|
|
{
|
2016-05-20 00:13:38 +00:00
|
|
|
unsigned int alloc_flags = ALLOC_WMARK_MIN | ALLOC_CPUSET;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2020-04-02 04:09:47 +00:00
|
|
|
/*
|
|
|
|
* __GFP_HIGH is assumed to be the same as ALLOC_HIGH
|
|
|
|
* and __GFP_KSWAPD_RECLAIM is assumed to be the same as ALLOC_KSWAPD
|
|
|
|
* to save two branches.
|
|
|
|
*/
|
2010-10-26 21:21:59 +00:00
|
|
|
BUILD_BUG_ON(__GFP_HIGH != (__force gfp_t) ALLOC_HIGH);
|
2020-04-02 04:09:47 +00:00
|
|
|
BUILD_BUG_ON(__GFP_KSWAPD_RECLAIM != (__force gfp_t) ALLOC_KSWAPD);
|
2006-12-08 10:39:45 +00:00
|
|
|
|
2009-06-16 22:32:02 +00:00
|
|
|
/*
|
|
|
|
* The caller may dip into page reserves a bit more if the caller
|
|
|
|
* cannot run direct reclaim, or if the caller has realtime scheduling
|
|
|
|
* policy or is asking for __GFP_HIGH memory. GFP_ATOMIC requests will
|
2015-11-07 00:28:21 +00:00
|
|
|
* set both ALLOC_HARDER (__GFP_ATOMIC) and ALLOC_HIGH (__GFP_HIGH).
|
2009-06-16 22:32:02 +00:00
|
|
|
*/
|
2020-04-02 04:09:47 +00:00
|
|
|
alloc_flags |= (__force int)
|
|
|
|
(gfp_mask & (__GFP_HIGH | __GFP_KSWAPD_RECLAIM));
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2015-11-07 00:28:21 +00:00
|
|
|
if (gfp_mask & __GFP_ATOMIC) {
|
2011-01-13 23:46:49 +00:00
|
|
|
/*
|
2014-07-30 23:08:24 +00:00
|
|
|
* Not worth trying to allocate harder for __GFP_NOMEMALLOC even
|
|
|
|
* if it can't schedule.
|
2011-01-13 23:46:49 +00:00
|
|
|
*/
|
2014-07-30 23:08:24 +00:00
|
|
|
if (!(gfp_mask & __GFP_NOMEMALLOC))
|
2011-01-13 23:46:49 +00:00
|
|
|
alloc_flags |= ALLOC_HARDER;
|
2007-10-16 08:25:37 +00:00
|
|
|
/*
|
2014-07-30 23:08:24 +00:00
|
|
|
* Ignore cpuset mems for GFP_ATOMIC rather than fail, see the
|
2014-10-20 11:50:30 +00:00
|
|
|
* comment for __cpuset_node_allowed().
|
2007-10-16 08:25:37 +00:00
|
|
|
*/
|
2009-06-16 22:32:02 +00:00
|
|
|
alloc_flags &= ~ALLOC_CPUSET;
|
2011-01-13 23:47:32 +00:00
|
|
|
} else if (unlikely(rt_task(current)) && !in_interrupt())
|
2009-06-16 22:32:02 +00:00
|
|
|
alloc_flags |= ALLOC_HARDER;
|
|
|
|
|
2020-08-07 06:26:04 +00:00
|
|
|
alloc_flags = current_alloc_flags(gfp_mask, alloc_flags);
|
|
|
|
|
2009-06-16 22:32:02 +00:00
|
|
|
return alloc_flags;
|
|
|
|
}
|
|
|
|
|
2017-09-06 23:24:50 +00:00
|
|
|
static bool oom_reserves_allowed(struct task_struct *tsk)
|
mm: sl[au]b: add knowledge of PFMEMALLOC reserve pages
When a user or administrator requires swap for their application, they
create a swap partition and file, format it with mkswap and activate it
with swapon. Swap over the network is considered as an option in diskless
systems. The two likely scenarios are when blade servers are used as part
of a cluster where the form factor or maintenance costs do not allow the
use of disks and thin clients.
The Linux Terminal Server Project recommends the use of the Network Block
Device (NBD) for swap according to the manual at
https://sourceforge.net/projects/ltsp/files/Docs-Admin-Guide/LTSPManual.pdf/download
There is also documentation and tutorials on how to setup swap over NBD at
places like https://help.ubuntu.com/community/UbuntuLTSP/EnableNBDSWAP The
nbd-client also documents the use of NBD as swap. Despite this, the fact
is that a machine using NBD for swap can deadlock within minutes if swap
is used intensively. This patch series addresses the problem.
The core issue is that network block devices do not use mempools like
normal block devices do. As the host cannot control where they receive
packets from, they cannot reliably work out in advance how much memory
they might need. Some years ago, Peter Zijlstra developed a series of
patches that supported swap over an NFS that at least one distribution is
carrying within their kernels. This patch series borrows very heavily
from Peter's work to support swapping over NBD as a pre-requisite to
supporting swap-over-NFS. The bulk of the complexity is concerned with
preserving memory that is allocated from the PFMEMALLOC reserves for use
by the network layer which is needed for both NBD and NFS.
Patch 1 adds knowledge of the PFMEMALLOC reserves to SLAB and SLUB to
preserve access to pages allocated under low memory situations
to callers that are freeing memory.
Patch 2 optimises the SLUB fast path to avoid pfmemalloc checks
Patch 3 introduces __GFP_MEMALLOC to allow access to the PFMEMALLOC
reserves without setting PFMEMALLOC.
Patch 4 opens the possibility for softirqs to use PFMEMALLOC reserves
for later use by network packet processing.
Patch 5 only sets page->pfmemalloc when ALLOC_NO_WATERMARKS was required
Patch 6 ignores memory policies when ALLOC_NO_WATERMARKS is set.
Patches 7-12 allows network processing to use PFMEMALLOC reserves when
the socket has been marked as being used by the VM to clean pages. If
packets are received and stored in pages that were allocated under
low-memory situations and are unrelated to the VM, the packets
are dropped.
Patch 11 reintroduces __skb_alloc_page which the networking
folk may object to but is needed in some cases to propogate
pfmemalloc from a newly allocated page to an skb. If there is a
strong objection, this patch can be dropped with the impact being
that swap-over-network will be slower in some cases but it should
not fail.
Patch 13 is a micro-optimisation to avoid a function call in the
common case.
Patch 14 tags NBD sockets as being SOCK_MEMALLOC so they can use
PFMEMALLOC if necessary.
Patch 15 notes that it is still possible for the PFMEMALLOC reserve
to be depleted. To prevent this, direct reclaimers get throttled on
a waitqueue if 50% of the PFMEMALLOC reserves are depleted. It is
expected that kswapd and the direct reclaimers already running
will clean enough pages for the low watermark to be reached and
the throttled processes are woken up.
Patch 16 adds a statistic to track how often processes get throttled
Some basic performance testing was run using kernel builds, netperf on
loopback for UDP and TCP, hackbench (pipes and sockets), iozone and
sysbench. Each of them were expected to use the sl*b allocators
reasonably heavily but there did not appear to be significant performance
variances.
For testing swap-over-NBD, a machine was booted with 2G of RAM with a
swapfile backed by NBD. 8*NUM_CPU processes were started that create
anonymous memory mappings and read them linearly in a loop. The total
size of the mappings were 4*PHYSICAL_MEMORY to use swap heavily under
memory pressure.
Without the patches and using SLUB, the machine locks up within minutes
and runs to completion with them applied. With SLAB, the story is
different as an unpatched kernel run to completion. However, the patched
kernel completed the test 45% faster.
MICRO
3.5.0-rc2 3.5.0-rc2
vanilla swapnbd
Unrecognised test vmscan-anon-mmap-write
MMTests Statistics: duration
Sys Time Running Test (seconds) 197.80 173.07
User+Sys Time Running Test (seconds) 206.96 182.03
Total Elapsed Time (seconds) 3240.70 1762.09
This patch: mm: sl[au]b: add knowledge of PFMEMALLOC reserve pages
Allocations of pages below the min watermark run a risk of the machine
hanging due to a lack of memory. To prevent this, only callers who have
PF_MEMALLOC or TIF_MEMDIE set and are not processing an interrupt are
allowed to allocate with ALLOC_NO_WATERMARKS. Once they are allocated to
a slab though, nothing prevents other callers consuming free objects
within those slabs. This patch limits access to slab pages that were
alloced from the PFMEMALLOC reserves.
When this patch is applied, pages allocated from below the low watermark
are returned with page->pfmemalloc set and it is up to the caller to
determine how the page should be protected. SLAB restricts access to any
page with page->pfmemalloc set to callers which are known to able to
access the PFMEMALLOC reserve. If one is not available, an attempt is
made to allocate a new page rather than use a reserve. SLUB is a bit more
relaxed in that it only records if the current per-CPU page was allocated
from PFMEMALLOC reserve and uses another partial slab if the caller does
not have the necessary GFP or process flags. This was found to be
sufficient in tests to avoid hangs due to SLUB generally maintaining
smaller lists than SLAB.
In low-memory conditions it does mean that !PFMEMALLOC allocators can fail
a slab allocation even though free objects are available because they are
being preserved for callers that are freeing pages.
[a.p.zijlstra@chello.nl: Original implementation]
[sebastian@breakpoint.cc: Correct order of page flag clearing]
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: David Miller <davem@davemloft.net>
Cc: Neil Brown <neilb@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-07-31 23:43:58 +00:00
|
|
|
{
|
2017-09-06 23:24:50 +00:00
|
|
|
if (!tsk_is_oom_victim(tsk))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* !MMU doesn't have oom reaper so give access to memory reserves
|
|
|
|
* only to the thread with TIF_MEMDIE set
|
|
|
|
*/
|
|
|
|
if (!IS_ENABLED(CONFIG_MMU) && !test_thread_flag(TIF_MEMDIE))
|
2016-07-28 22:49:13 +00:00
|
|
|
return false;
|
|
|
|
|
2017-09-06 23:24:50 +00:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Distinguish requests which really need access to full memory
|
|
|
|
* reserves from oom victims which can live with a portion of it
|
|
|
|
*/
|
|
|
|
static inline int __gfp_pfmemalloc_flags(gfp_t gfp_mask)
|
|
|
|
{
|
|
|
|
if (unlikely(gfp_mask & __GFP_NOMEMALLOC))
|
|
|
|
return 0;
|
2016-07-28 22:49:13 +00:00
|
|
|
if (gfp_mask & __GFP_MEMALLOC)
|
2017-09-06 23:24:50 +00:00
|
|
|
return ALLOC_NO_WATERMARKS;
|
2016-07-28 22:49:13 +00:00
|
|
|
if (in_serving_softirq() && (current->flags & PF_MEMALLOC))
|
2017-09-06 23:24:50 +00:00
|
|
|
return ALLOC_NO_WATERMARKS;
|
|
|
|
if (!in_interrupt()) {
|
|
|
|
if (current->flags & PF_MEMALLOC)
|
|
|
|
return ALLOC_NO_WATERMARKS;
|
|
|
|
else if (oom_reserves_allowed(current))
|
|
|
|
return ALLOC_OOM;
|
|
|
|
}
|
2016-07-28 22:49:13 +00:00
|
|
|
|
2017-09-06 23:24:50 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool gfp_pfmemalloc_allowed(gfp_t gfp_mask)
|
|
|
|
{
|
|
|
|
return !!__gfp_pfmemalloc_flags(gfp_mask);
|
mm: sl[au]b: add knowledge of PFMEMALLOC reserve pages
When a user or administrator requires swap for their application, they
create a swap partition and file, format it with mkswap and activate it
with swapon. Swap over the network is considered as an option in diskless
systems. The two likely scenarios are when blade servers are used as part
of a cluster where the form factor or maintenance costs do not allow the
use of disks and thin clients.
The Linux Terminal Server Project recommends the use of the Network Block
Device (NBD) for swap according to the manual at
https://sourceforge.net/projects/ltsp/files/Docs-Admin-Guide/LTSPManual.pdf/download
There is also documentation and tutorials on how to setup swap over NBD at
places like https://help.ubuntu.com/community/UbuntuLTSP/EnableNBDSWAP The
nbd-client also documents the use of NBD as swap. Despite this, the fact
is that a machine using NBD for swap can deadlock within minutes if swap
is used intensively. This patch series addresses the problem.
The core issue is that network block devices do not use mempools like
normal block devices do. As the host cannot control where they receive
packets from, they cannot reliably work out in advance how much memory
they might need. Some years ago, Peter Zijlstra developed a series of
patches that supported swap over an NFS that at least one distribution is
carrying within their kernels. This patch series borrows very heavily
from Peter's work to support swapping over NBD as a pre-requisite to
supporting swap-over-NFS. The bulk of the complexity is concerned with
preserving memory that is allocated from the PFMEMALLOC reserves for use
by the network layer which is needed for both NBD and NFS.
Patch 1 adds knowledge of the PFMEMALLOC reserves to SLAB and SLUB to
preserve access to pages allocated under low memory situations
to callers that are freeing memory.
Patch 2 optimises the SLUB fast path to avoid pfmemalloc checks
Patch 3 introduces __GFP_MEMALLOC to allow access to the PFMEMALLOC
reserves without setting PFMEMALLOC.
Patch 4 opens the possibility for softirqs to use PFMEMALLOC reserves
for later use by network packet processing.
Patch 5 only sets page->pfmemalloc when ALLOC_NO_WATERMARKS was required
Patch 6 ignores memory policies when ALLOC_NO_WATERMARKS is set.
Patches 7-12 allows network processing to use PFMEMALLOC reserves when
the socket has been marked as being used by the VM to clean pages. If
packets are received and stored in pages that were allocated under
low-memory situations and are unrelated to the VM, the packets
are dropped.
Patch 11 reintroduces __skb_alloc_page which the networking
folk may object to but is needed in some cases to propogate
pfmemalloc from a newly allocated page to an skb. If there is a
strong objection, this patch can be dropped with the impact being
that swap-over-network will be slower in some cases but it should
not fail.
Patch 13 is a micro-optimisation to avoid a function call in the
common case.
Patch 14 tags NBD sockets as being SOCK_MEMALLOC so they can use
PFMEMALLOC if necessary.
Patch 15 notes that it is still possible for the PFMEMALLOC reserve
to be depleted. To prevent this, direct reclaimers get throttled on
a waitqueue if 50% of the PFMEMALLOC reserves are depleted. It is
expected that kswapd and the direct reclaimers already running
will clean enough pages for the low watermark to be reached and
the throttled processes are woken up.
Patch 16 adds a statistic to track how often processes get throttled
Some basic performance testing was run using kernel builds, netperf on
loopback for UDP and TCP, hackbench (pipes and sockets), iozone and
sysbench. Each of them were expected to use the sl*b allocators
reasonably heavily but there did not appear to be significant performance
variances.
For testing swap-over-NBD, a machine was booted with 2G of RAM with a
swapfile backed by NBD. 8*NUM_CPU processes were started that create
anonymous memory mappings and read them linearly in a loop. The total
size of the mappings were 4*PHYSICAL_MEMORY to use swap heavily under
memory pressure.
Without the patches and using SLUB, the machine locks up within minutes
and runs to completion with them applied. With SLAB, the story is
different as an unpatched kernel run to completion. However, the patched
kernel completed the test 45% faster.
MICRO
3.5.0-rc2 3.5.0-rc2
vanilla swapnbd
Unrecognised test vmscan-anon-mmap-write
MMTests Statistics: duration
Sys Time Running Test (seconds) 197.80 173.07
User+Sys Time Running Test (seconds) 206.96 182.03
Total Elapsed Time (seconds) 3240.70 1762.09
This patch: mm: sl[au]b: add knowledge of PFMEMALLOC reserve pages
Allocations of pages below the min watermark run a risk of the machine
hanging due to a lack of memory. To prevent this, only callers who have
PF_MEMALLOC or TIF_MEMDIE set and are not processing an interrupt are
allowed to allocate with ALLOC_NO_WATERMARKS. Once they are allocated to
a slab though, nothing prevents other callers consuming free objects
within those slabs. This patch limits access to slab pages that were
alloced from the PFMEMALLOC reserves.
When this patch is applied, pages allocated from below the low watermark
are returned with page->pfmemalloc set and it is up to the caller to
determine how the page should be protected. SLAB restricts access to any
page with page->pfmemalloc set to callers which are known to able to
access the PFMEMALLOC reserve. If one is not available, an attempt is
made to allocate a new page rather than use a reserve. SLUB is a bit more
relaxed in that it only records if the current per-CPU page was allocated
from PFMEMALLOC reserve and uses another partial slab if the caller does
not have the necessary GFP or process flags. This was found to be
sufficient in tests to avoid hangs due to SLUB generally maintaining
smaller lists than SLAB.
In low-memory conditions it does mean that !PFMEMALLOC allocators can fail
a slab allocation even though free objects are available because they are
being preserved for callers that are freeing pages.
[a.p.zijlstra@chello.nl: Original implementation]
[sebastian@breakpoint.cc: Correct order of page flag clearing]
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: David Miller <davem@davemloft.net>
Cc: Neil Brown <neilb@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-07-31 23:43:58 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 23:57:00 +00:00
|
|
|
/*
|
|
|
|
* Checks whether it makes sense to retry the reclaim to make a forward progress
|
|
|
|
* for the given allocation request.
|
2017-05-03 21:52:16 +00:00
|
|
|
*
|
|
|
|
* We give up when we either have tried MAX_RECLAIM_RETRIES in a row
|
|
|
|
* without success, or when we couldn't even meet the watermark if we
|
|
|
|
* reclaimed all remaining pages on the LRU lists.
|
2016-05-20 23:57:00 +00:00
|
|
|
*
|
|
|
|
* Returns true if a retry is viable or false to enter the oom path.
|
|
|
|
*/
|
|
|
|
static inline bool
|
|
|
|
should_reclaim_retry(gfp_t gfp_mask, unsigned order,
|
|
|
|
struct alloc_context *ac, int alloc_flags,
|
2016-10-08 00:00:40 +00:00
|
|
|
bool did_some_progress, int *no_progress_loops)
|
2016-05-20 23:57:00 +00:00
|
|
|
{
|
|
|
|
struct zone *zone;
|
|
|
|
struct zoneref *z;
|
2018-10-26 22:03:31 +00:00
|
|
|
bool ret = false;
|
2016-05-20 23:57:00 +00:00
|
|
|
|
2016-10-08 00:00:40 +00:00
|
|
|
/*
|
|
|
|
* Costly allocations might have made a progress but this doesn't mean
|
|
|
|
* their order will become available due to high fragmentation so
|
|
|
|
* always increment the no progress counter for them
|
|
|
|
*/
|
|
|
|
if (did_some_progress && order <= PAGE_ALLOC_COSTLY_ORDER)
|
|
|
|
*no_progress_loops = 0;
|
|
|
|
else
|
|
|
|
(*no_progress_loops)++;
|
|
|
|
|
2016-05-20 23:57:00 +00:00
|
|
|
/*
|
|
|
|
* Make sure we converge to OOM if we cannot make any progress
|
|
|
|
* several times in the row.
|
|
|
|
*/
|
mm: try to exhaust highatomic reserve before the OOM
I got OOM report from production team with v4.4 kernel. It had enough
free memory but failed to allocate GFP_KERNEL order-0 page and finally
encountered OOM kill. It occured during QA process which launches
several apps, switching and so on. It happned rarely. IOW, In normal
situation, it was not a problem but if we are unluck so that several
apps uses peak memory at the same time, it can happen. If we manage to
pass the phase, the system can go working well.
I could reproduce it with my test(memory spike easily. Look at below.
The reason is free pages(19M) of DMA32 zone are reserved for
HIGHORDERATOMIC and doesn't unreserved before the OOM.
balloon invoked oom-killer: gfp_mask=0x24280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), order=0, oom_score_adj=0
balloon cpuset=/ mems_allowed=0
CPU: 1 PID: 8473 Comm: balloon Tainted: G W OE 4.8.0-rc7-00219-g3f74c9559583-dirty #3161
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
dump_header+0x5c/0x1ce
oom_kill_process+0x22e/0x400
out_of_memory+0x1ac/0x210
__alloc_pages_nodemask+0x101e/0x1040
handle_mm_fault+0xa0a/0xbf0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:383949 inactive_anon:106724 isolated_anon:0
active_file:15 inactive_file:44 isolated_file:0
unevictable:0 dirty:0 writeback:24 unstable:0
slab_reclaimable:2483 slab_unreclaimable:3326
mapped:0 shmem:0 pagetables:1906 bounce:0
free:6898 free_pcp:291 free_cma:0
Node 0 active_anon:1535796kB inactive_anon:426896kB active_file:60kB inactive_file:176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:96kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1418 all_unreclaimable? no
DMA free:8188kB min:44kB low:56kB high:68kB active_anon:7648kB inactive_anon:0kB active_file:0kB inactive_file:4kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:20kB kernel_stack:0kB pagetables:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:19404kB min:5628kB low:7624kB high:9620kB active_anon:1528148kB inactive_anon:426896kB active_file:60kB inactive_file:420kB unevictable:0kB writepending:96kB present:2080640kB managed:2030092kB mlocked:0kB slab_reclaimable:9932kB slab_unreclaimable:13284kB kernel_stack:2496kB pagetables:7624kB bounce:0kB free_pcp:900kB local_pcp:112kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 0*4kB 0*8kB 0*16kB 0*32kB 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 2*4096kB (H) = 8192kB
DMA32: 7*4kB (H) 8*8kB (H) 30*16kB (H) 31*32kB (H) 14*64kB (H) 9*128kB (H) 2*256kB (H) 2*512kB (H) 4*1024kB (H) 5*2048kB (H) 0*4096kB = 19484kB
51131 total pagecache pages
50795 pages in swap cache
Swap cache stats: add 3532405601, delete 3532354806, find 124289150/1822712228
Free swap = 8kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12658 pages reserved
0 pages cma reserved
0 pages hwpoisoned
Another example exceeded the limit by the race is
in:imklog: page allocation failure: order:0, mode:0x2280020(GFP_ATOMIC|__GFP_NOTRACK)
CPU: 0 PID: 476 Comm: in:imklog Tainted: G E 4.8.0-rc7-00217-g266ef83c51e5-dirty #3135
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
warn_alloc_failed+0xdb/0x130
__alloc_pages_nodemask+0x4d6/0xdb0
new_slab+0x339/0x490
___slab_alloc.constprop.74+0x367/0x480
__slab_alloc.constprop.73+0x20/0x40
__kmalloc+0x1a4/0x1e0
alloc_indirect.isra.14+0x1d/0x50
virtqueue_add_sgs+0x1c4/0x470
__virtblk_add_req+0xae/0x1f0
virtio_queue_rq+0x12d/0x290
__blk_mq_run_hw_queue+0x239/0x370
blk_mq_run_hw_queue+0x8f/0xb0
blk_mq_insert_requests+0x18c/0x1a0
blk_mq_flush_plug_list+0x125/0x140
blk_flush_plug_list+0xc7/0x220
blk_finish_plug+0x2c/0x40
__do_page_cache_readahead+0x196/0x230
filemap_fault+0x448/0x4f0
ext4_filemap_fault+0x36/0x50
__do_fault+0x75/0x140
handle_mm_fault+0x84d/0xbe0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:363826 inactive_anon:121283 isolated_anon:32
active_file:65 inactive_file:152 isolated_file:0
unevictable:0 dirty:0 writeback:46 unstable:0
slab_reclaimable:2778 slab_unreclaimable:3070
mapped:112 shmem:0 pagetables:1822 bounce:0
free:9469 free_pcp:231 free_cma:0
Node 0 active_anon:1455304kB inactive_anon:485132kB active_file:260kB inactive_file:608kB unevictable:0kB isolated(anon):128kB isolated(file):0kB mapped:448kB dirty:0kB writeback:184kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:13641 all_unreclaimable? no
DMA free:7748kB min:44kB low:56kB high:68kB active_anon:7944kB inactive_anon:104kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:108kB kernel_stack:0kB pagetables:4kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:30128kB min:5628kB low:7624kB high:9620kB active_anon:1447360kB inactive_anon:485028kB active_file:260kB inactive_file:608kB unevictable:0kB writepending:184kB present:2080640kB managed:2030132kB mlocked:0kB slab_reclaimable:11112kB slab_unreclaimable:12172kB kernel_stack:2400kB pagetables:7284kB bounce:0kB free_pcp:924kB local_pcp:72kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 7*4kB (UE) 3*8kB (UH) 1*16kB (M) 0*32kB 2*64kB (U) 1*128kB (M) 1*256kB (U) 0*512kB 1*1024kB (U) 1*2048kB (U) 1*4096kB (H) = 7748kB
DMA32: 10*4kB (H) 3*8kB (H) 47*16kB (H) 38*32kB (H) 5*64kB (H) 1*128kB (H) 2*256kB (H) 3*512kB (H) 3*1024kB (H) 3*2048kB (H) 4*4096kB (H) = 30128kB
2775 total pagecache pages
2536 pages in swap cache
Swap cache stats: add 206786828, delete 206784292, find 7323106/106686077
Free swap = 108744kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12648 pages reserved
0 pages cma reserved
0 pages hwpoisoned
It's weird to show that zone has enough free memory above min watermark
but OOMed with 4K GFP_KERNEL allocation due to reserved highatomic
pages. As last resort, try to unreserve highatomic pages again and if
it has moved pages to non-highatmoc free list, retry reclaim once more.
Link: http://lkml.kernel.org/r/1476259429-18279-4-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Sangseok Lee <sangseok.lee@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-12-13 00:42:11 +00:00
|
|
|
if (*no_progress_loops > MAX_RECLAIM_RETRIES) {
|
|
|
|
/* Before OOM, exhaust highatomic_reserve */
|
2016-12-13 00:42:14 +00:00
|
|
|
return unreserve_highatomic_pageblock(ac, true);
|
mm: try to exhaust highatomic reserve before the OOM
I got OOM report from production team with v4.4 kernel. It had enough
free memory but failed to allocate GFP_KERNEL order-0 page and finally
encountered OOM kill. It occured during QA process which launches
several apps, switching and so on. It happned rarely. IOW, In normal
situation, it was not a problem but if we are unluck so that several
apps uses peak memory at the same time, it can happen. If we manage to
pass the phase, the system can go working well.
I could reproduce it with my test(memory spike easily. Look at below.
The reason is free pages(19M) of DMA32 zone are reserved for
HIGHORDERATOMIC and doesn't unreserved before the OOM.
balloon invoked oom-killer: gfp_mask=0x24280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), order=0, oom_score_adj=0
balloon cpuset=/ mems_allowed=0
CPU: 1 PID: 8473 Comm: balloon Tainted: G W OE 4.8.0-rc7-00219-g3f74c9559583-dirty #3161
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
dump_header+0x5c/0x1ce
oom_kill_process+0x22e/0x400
out_of_memory+0x1ac/0x210
__alloc_pages_nodemask+0x101e/0x1040
handle_mm_fault+0xa0a/0xbf0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:383949 inactive_anon:106724 isolated_anon:0
active_file:15 inactive_file:44 isolated_file:0
unevictable:0 dirty:0 writeback:24 unstable:0
slab_reclaimable:2483 slab_unreclaimable:3326
mapped:0 shmem:0 pagetables:1906 bounce:0
free:6898 free_pcp:291 free_cma:0
Node 0 active_anon:1535796kB inactive_anon:426896kB active_file:60kB inactive_file:176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:96kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1418 all_unreclaimable? no
DMA free:8188kB min:44kB low:56kB high:68kB active_anon:7648kB inactive_anon:0kB active_file:0kB inactive_file:4kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:20kB kernel_stack:0kB pagetables:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:19404kB min:5628kB low:7624kB high:9620kB active_anon:1528148kB inactive_anon:426896kB active_file:60kB inactive_file:420kB unevictable:0kB writepending:96kB present:2080640kB managed:2030092kB mlocked:0kB slab_reclaimable:9932kB slab_unreclaimable:13284kB kernel_stack:2496kB pagetables:7624kB bounce:0kB free_pcp:900kB local_pcp:112kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 0*4kB 0*8kB 0*16kB 0*32kB 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 2*4096kB (H) = 8192kB
DMA32: 7*4kB (H) 8*8kB (H) 30*16kB (H) 31*32kB (H) 14*64kB (H) 9*128kB (H) 2*256kB (H) 2*512kB (H) 4*1024kB (H) 5*2048kB (H) 0*4096kB = 19484kB
51131 total pagecache pages
50795 pages in swap cache
Swap cache stats: add 3532405601, delete 3532354806, find 124289150/1822712228
Free swap = 8kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12658 pages reserved
0 pages cma reserved
0 pages hwpoisoned
Another example exceeded the limit by the race is
in:imklog: page allocation failure: order:0, mode:0x2280020(GFP_ATOMIC|__GFP_NOTRACK)
CPU: 0 PID: 476 Comm: in:imklog Tainted: G E 4.8.0-rc7-00217-g266ef83c51e5-dirty #3135
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
dump_stack+0x63/0x90
warn_alloc_failed+0xdb/0x130
__alloc_pages_nodemask+0x4d6/0xdb0
new_slab+0x339/0x490
___slab_alloc.constprop.74+0x367/0x480
__slab_alloc.constprop.73+0x20/0x40
__kmalloc+0x1a4/0x1e0
alloc_indirect.isra.14+0x1d/0x50
virtqueue_add_sgs+0x1c4/0x470
__virtblk_add_req+0xae/0x1f0
virtio_queue_rq+0x12d/0x290
__blk_mq_run_hw_queue+0x239/0x370
blk_mq_run_hw_queue+0x8f/0xb0
blk_mq_insert_requests+0x18c/0x1a0
blk_mq_flush_plug_list+0x125/0x140
blk_flush_plug_list+0xc7/0x220
blk_finish_plug+0x2c/0x40
__do_page_cache_readahead+0x196/0x230
filemap_fault+0x448/0x4f0
ext4_filemap_fault+0x36/0x50
__do_fault+0x75/0x140
handle_mm_fault+0x84d/0xbe0
__do_page_fault+0x1dd/0x4d0
trace_do_page_fault+0x43/0x130
do_async_page_fault+0x1a/0xa0
async_page_fault+0x28/0x30
Mem-Info:
active_anon:363826 inactive_anon:121283 isolated_anon:32
active_file:65 inactive_file:152 isolated_file:0
unevictable:0 dirty:0 writeback:46 unstable:0
slab_reclaimable:2778 slab_unreclaimable:3070
mapped:112 shmem:0 pagetables:1822 bounce:0
free:9469 free_pcp:231 free_cma:0
Node 0 active_anon:1455304kB inactive_anon:485132kB active_file:260kB inactive_file:608kB unevictable:0kB isolated(anon):128kB isolated(file):0kB mapped:448kB dirty:0kB writeback:184kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:13641 all_unreclaimable? no
DMA free:7748kB min:44kB low:56kB high:68kB active_anon:7944kB inactive_anon:104kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:108kB kernel_stack:0kB pagetables:4kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 1952 1952 1952
DMA32 free:30128kB min:5628kB low:7624kB high:9620kB active_anon:1447360kB inactive_anon:485028kB active_file:260kB inactive_file:608kB unevictable:0kB writepending:184kB present:2080640kB managed:2030132kB mlocked:0kB slab_reclaimable:11112kB slab_unreclaimable:12172kB kernel_stack:2400kB pagetables:7284kB bounce:0kB free_pcp:924kB local_pcp:72kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 7*4kB (UE) 3*8kB (UH) 1*16kB (M) 0*32kB 2*64kB (U) 1*128kB (M) 1*256kB (U) 0*512kB 1*1024kB (U) 1*2048kB (U) 1*4096kB (H) = 7748kB
DMA32: 10*4kB (H) 3*8kB (H) 47*16kB (H) 38*32kB (H) 5*64kB (H) 1*128kB (H) 2*256kB (H) 3*512kB (H) 3*1024kB (H) 3*2048kB (H) 4*4096kB (H) = 30128kB
2775 total pagecache pages
2536 pages in swap cache
Swap cache stats: add 206786828, delete 206784292, find 7323106/106686077
Free swap = 108744kB
Total swap = 255996kB
524158 pages RAM
0 pages HighMem/MovableOnly
12648 pages reserved
0 pages cma reserved
0 pages hwpoisoned
It's weird to show that zone has enough free memory above min watermark
but OOMed with 4K GFP_KERNEL allocation due to reserved highatomic
pages. As last resort, try to unreserve highatomic pages again and if
it has moved pages to non-highatmoc free list, retry reclaim once more.
Link: http://lkml.kernel.org/r/1476259429-18279-4-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Sangseok Lee <sangseok.lee@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-12-13 00:42:11 +00:00
|
|
|
}
|
2016-05-20 23:57:00 +00:00
|
|
|
|
2016-07-28 22:47:05 +00:00
|
|
|
/*
|
|
|
|
* Keep reclaiming pages while there is a chance this will lead
|
|
|
|
* somewhere. If none of the target zones can satisfy our allocation
|
|
|
|
* request even if all reclaimable pages are considered then we are
|
|
|
|
* screwed and have to go OOM.
|
2016-05-20 23:57:00 +00:00
|
|
|
*/
|
2020-06-03 22:59:01 +00:00
|
|
|
for_each_zone_zonelist_nodemask(zone, z, ac->zonelist,
|
|
|
|
ac->highest_zoneidx, ac->nodemask) {
|
2016-05-20 23:57:00 +00:00
|
|
|
unsigned long available;
|
2016-05-20 23:57:03 +00:00
|
|
|
unsigned long reclaimable;
|
2017-02-22 23:42:00 +00:00
|
|
|
unsigned long min_wmark = min_wmark_pages(zone);
|
|
|
|
bool wmark;
|
2016-05-20 23:57:00 +00:00
|
|
|
|
2016-07-28 22:47:31 +00:00
|
|
|
available = reclaimable = zone_reclaimable_pages(zone);
|
|
|
|
available += zone_page_state_snapshot(zone, NR_FREE_PAGES);
|
2016-05-20 23:57:00 +00:00
|
|
|
|
|
|
|
/*
|
2017-05-03 21:52:16 +00:00
|
|
|
* Would the allocation succeed if we reclaimed all
|
|
|
|
* reclaimable pages?
|
2016-05-20 23:57:00 +00:00
|
|
|
*/
|
2017-02-22 23:42:00 +00:00
|
|
|
wmark = __zone_watermark_ok(zone, order, min_wmark,
|
2020-06-03 22:59:01 +00:00
|
|
|
ac->highest_zoneidx, alloc_flags, available);
|
2017-02-22 23:42:00 +00:00
|
|
|
trace_reclaim_retry_zone(z, order, reclaimable,
|
|
|
|
available, min_wmark, *no_progress_loops, wmark);
|
|
|
|
if (wmark) {
|
2016-05-20 23:57:03 +00:00
|
|
|
/*
|
|
|
|
* If we didn't make any progress and have a lot of
|
|
|
|
* dirty + writeback pages then we should wait for
|
|
|
|
* an IO to complete to slow down the reclaim and
|
|
|
|
* prevent from pre mature OOM
|
|
|
|
*/
|
|
|
|
if (!did_some_progress) {
|
2016-07-28 22:46:20 +00:00
|
|
|
unsigned long write_pending;
|
2016-05-20 23:57:03 +00:00
|
|
|
|
2016-07-28 22:47:31 +00:00
|
|
|
write_pending = zone_page_state_snapshot(zone,
|
|
|
|
NR_ZONE_WRITE_PENDING);
|
2016-05-20 23:57:03 +00:00
|
|
|
|
2016-07-28 22:46:20 +00:00
|
|
|
if (2 * write_pending > reclaimable) {
|
2016-05-20 23:57:03 +00:00
|
|
|
congestion_wait(BLK_RW_ASYNC, HZ/10);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
2016-07-28 22:47:31 +00:00
|
|
|
|
2018-10-26 22:03:31 +00:00
|
|
|
ret = true;
|
|
|
|
goto out;
|
2016-05-20 23:57:00 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-10-26 22:03:31 +00:00
|
|
|
out:
|
|
|
|
/*
|
|
|
|
* Memory allocation/reclaim might be called from a WQ context and the
|
|
|
|
* current implementation of the WQ concurrency control doesn't
|
|
|
|
* recognize that a particular WQ is congested if the worker thread is
|
|
|
|
* looping without ever sleeping. Therefore we have to do a short sleep
|
|
|
|
* here rather than calling cond_resched().
|
|
|
|
*/
|
|
|
|
if (current->flags & PF_WQ_WORKER)
|
|
|
|
schedule_timeout_uninterruptible(1);
|
|
|
|
else
|
|
|
|
cond_resched();
|
|
|
|
return ret;
|
2016-05-20 23:57:00 +00:00
|
|
|
}
|
|
|
|
|
mm, page_alloc: fix more premature OOM due to race with cpuset update
I would like to stress that this patchset aims to fix issues and cleanup
the code *within the existing documented semantics*, i.e. patch 1
ignores mempolicy restrictions if the set of allowed nodes has no
intersection with set of nodes allowed by cpuset. I believe discussing
potential changes of the semantics can be better done once we have a
baseline with no known bugs of the current semantics.
I've recently summarized the cpuset/mempolicy issues in a LSF/MM
proposal [1] and the discussion itself [2]. I've been trying to rewrite
the handling as proposed, with the idea that changing semantics to make
all mempolicies static wrt cpuset updates (and discarding the relative
and default modes) can be tried on top, as there's a high risk of being
rejected/reverted because somebody might still care about the removed
modes.
However I haven't yet figured out how to properly:
1) make mempolicies swappable instead of rebinding in place. I thought
mbind() already works that way and uses refcounting to avoid
use-after-free of the old policy by a parallel allocation, but turns
out true refcounting is only done for shared (shmem) mempolicies, and
the actual protection for mbind() comes from mmap_sem. Extending the
refcounting means more overhead in allocator hot path. Also swapping
whole mempolicies means that we have to allocate the new ones, which
can fail, and reverting of the partially done work also means
allocating (note that mbind() doesn't care and will just leave part
of the range updated and part not updated when returning -ENOMEM...).
2) make cpuset's task->mems_allowed also swappable (after converting it
from nodemask to zonelist, which is the easy part) for mostly the
same reasons.
The good news is that while trying to do the above, I've at least
figured out how to hopefully close the remaining premature OOM's, and do
a buch of cleanups on top, removing quite some of the code that was also
supposed to prevent the cpuset update races, but doesn't work anymore
nowadays. This should fix the most pressing concerns with this topic
and give us a better baseline before either proceeding with the original
proposal, or pushing a change of semantics that removes the problem 1)
above. I'd be then fine with trying to change the semantic first and
rewrite later.
Patchset has been tested with the LTP cpuset01 stress test.
[1] https://lkml.kernel.org/r/4c44a589-5fd8-08d0-892c-e893bb525b71@suse.cz
[2] https://lwn.net/Articles/717797/
[3] https://marc.info/?l=linux-mm&m=149191957922828&w=2
This patch (of 6):
Commit e47483bca2cc ("mm, page_alloc: fix premature OOM when racing with
cpuset mems update") has fixed known recent regressions found by LTP's
cpuset01 testcase. I have however found that by modifying the testcase
to use per-vma mempolicies via bind(2) instead of per-task mempolicies
via set_mempolicy(2), the premature OOM still happens and the issue is
much older.
The root of the problem is that the cpuset's mems_allowed and
mempolicy's nodemask can temporarily have no intersection, thus
get_page_from_freelist() cannot find any usable zone. The current
semantic for empty intersection is to ignore mempolicy's nodemask and
honour cpuset restrictions. This is checked in node_zonelist(), but the
racy update can happen after we already passed the check. Such races
should be protected by the seqlock task->mems_allowed_seq, but it
doesn't work here, because 1) mpol_rebind_mm() does not happen under
seqlock for write, and doing so would lead to deadlock, as it takes
mmap_sem for write, while the allocation can have mmap_sem for read when
it's taking the seqlock for read. And 2) the seqlock cookie of callers
of node_zonelist() (alloc_pages_vma() and alloc_pages_current()) is
different than the one of __alloc_pages_slowpath(), so there's still a
potential race window.
This patch fixes the issue by having __alloc_pages_slowpath() check for
empty intersection of cpuset and ac->nodemask before OOM or allocation
failure. If it's indeed empty, the nodemask is ignored and allocation
retried, which mimics node_zonelist(). This works fine, because almost
all callers of __alloc_pages_nodemask are obtaining the nodemask via
node_zonelist(). The only exception is new_node_page() from hotplug,
where the potential violation of nodemask isn't an issue, as there's
already a fallback allocation attempt without any nodemask. If there's
a future caller that needs to have its specific nodemask honoured over
task's cpuset restrictions, we'll have to e.g. add a gfp flag for that.
Link: http://lkml.kernel.org/r/20170517081140.30654-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-06 22:39:56 +00:00
|
|
|
static inline bool
|
|
|
|
check_retry_cpuset(int cpuset_mems_cookie, struct alloc_context *ac)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* It's possible that cpuset's mems_allowed and the nodemask from
|
|
|
|
* mempolicy don't intersect. This should be normally dealt with by
|
|
|
|
* policy_nodemask(), but it's possible to race with cpuset update in
|
|
|
|
* such a way the check therein was true, and then it became false
|
|
|
|
* before we got our cpuset_mems_cookie here.
|
|
|
|
* This assumes that for all allocations, ac->nodemask can come only
|
|
|
|
* from MPOL_BIND mempolicy (whose documented semantics is to be ignored
|
|
|
|
* when it does not intersect with the cpuset restrictions) or the
|
|
|
|
* caller can deal with a violated nodemask.
|
|
|
|
*/
|
|
|
|
if (cpusets_enabled() && ac->nodemask &&
|
|
|
|
!cpuset_nodemask_valid_mems_allowed(ac->nodemask)) {
|
|
|
|
ac->nodemask = NULL;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* When updating a task's mems_allowed or mempolicy nodemask, it is
|
|
|
|
* possible to race with parallel threads in such a way that our
|
|
|
|
* allocation can fail while the mask is being updated. If we are about
|
|
|
|
* to fail, check if the cpuset changed during allocation and if so,
|
|
|
|
* retry.
|
|
|
|
*/
|
|
|
|
if (read_mems_allowed_retry(cpuset_mems_cookie))
|
|
|
|
return true;
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2009-06-16 22:31:57 +00:00
|
|
|
static inline struct page *
|
|
|
|
__alloc_pages_slowpath(gfp_t gfp_mask, unsigned int order,
|
2015-02-11 23:25:41 +00:00
|
|
|
struct alloc_context *ac)
|
2009-06-16 22:31:57 +00:00
|
|
|
{
|
2015-11-07 00:28:21 +00:00
|
|
|
bool can_direct_reclaim = gfp_mask & __GFP_DIRECT_RECLAIM;
|
mm, compaction: restrict async compaction to pageblocks of same migratetype
The migrate scanner in async compaction is currently limited to
MIGRATE_MOVABLE pageblocks. This is a heuristic intended to reduce
latency, based on the assumption that non-MOVABLE pageblocks are
unlikely to contain movable pages.
However, with the exception of THP's, most high-order allocations are
not movable. Should the async compaction succeed, this increases the
chance that the non-MOVABLE allocations will fallback to a MOVABLE
pageblock, making the long-term fragmentation worse.
This patch attempts to help the situation by changing async direct
compaction so that the migrate scanner only scans the pageblocks of the
requested migratetype. If it's a non-MOVABLE type and there are such
pageblocks that do contain movable pages, chances are that the
allocation can succeed within one of such pageblocks, removing the need
for a fallback. If that fails, the subsequent sync attempt will ignore
this restriction.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 30%. The number of movable allocations falling back is reduced by
12%.
Link: http://lkml.kernel.org/r/20170307131545.28577-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:49 +00:00
|
|
|
const bool costly_order = order > PAGE_ALLOC_COSTLY_ORDER;
|
2009-06-16 22:31:57 +00:00
|
|
|
struct page *page = NULL;
|
2016-05-20 00:13:38 +00:00
|
|
|
unsigned int alloc_flags;
|
2009-06-16 22:31:57 +00:00
|
|
|
unsigned long did_some_progress;
|
2017-01-24 23:18:38 +00:00
|
|
|
enum compact_priority compact_priority;
|
2016-05-20 23:56:53 +00:00
|
|
|
enum compact_result compact_result;
|
2017-01-24 23:18:38 +00:00
|
|
|
int compaction_retries;
|
|
|
|
int no_progress_loops;
|
|
|
|
unsigned int cpuset_mems_cookie;
|
2017-09-06 23:24:50 +00:00
|
|
|
int reserve_flags;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2015-11-07 00:28:21 +00:00
|
|
|
/*
|
|
|
|
* We also sanity check to catch abuse of atomic reserves being used by
|
|
|
|
* callers that are not in atomic context.
|
|
|
|
*/
|
|
|
|
if (WARN_ON_ONCE((gfp_mask & (__GFP_ATOMIC|__GFP_DIRECT_RECLAIM)) ==
|
|
|
|
(__GFP_ATOMIC|__GFP_DIRECT_RECLAIM)))
|
|
|
|
gfp_mask &= ~__GFP_ATOMIC;
|
|
|
|
|
2017-01-24 23:18:38 +00:00
|
|
|
retry_cpuset:
|
|
|
|
compaction_retries = 0;
|
|
|
|
no_progress_loops = 0;
|
|
|
|
compact_priority = DEF_COMPACT_PRIORITY;
|
|
|
|
cpuset_mems_cookie = read_mems_allowed_begin();
|
2017-02-22 23:46:19 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The fast path uses conservative alloc_flags to succeed only until
|
|
|
|
* kswapd needs to be woken up, and to avoid the cost of setting up
|
|
|
|
* alloc_flags precisely. So we do that now.
|
|
|
|
*/
|
|
|
|
alloc_flags = gfp_to_alloc_flags(gfp_mask);
|
|
|
|
|
mm, page_alloc: fix premature OOM when racing with cpuset mems update
Ganapatrao Kulkarni reported that the LTP test cpuset01 in stress mode
triggers OOM killer in few seconds, despite lots of free memory. The
test attempts to repeatedly fault in memory in one process in a cpuset,
while changing allowed nodes of the cpuset between 0 and 1 in another
process.
The problem comes from insufficient protection against cpuset changes,
which can cause get_page_from_freelist() to consider all zones as
non-eligible due to nodemask and/or current->mems_allowed. This was
masked in the past by sufficient retries, but since commit 682a3385e773
("mm, page_alloc: inline the fast path of the zonelist iterator") we fix
the preferred_zoneref once, and don't iterate over the whole zonelist in
further attempts, thus the only eligible zones might be placed in the
zonelist before our starting point and we always miss them.
A previous patch fixed this problem for current->mems_allowed. However,
cpuset changes also update the task's mempolicy nodemask. The fix has
two parts. We have to repeat the preferred_zoneref search when we
detect cpuset update by way of seqcount, and we have to check the
seqcount before considering OOM.
[akpm@linux-foundation.org: fix typo in comment]
Link: http://lkml.kernel.org/r/20170120103843.24587-5-vbabka@suse.cz
Fixes: c33d6c06f60f ("mm, page_alloc: avoid looking up the first zone in a zonelist twice")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Ganapatrao Kulkarni <gpkulkarni@gmail.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-24 23:18:41 +00:00
|
|
|
/*
|
|
|
|
* We need to recalculate the starting point for the zonelist iterator
|
|
|
|
* because we might have used different nodemask in the fast path, or
|
|
|
|
* there was a cpuset modification and we are retrying - otherwise we
|
|
|
|
* could end up iterating over non-eligible zones endlessly.
|
|
|
|
*/
|
|
|
|
ac->preferred_zoneref = first_zones_zonelist(ac->zonelist,
|
2020-06-03 22:59:01 +00:00
|
|
|
ac->highest_zoneidx, ac->nodemask);
|
mm, page_alloc: fix premature OOM when racing with cpuset mems update
Ganapatrao Kulkarni reported that the LTP test cpuset01 in stress mode
triggers OOM killer in few seconds, despite lots of free memory. The
test attempts to repeatedly fault in memory in one process in a cpuset,
while changing allowed nodes of the cpuset between 0 and 1 in another
process.
The problem comes from insufficient protection against cpuset changes,
which can cause get_page_from_freelist() to consider all zones as
non-eligible due to nodemask and/or current->mems_allowed. This was
masked in the past by sufficient retries, but since commit 682a3385e773
("mm, page_alloc: inline the fast path of the zonelist iterator") we fix
the preferred_zoneref once, and don't iterate over the whole zonelist in
further attempts, thus the only eligible zones might be placed in the
zonelist before our starting point and we always miss them.
A previous patch fixed this problem for current->mems_allowed. However,
cpuset changes also update the task's mempolicy nodemask. The fix has
two parts. We have to repeat the preferred_zoneref search when we
detect cpuset update by way of seqcount, and we have to check the
seqcount before considering OOM.
[akpm@linux-foundation.org: fix typo in comment]
Link: http://lkml.kernel.org/r/20170120103843.24587-5-vbabka@suse.cz
Fixes: c33d6c06f60f ("mm, page_alloc: avoid looking up the first zone in a zonelist twice")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Ganapatrao Kulkarni <gpkulkarni@gmail.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-24 23:18:41 +00:00
|
|
|
if (!ac->preferred_zoneref->zone)
|
|
|
|
goto nopage;
|
|
|
|
|
2018-12-28 08:35:48 +00:00
|
|
|
if (alloc_flags & ALLOC_KSWAPD)
|
2018-04-05 23:25:16 +00:00
|
|
|
wake_all_kswapds(order, gfp_mask, ac);
|
2016-07-28 22:49:16 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The adjusted alloc_flags might result in immediate success, so try
|
|
|
|
* that first
|
|
|
|
*/
|
|
|
|
page = get_page_from_freelist(gfp_mask, order, alloc_flags, ac);
|
|
|
|
if (page)
|
|
|
|
goto got_pg;
|
|
|
|
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
/*
|
|
|
|
* For costly allocations, try direct compaction first, as it's likely
|
mm, compaction: restrict async compaction to pageblocks of same migratetype
The migrate scanner in async compaction is currently limited to
MIGRATE_MOVABLE pageblocks. This is a heuristic intended to reduce
latency, based on the assumption that non-MOVABLE pageblocks are
unlikely to contain movable pages.
However, with the exception of THP's, most high-order allocations are
not movable. Should the async compaction succeed, this increases the
chance that the non-MOVABLE allocations will fallback to a MOVABLE
pageblock, making the long-term fragmentation worse.
This patch attempts to help the situation by changing async direct
compaction so that the migrate scanner only scans the pageblocks of the
requested migratetype. If it's a non-MOVABLE type and there are such
pageblocks that do contain movable pages, chances are that the
allocation can succeed within one of such pageblocks, removing the need
for a fallback. If that fails, the subsequent sync attempt will ignore
this restriction.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 30%. The number of movable allocations falling back is reduced by
12%.
Link: http://lkml.kernel.org/r/20170307131545.28577-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:49 +00:00
|
|
|
* that we have enough base pages and don't need to reclaim. For non-
|
|
|
|
* movable high-order allocations, do that as well, as compaction will
|
|
|
|
* try prevent permanent fragmentation by migrating from blocks of the
|
|
|
|
* same migratetype.
|
|
|
|
* Don't try this for allocations that are allowed to ignore
|
|
|
|
* watermarks, as the ALLOC_NO_WATERMARKS attempt didn't yet happen.
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
*/
|
mm, compaction: restrict async compaction to pageblocks of same migratetype
The migrate scanner in async compaction is currently limited to
MIGRATE_MOVABLE pageblocks. This is a heuristic intended to reduce
latency, based on the assumption that non-MOVABLE pageblocks are
unlikely to contain movable pages.
However, with the exception of THP's, most high-order allocations are
not movable. Should the async compaction succeed, this increases the
chance that the non-MOVABLE allocations will fallback to a MOVABLE
pageblock, making the long-term fragmentation worse.
This patch attempts to help the situation by changing async direct
compaction so that the migrate scanner only scans the pageblocks of the
requested migratetype. If it's a non-MOVABLE type and there are such
pageblocks that do contain movable pages, chances are that the
allocation can succeed within one of such pageblocks, removing the need
for a fallback. If that fails, the subsequent sync attempt will ignore
this restriction.
In testing based on 4.9 kernel with stress-highalloc from mmtests
configured for order-4 GFP_KERNEL allocations, this patch has reduced
the number of unmovable allocations falling back to movable pageblocks
by 30%. The number of movable allocations falling back is reduced by
12%.
Link: http://lkml.kernel.org/r/20170307131545.28577-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-08 22:54:49 +00:00
|
|
|
if (can_direct_reclaim &&
|
|
|
|
(costly_order ||
|
|
|
|
(order > 0 && ac->migratetype != MIGRATE_MOVABLE))
|
|
|
|
&& !gfp_pfmemalloc_allowed(gfp_mask)) {
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
page = __alloc_pages_direct_compact(gfp_mask, order,
|
|
|
|
alloc_flags, ac,
|
2016-07-28 22:49:28 +00:00
|
|
|
INIT_COMPACT_PRIORITY,
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
&compact_result);
|
|
|
|
if (page)
|
|
|
|
goto got_pg;
|
|
|
|
|
mm, thp: tweak reclaim/compaction effort of local-only and all-node allocations
THP page faults now attempt a __GFP_THISNODE allocation first, which
should only compact existing free memory, followed by another attempt
that can allocate from any node using reclaim/compaction effort
specified by global defrag setting and madvise.
This patch makes the following changes to the scheme:
- Before the patch, the first allocation relies on a check for
pageblock order and __GFP_IO to prevent excessive reclaim. This
however affects also the second attempt, which is not limited to
single node.
Instead of that, reuse the existing check for costly order
__GFP_NORETRY allocations, and make sure the first THP attempt uses
__GFP_NORETRY. As a side-effect, all costly order __GFP_NORETRY
allocations will bail out if compaction needs reclaim, while
previously they only bailed out when compaction was deferred due to
previous failures.
This should be still acceptable within the __GFP_NORETRY semantics.
- Before the patch, the second allocation attempt (on all nodes) was
passing __GFP_NORETRY. This is redundant as the check for pageblock
order (discussed above) was stronger. It's also contrary to
madvise(MADV_HUGEPAGE) which means some effort to allocate THP is
requested.
After this patch, the second attempt doesn't pass __GFP_THISNODE nor
__GFP_NORETRY.
To sum up, THP page faults now try the following attempts:
1. local node only THP allocation with no reclaim, just compaction.
2. for madvised VMA's or when synchronous compaction is enabled always - THP
allocation from any node with effort determined by global defrag setting
and VMA madvise
3. fallback to base pages on any node
Link: http://lkml.kernel.org/r/08a3f4dd-c3ce-0009-86c5-9ee51aba8557@suse.cz
Fixes: b39d0ee2632d ("mm, page_alloc: avoid expensive reclaim when compaction may not succeed")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-14 00:29:04 +00:00
|
|
|
/*
|
|
|
|
* Checks for costly allocations with __GFP_NORETRY, which
|
|
|
|
* includes some THP page fault allocations
|
|
|
|
*/
|
|
|
|
if (costly_order && (gfp_mask & __GFP_NORETRY)) {
|
2019-09-04 19:54:22 +00:00
|
|
|
/*
|
|
|
|
* If allocating entire pageblock(s) and compaction
|
|
|
|
* failed because all zones are below low watermarks
|
|
|
|
* or is prohibited because it recently failed at this
|
mm, hugetlb: allow hugepage allocations to reclaim as needed
Commit b39d0ee2632d ("mm, page_alloc: avoid expensive reclaim when
compaction may not succeed") has chnaged the allocator to bail out from
the allocator early to prevent from a potentially excessive memory
reclaim. __GFP_RETRY_MAYFAIL is designed to retry the allocation,
reclaim and compaction loop as long as there is a reasonable chance to
make forward progress. Neither COMPACT_SKIPPED nor COMPACT_DEFERRED at
the INIT_COMPACT_PRIORITY compaction attempt gives this feedback.
The most obvious affected subsystem is hugetlbfs which allocates huge
pages based on an admin request (or via admin configured overcommit). I
have done a simple test which tries to allocate half of the memory for
hugetlb pages while the memory is full of a clean page cache. This is
not an unusual situation because we try to cache as much of the memory
as possible and sysctl/sysfs interface to allocate huge pages is there
for flexibility to allocate hugetlb pages at any time.
System has 1GB of RAM and we are requesting 515MB worth of hugetlb pages
after the memory is prefilled by a clean page cache:
root@test1:~# cat hugetlb_test.sh
set -x
echo 0 > /proc/sys/vm/nr_hugepages
echo 3 > /proc/sys/vm/drop_caches
echo 1 > /proc/sys/vm/compact_memory
dd if=/mnt/data/file-1G of=/dev/null bs=$((4<<10))
TS=$(date +%s)
echo 256 > /proc/sys/vm/nr_hugepages
cat /proc/sys/vm/nr_hugepages
The results for 2 consecutive runs on clean 5.3
root@test1:~# sh hugetlb_test.sh
+ echo 0
+ echo 3
+ echo 1
+ dd if=/mnt/data/file-1G of=/dev/null bs=4096
262144+0 records in
262144+0 records out
1073741824 bytes (1.1 GB) copied, 21.0694 s, 51.0 MB/s
+ date +%s
+ TS=1569905284
+ echo 256
+ cat /proc/sys/vm/nr_hugepages
256
root@test1:~# sh hugetlb_test.sh
+ echo 0
+ echo 3
+ echo 1
+ dd if=/mnt/data/file-1G of=/dev/null bs=4096
262144+0 records in
262144+0 records out
1073741824 bytes (1.1 GB) copied, 21.7548 s, 49.4 MB/s
+ date +%s
+ TS=1569905311
+ echo 256
+ cat /proc/sys/vm/nr_hugepages
256
Now with b39d0ee2632d applied
root@test1:~# sh hugetlb_test.sh
+ echo 0
+ echo 3
+ echo 1
+ dd if=/mnt/data/file-1G of=/dev/null bs=4096
262144+0 records in
262144+0 records out
1073741824 bytes (1.1 GB) copied, 20.1815 s, 53.2 MB/s
+ date +%s
+ TS=1569905516
+ echo 256
+ cat /proc/sys/vm/nr_hugepages
11
root@test1:~# sh hugetlb_test.sh
+ echo 0
+ echo 3
+ echo 1
+ dd if=/mnt/data/file-1G of=/dev/null bs=4096
262144+0 records in
262144+0 records out
1073741824 bytes (1.1 GB) copied, 21.9485 s, 48.9 MB/s
+ date +%s
+ TS=1569905541
+ echo 256
+ cat /proc/sys/vm/nr_hugepages
12
The success rate went down by factor of 20!
Although hugetlb allocation requests might fail and it is reasonable to
expect them to under extremely fragmented memory or when the memory is
under a heavy pressure but the above situation is not that case.
Fix the regression by reverting back to the previous behavior for
__GFP_RETRY_MAYFAIL requests and disable the beail out heuristic for
those requests.
Mike said:
: hugetlbfs allocations are commonly done via sysctl/sysfs shortly after
: boot where this may not be as much of an issue. However, I am aware of at
: least three use cases where allocations are made after the system has been
: up and running for quite some time:
:
: - DB reconfiguration. If sysctl/sysfs fails to get required number of
: huge pages, system is rebooted to perform allocation after boot.
:
: - VM provisioning. If unable get required number of huge pages, fall
: back to base pages.
:
: - An application that does not preallocate pool, but rather allocates
: pages at fault time for optimal NUMA locality.
:
: In all cases, I would expect b39d0ee2632d to cause regressions and
: noticable behavior changes.
:
: My quick/limited testing in
: https://lkml.kernel.org/r/3468b605-a3a9-6978-9699-57c52a90bd7e@oracle.com
: was insufficient. It was also mentioned that if something like
: b39d0ee2632d went forward, I would like exemptions for __GFP_RETRY_MAYFAIL
: requests as in this patch.
[mhocko@suse.com: reworded changelog]
Link: http://lkml.kernel.org/r/20191007075548.12456-1-mhocko@kernel.org
Fixes: b39d0ee2632d ("mm, page_alloc: avoid expensive reclaim when compaction may not succeed")
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-10-14 21:12:04 +00:00
|
|
|
* order, fail immediately unless the allocator has
|
|
|
|
* requested compaction and reclaim retry.
|
2019-09-04 19:54:22 +00:00
|
|
|
*
|
|
|
|
* Reclaim is
|
|
|
|
* - potentially very expensive because zones are far
|
|
|
|
* below their low watermarks or this is part of very
|
|
|
|
* bursty high order allocations,
|
|
|
|
* - not guaranteed to help because isolate_freepages()
|
|
|
|
* may not iterate over freed pages as part of its
|
|
|
|
* linear scan, and
|
|
|
|
* - unlikely to make entire pageblocks free on its
|
|
|
|
* own.
|
|
|
|
*/
|
|
|
|
if (compact_result == COMPACT_SKIPPED ||
|
|
|
|
compact_result == COMPACT_DEFERRED)
|
|
|
|
goto nopage;
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
|
|
|
|
/*
|
mm, page_alloc: make THP-specific decisions more generic
Since THP allocations during page faults can be costly, extra decisions
are employed for them to avoid excessive reclaim and compaction, if the
initial compaction doesn't look promising. The detection has never been
perfect as there is no gfp flag specific to THP allocations. At this
moment it checks the whole combination of flags that makes up
GFP_TRANSHUGE, and hopes that no other users of such combination exist,
or would mind being treated the same way. Extra care is also taken to
separate allocations from khugepaged, where latency doesn't matter that
much.
It is however possible to distinguish these allocations in a simpler and
more reliable way. The key observation is that after the initial
compaction followed by the first iteration of "standard"
reclaim/compaction, both __GFP_NORETRY allocations and costly
allocations without __GFP_REPEAT are declared as failures:
/* Do not loop if specifically requested */
if (gfp_mask & __GFP_NORETRY)
goto nopage;
/*
* Do not retry costly high order allocations unless they are
* __GFP_REPEAT
*/
if (order > PAGE_ALLOC_COSTLY_ORDER && !(gfp_mask & __GFP_REPEAT))
goto nopage;
This means we can further distinguish allocations that are costly order
*and* additionally include the __GFP_NORETRY flag. As it happens,
GFP_TRANSHUGE allocations do already fall into this category. This will
also allow other costly allocations with similar high-order benefit vs
latency considerations to use this semantic. Furthermore, we can
distinguish THP allocations that should try a bit harder (such as from
khugepageed) by removing __GFP_NORETRY, as will be done in the next
patch.
Link: http://lkml.kernel.org/r/20160721073614.24395-6-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:22 +00:00
|
|
|
* Looks like reclaim/compaction is worth trying, but
|
|
|
|
* sync compaction could be very expensive, so keep
|
mm, thp: remove __GFP_NORETRY from khugepaged and madvised allocations
After the previous patch, we can distinguish costly allocations that
should be really lightweight, such as THP page faults, with
__GFP_NORETRY. This means we don't need to recognize khugepaged
allocations via PF_KTHREAD anymore. We can also change THP page faults
in areas where madvise(MADV_HUGEPAGE) was used to try as hard as
khugepaged, as the process has indicated that it benefits from THP's and
is willing to pay some initial latency costs.
We can also make the flags handling less cryptic by distinguishing
GFP_TRANSHUGE_LIGHT (no reclaim at all, default mode in page fault) from
GFP_TRANSHUGE (only direct reclaim, khugepaged default). Adding
__GFP_NORETRY or __GFP_KSWAPD_RECLAIM is done where needed.
The patch effectively changes the current GFP_TRANSHUGE users as
follows:
* get_huge_zero_page() - the zero page lifetime should be relatively
long and it's shared by multiple users, so it's worth spending some
effort on it. We use GFP_TRANSHUGE, and __GFP_NORETRY is not added.
This also restores direct reclaim to this allocation, which was
unintentionally removed by commit e4a49efe4e7e ("mm: thp: set THP defrag
by default to madvise and add a stall-free defrag option")
* alloc_hugepage_khugepaged_gfpmask() - this is khugepaged, so latency
is not an issue. So if khugepaged "defrag" is enabled (the default), do
reclaim via GFP_TRANSHUGE without __GFP_NORETRY. We can remove the
PF_KTHREAD check from page alloc.
As a side-effect, khugepaged will now no longer check if the initial
compaction was deferred or contended. This is OK, as khugepaged sleep
times between collapsion attempts are long enough to prevent noticeable
disruption, so we should allow it to spend some effort.
* migrate_misplaced_transhuge_page() - already was masking out
__GFP_RECLAIM, so just convert to GFP_TRANSHUGE_LIGHT which is
equivalent.
* alloc_hugepage_direct_gfpmask() - vma's with VM_HUGEPAGE (via madvise)
are now allocating without __GFP_NORETRY. Other vma's keep using
__GFP_NORETRY if direct reclaim/compaction is at all allowed (by default
it's allowed only for madvised vma's). The rest is conversion to
GFP_TRANSHUGE(_LIGHT).
[mhocko@suse.com: suggested GFP_TRANSHUGE_LIGHT]
Link: http://lkml.kernel.org/r/20160721073614.24395-7-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:25 +00:00
|
|
|
* using async compaction.
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
*/
|
2016-07-28 22:49:28 +00:00
|
|
|
compact_priority = INIT_COMPACT_PRIORITY;
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
}
|
|
|
|
}
|
2016-07-28 22:49:16 +00:00
|
|
|
|
2016-07-28 22:49:13 +00:00
|
|
|
retry:
|
2016-07-28 22:49:16 +00:00
|
|
|
/* Ensure kswapd doesn't accidentally go to sleep as long as we loop */
|
2018-12-28 08:35:48 +00:00
|
|
|
if (alloc_flags & ALLOC_KSWAPD)
|
2018-04-05 23:25:16 +00:00
|
|
|
wake_all_kswapds(order, gfp_mask, ac);
|
2016-07-28 22:49:13 +00:00
|
|
|
|
2017-09-06 23:24:50 +00:00
|
|
|
reserve_flags = __gfp_pfmemalloc_flags(gfp_mask);
|
|
|
|
if (reserve_flags)
|
2020-08-07 06:26:04 +00:00
|
|
|
alloc_flags = current_alloc_flags(gfp_mask, reserve_flags);
|
2016-07-28 22:49:16 +00:00
|
|
|
|
2016-06-03 21:56:01 +00:00
|
|
|
/*
|
mm, page_alloc: actually ignore mempolicies for high priority allocations
__alloc_pages_slowpath() has for a long time contained code to ignore
node restrictions from memory policies for high priority allocations.
The current code that resets the zonelist iterator however does
effectively nothing after commit 7810e6781e0f ("mm, page_alloc: do not
break __GFP_THISNODE by zonelist reset") removed a buggy zonelist reset.
Even before that commit, mempolicy restrictions were still not ignored,
as they are passed in ac->nodemask which is untouched by the code.
We can either remove the code, or make it work as intended. Since
ac->nodemask can be set from task's mempolicy via alloc_pages_current()
and thus also alloc_pages(), it may indeed affect kernel allocations,
and it makes sense to ignore it to allow progress for high priority
allocations.
Thus, this patch resets ac->nodemask to NULL in such cases. This
assumes all callers can handle it (i.e. there are no guarantees as in
the case of __GFP_THISNODE) which seems to be the case. The same
assumption is already present in check_retry_cpuset() for some time.
The expected effect is that high priority kernel allocations in the
context of userspace tasks (e.g. OOM victims) restricted by mempolicies
will have higher chance to succeed if they are restricted to nodes with
depleted memory, while there are other nodes with free memory left.
It's not a new intention, but for the first time the code will match the
intention, AFAICS. It was intended by commit 183f6371aac2 ("mm: ignore
mempolicies when using ALLOC_NO_WATERMARK") in v3.6 but I think it never
really worked, as mempolicy restriction was already encoded in nodemask,
not zonelist, at that time.
So originally that was for ALLOC_NO_WATERMARK only. Then it was
adjusted by e46e7b77c909 ("mm, page_alloc: recalculate the preferred
zoneref if the context can ignore memory policies") and cd04ae1e2dc8
("mm, oom: do not rely on TIF_MEMDIE for memory reserves access") to the
current state. So even GFP_ATOMIC would now ignore mempolicies after
the initial attempts fail - if the code worked as people thought it
does.
Link: http://lkml.kernel.org/r/20180612122624.8045-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:05 +00:00
|
|
|
* Reset the nodemask and zonelist iterators if memory policies can be
|
|
|
|
* ignored. These allocations are high priority and system rather than
|
|
|
|
* user oriented.
|
2016-06-03 21:56:01 +00:00
|
|
|
*/
|
2017-09-06 23:24:50 +00:00
|
|
|
if (!(alloc_flags & ALLOC_CPUSET) || reserve_flags) {
|
mm, page_alloc: actually ignore mempolicies for high priority allocations
__alloc_pages_slowpath() has for a long time contained code to ignore
node restrictions from memory policies for high priority allocations.
The current code that resets the zonelist iterator however does
effectively nothing after commit 7810e6781e0f ("mm, page_alloc: do not
break __GFP_THISNODE by zonelist reset") removed a buggy zonelist reset.
Even before that commit, mempolicy restrictions were still not ignored,
as they are passed in ac->nodemask which is untouched by the code.
We can either remove the code, or make it work as intended. Since
ac->nodemask can be set from task's mempolicy via alloc_pages_current()
and thus also alloc_pages(), it may indeed affect kernel allocations,
and it makes sense to ignore it to allow progress for high priority
allocations.
Thus, this patch resets ac->nodemask to NULL in such cases. This
assumes all callers can handle it (i.e. there are no guarantees as in
the case of __GFP_THISNODE) which seems to be the case. The same
assumption is already present in check_retry_cpuset() for some time.
The expected effect is that high priority kernel allocations in the
context of userspace tasks (e.g. OOM victims) restricted by mempolicies
will have higher chance to succeed if they are restricted to nodes with
depleted memory, while there are other nodes with free memory left.
It's not a new intention, but for the first time the code will match the
intention, AFAICS. It was intended by commit 183f6371aac2 ("mm: ignore
mempolicies when using ALLOC_NO_WATERMARK") in v3.6 but I think it never
really worked, as mempolicy restriction was already encoded in nodemask,
not zonelist, at that time.
So originally that was for ALLOC_NO_WATERMARK only. Then it was
adjusted by e46e7b77c909 ("mm, page_alloc: recalculate the preferred
zoneref if the context can ignore memory policies") and cd04ae1e2dc8
("mm, oom: do not rely on TIF_MEMDIE for memory reserves access") to the
current state. So even GFP_ATOMIC would now ignore mempolicies after
the initial attempts fail - if the code worked as people thought it
does.
Link: http://lkml.kernel.org/r/20180612122624.8045-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:05 +00:00
|
|
|
ac->nodemask = NULL;
|
2016-06-03 21:56:01 +00:00
|
|
|
ac->preferred_zoneref = first_zones_zonelist(ac->zonelist,
|
2020-06-03 22:59:01 +00:00
|
|
|
ac->highest_zoneidx, ac->nodemask);
|
2016-06-03 21:56:01 +00:00
|
|
|
}
|
|
|
|
|
2016-07-28 22:49:16 +00:00
|
|
|
/* Attempt with potentially adjusted zonelist and alloc_flags */
|
2016-07-28 22:49:13 +00:00
|
|
|
page = get_page_from_freelist(gfp_mask, order, alloc_flags, ac);
|
2005-11-14 00:06:43 +00:00
|
|
|
if (page)
|
|
|
|
goto got_pg;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2015-11-07 00:28:21 +00:00
|
|
|
/* Caller is not willing to reclaim, we can't balance anything */
|
2017-02-22 23:46:19 +00:00
|
|
|
if (!can_direct_reclaim)
|
2005-04-16 22:20:36 +00:00
|
|
|
goto nopage;
|
|
|
|
|
2017-02-22 23:46:19 +00:00
|
|
|
/* Avoid recursion of direct reclaim */
|
|
|
|
if (current->flags & PF_MEMALLOC)
|
2009-07-29 22:02:06 +00:00
|
|
|
goto nopage;
|
|
|
|
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
/* Try direct reclaim and then allocating */
|
|
|
|
page = __alloc_pages_direct_reclaim(gfp_mask, order, alloc_flags, ac,
|
|
|
|
&did_some_progress);
|
|
|
|
if (page)
|
|
|
|
goto got_pg;
|
|
|
|
|
|
|
|
/* Try direct compaction and then allocating */
|
2015-02-11 23:25:41 +00:00
|
|
|
page = __alloc_pages_direct_compact(gfp_mask, order, alloc_flags, ac,
|
2016-07-28 22:49:28 +00:00
|
|
|
compact_priority, &compact_result);
|
2010-05-24 21:32:30 +00:00
|
|
|
if (page)
|
|
|
|
goto got_pg;
|
2014-06-04 23:08:30 +00:00
|
|
|
|
2015-06-24 23:57:21 +00:00
|
|
|
/* Do not loop if specifically requested */
|
|
|
|
if (gfp_mask & __GFP_NORETRY)
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
goto nopage;
|
2015-06-24 23:57:21 +00:00
|
|
|
|
2016-05-20 23:57:00 +00:00
|
|
|
/*
|
|
|
|
* Do not retry costly high order allocations unless they are
|
2017-07-12 21:36:45 +00:00
|
|
|
* __GFP_RETRY_MAYFAIL
|
2016-05-20 23:57:00 +00:00
|
|
|
*/
|
2017-07-12 21:36:45 +00:00
|
|
|
if (costly_order && !(gfp_mask & __GFP_RETRY_MAYFAIL))
|
mm, page_alloc: restructure direct compaction handling in slowpath
The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure. Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success. On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.
Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop. This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions. It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted. Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.
This patch attempts to restructure the code with only minimal functional
changes. The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.
The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation. This better
matches the checks in reclaim's shrink_zones().
There are two other smaller functional changes. One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction. This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special. It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.
The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt. As argued above, this is intentional. If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.
The main outcome of this patch should be simpler code. Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional. Now the code matches the logic better.
The change also enable the following patches.
Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:49:19 +00:00
|
|
|
goto nopage;
|
2016-05-20 23:57:00 +00:00
|
|
|
|
|
|
|
if (should_reclaim_retry(gfp_mask, order, ac, alloc_flags,
|
2016-10-08 00:00:40 +00:00
|
|
|
did_some_progress > 0, &no_progress_loops))
|
2016-05-20 23:57:00 +00:00
|
|
|
goto retry;
|
|
|
|
|
2016-05-20 23:57:06 +00:00
|
|
|
/*
|
|
|
|
* It doesn't make any sense to retry for the compaction if the order-0
|
|
|
|
* reclaim is not able to make any progress because the current
|
|
|
|
* implementation of the compaction depends on the sufficient amount
|
|
|
|
* of free memory (see __compaction_suitable)
|
|
|
|
*/
|
|
|
|
if (did_some_progress > 0 &&
|
2016-05-20 23:57:12 +00:00
|
|
|
should_compact_retry(ac, order, alloc_flags,
|
2016-07-28 22:49:28 +00:00
|
|
|
compact_result, &compact_priority,
|
2016-10-08 00:00:31 +00:00
|
|
|
&compaction_retries))
|
2016-05-20 23:57:06 +00:00
|
|
|
goto retry;
|
|
|
|
|
mm, page_alloc: fix more premature OOM due to race with cpuset update
I would like to stress that this patchset aims to fix issues and cleanup
the code *within the existing documented semantics*, i.e. patch 1
ignores mempolicy restrictions if the set of allowed nodes has no
intersection with set of nodes allowed by cpuset. I believe discussing
potential changes of the semantics can be better done once we have a
baseline with no known bugs of the current semantics.
I've recently summarized the cpuset/mempolicy issues in a LSF/MM
proposal [1] and the discussion itself [2]. I've been trying to rewrite
the handling as proposed, with the idea that changing semantics to make
all mempolicies static wrt cpuset updates (and discarding the relative
and default modes) can be tried on top, as there's a high risk of being
rejected/reverted because somebody might still care about the removed
modes.
However I haven't yet figured out how to properly:
1) make mempolicies swappable instead of rebinding in place. I thought
mbind() already works that way and uses refcounting to avoid
use-after-free of the old policy by a parallel allocation, but turns
out true refcounting is only done for shared (shmem) mempolicies, and
the actual protection for mbind() comes from mmap_sem. Extending the
refcounting means more overhead in allocator hot path. Also swapping
whole mempolicies means that we have to allocate the new ones, which
can fail, and reverting of the partially done work also means
allocating (note that mbind() doesn't care and will just leave part
of the range updated and part not updated when returning -ENOMEM...).
2) make cpuset's task->mems_allowed also swappable (after converting it
from nodemask to zonelist, which is the easy part) for mostly the
same reasons.
The good news is that while trying to do the above, I've at least
figured out how to hopefully close the remaining premature OOM's, and do
a buch of cleanups on top, removing quite some of the code that was also
supposed to prevent the cpuset update races, but doesn't work anymore
nowadays. This should fix the most pressing concerns with this topic
and give us a better baseline before either proceeding with the original
proposal, or pushing a change of semantics that removes the problem 1)
above. I'd be then fine with trying to change the semantic first and
rewrite later.
Patchset has been tested with the LTP cpuset01 stress test.
[1] https://lkml.kernel.org/r/4c44a589-5fd8-08d0-892c-e893bb525b71@suse.cz
[2] https://lwn.net/Articles/717797/
[3] https://marc.info/?l=linux-mm&m=149191957922828&w=2
This patch (of 6):
Commit e47483bca2cc ("mm, page_alloc: fix premature OOM when racing with
cpuset mems update") has fixed known recent regressions found by LTP's
cpuset01 testcase. I have however found that by modifying the testcase
to use per-vma mempolicies via bind(2) instead of per-task mempolicies
via set_mempolicy(2), the premature OOM still happens and the issue is
much older.
The root of the problem is that the cpuset's mems_allowed and
mempolicy's nodemask can temporarily have no intersection, thus
get_page_from_freelist() cannot find any usable zone. The current
semantic for empty intersection is to ignore mempolicy's nodemask and
honour cpuset restrictions. This is checked in node_zonelist(), but the
racy update can happen after we already passed the check. Such races
should be protected by the seqlock task->mems_allowed_seq, but it
doesn't work here, because 1) mpol_rebind_mm() does not happen under
seqlock for write, and doing so would lead to deadlock, as it takes
mmap_sem for write, while the allocation can have mmap_sem for read when
it's taking the seqlock for read. And 2) the seqlock cookie of callers
of node_zonelist() (alloc_pages_vma() and alloc_pages_current()) is
different than the one of __alloc_pages_slowpath(), so there's still a
potential race window.
This patch fixes the issue by having __alloc_pages_slowpath() check for
empty intersection of cpuset and ac->nodemask before OOM or allocation
failure. If it's indeed empty, the nodemask is ignored and allocation
retried, which mimics node_zonelist(). This works fine, because almost
all callers of __alloc_pages_nodemask are obtaining the nodemask via
node_zonelist(). The only exception is new_node_page() from hotplug,
where the potential violation of nodemask isn't an issue, as there's
already a fallback allocation attempt without any nodemask. If there's
a future caller that needs to have its specific nodemask honoured over
task's cpuset restrictions, we'll have to e.g. add a gfp flag for that.
Link: http://lkml.kernel.org/r/20170517081140.30654-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-06 22:39:56 +00:00
|
|
|
|
|
|
|
/* Deal with possible cpuset update races before we start OOM killing */
|
|
|
|
if (check_retry_cpuset(cpuset_mems_cookie, ac))
|
mm, page_alloc: fix premature OOM when racing with cpuset mems update
Ganapatrao Kulkarni reported that the LTP test cpuset01 in stress mode
triggers OOM killer in few seconds, despite lots of free memory. The
test attempts to repeatedly fault in memory in one process in a cpuset,
while changing allowed nodes of the cpuset between 0 and 1 in another
process.
The problem comes from insufficient protection against cpuset changes,
which can cause get_page_from_freelist() to consider all zones as
non-eligible due to nodemask and/or current->mems_allowed. This was
masked in the past by sufficient retries, but since commit 682a3385e773
("mm, page_alloc: inline the fast path of the zonelist iterator") we fix
the preferred_zoneref once, and don't iterate over the whole zonelist in
further attempts, thus the only eligible zones might be placed in the
zonelist before our starting point and we always miss them.
A previous patch fixed this problem for current->mems_allowed. However,
cpuset changes also update the task's mempolicy nodemask. The fix has
two parts. We have to repeat the preferred_zoneref search when we
detect cpuset update by way of seqcount, and we have to check the
seqcount before considering OOM.
[akpm@linux-foundation.org: fix typo in comment]
Link: http://lkml.kernel.org/r/20170120103843.24587-5-vbabka@suse.cz
Fixes: c33d6c06f60f ("mm, page_alloc: avoid looking up the first zone in a zonelist twice")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Ganapatrao Kulkarni <gpkulkarni@gmail.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-24 23:18:41 +00:00
|
|
|
goto retry_cpuset;
|
|
|
|
|
2015-06-24 23:57:21 +00:00
|
|
|
/* Reclaim has failed us, start killing things */
|
|
|
|
page = __alloc_pages_may_oom(gfp_mask, order, ac, &did_some_progress);
|
|
|
|
if (page)
|
|
|
|
goto got_pg;
|
|
|
|
|
2017-02-22 23:46:19 +00:00
|
|
|
/* Avoid allocations with no watermarks from looping endlessly */
|
2017-09-06 23:24:50 +00:00
|
|
|
if (tsk_is_oom_victim(current) &&
|
2020-08-07 06:26:04 +00:00
|
|
|
(alloc_flags & ALLOC_OOM ||
|
mm/page_alloc.c: make sure OOM victim can try allocations with no watermarks once
Roman Gushchin has reported that the OOM killer can trivially selects
next OOM victim when a thread doing memory allocation from page fault
path was selected as first OOM victim.
allocate invoked oom-killer: gfp_mask=0x14280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), nodemask=(null), order=0, oom_score_adj=0
allocate cpuset=/ mems_allowed=0
CPU: 1 PID: 492 Comm: allocate Not tainted 4.12.0-rc1-mm1+ #181
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
oom_kill_process+0x219/0x3e0
out_of_memory+0x11d/0x480
__alloc_pages_slowpath+0xc84/0xd40
__alloc_pages_nodemask+0x245/0x260
alloc_pages_vma+0xa2/0x270
__handle_mm_fault+0xca9/0x10c0
handle_mm_fault+0xf3/0x210
__do_page_fault+0x240/0x4e0
trace_do_page_fault+0x37/0xe0
do_async_page_fault+0x19/0x70
async_page_fault+0x28/0x30
...
Out of memory: Kill process 492 (allocate) score 899 or sacrifice child
Killed process 492 (allocate) total-vm:2052368kB, anon-rss:1894576kB, file-rss:4kB, shmem-rss:0kB
allocate: page allocation failure: order:0, mode:0x14280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), nodemask=(null)
allocate cpuset=/ mems_allowed=0
CPU: 1 PID: 492 Comm: allocate Not tainted 4.12.0-rc1-mm1+ #181
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
__alloc_pages_slowpath+0xd32/0xd40
__alloc_pages_nodemask+0x245/0x260
alloc_pages_vma+0xa2/0x270
__handle_mm_fault+0xca9/0x10c0
handle_mm_fault+0xf3/0x210
__do_page_fault+0x240/0x4e0
trace_do_page_fault+0x37/0xe0
do_async_page_fault+0x19/0x70
async_page_fault+0x28/0x30
...
oom_reaper: reaped process 492 (allocate), now anon-rss:0kB, file-rss:0kB, shmem-rss:0kB
...
allocate invoked oom-killer: gfp_mask=0x0(), nodemask=(null), order=0, oom_score_adj=0
allocate cpuset=/ mems_allowed=0
CPU: 1 PID: 492 Comm: allocate Not tainted 4.12.0-rc1-mm1+ #181
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
oom_kill_process+0x219/0x3e0
out_of_memory+0x11d/0x480
pagefault_out_of_memory+0x68/0x80
mm_fault_error+0x8f/0x190
? handle_mm_fault+0xf3/0x210
__do_page_fault+0x4b2/0x4e0
trace_do_page_fault+0x37/0xe0
do_async_page_fault+0x19/0x70
async_page_fault+0x28/0x30
...
Out of memory: Kill process 233 (firewalld) score 10 or sacrifice child
Killed process 233 (firewalld) total-vm:246076kB, anon-rss:20956kB, file-rss:0kB, shmem-rss:0kB
There is a race window that the OOM reaper completes reclaiming the
first victim's memory while nothing but mutex_trylock() prevents the
first victim from calling out_of_memory() from pagefault_out_of_memory()
after memory allocation for page fault path failed due to being selected
as an OOM victim.
This is a side effect of commit 9a67f6488eca926f ("mm: consolidate
GFP_NOFAIL checks in the allocator slowpath") because that commit
silently changed the behavior from
/* Avoid allocations with no watermarks from looping endlessly */
to
/*
* Give up allocations without trying memory reserves if selected
* as an OOM victim
*/
in __alloc_pages_slowpath() by moving the location to check TIF_MEMDIE
flag. I have noticed this change but I didn't post a patch because I
thought it is an acceptable change other than noise by warn_alloc()
because !__GFP_NOFAIL allocations are allowed to fail. But we
overlooked that failing memory allocation from page fault path makes
difference due to the race window explained above.
While it might be possible to add a check to pagefault_out_of_memory()
that prevents the first victim from calling out_of_memory() or remove
out_of_memory() from pagefault_out_of_memory(), changing
pagefault_out_of_memory() does not suppress noise by warn_alloc() when
allocating thread was selected as an OOM victim. There is little point
with printing similar backtraces and memory information from both
out_of_memory() and warn_alloc().
Instead, if we guarantee that current thread can try allocations with no
watermarks once when current thread looping inside
__alloc_pages_slowpath() was selected as an OOM victim, we can follow "who
can use memory reserves" rules and suppress noise by warn_alloc() and
prevent memory allocations from page fault path from calling
pagefault_out_of_memory().
If we take the comment literally, this patch would do
- if (test_thread_flag(TIF_MEMDIE))
- goto nopage;
+ if (alloc_flags == ALLOC_NO_WATERMARKS || (gfp_mask & __GFP_NOMEMALLOC))
+ goto nopage;
because gfp_pfmemalloc_allowed() returns false if __GFP_NOMEMALLOC is
given. But if I recall correctly (I couldn't find the message), the
condition is meant to apply to only OOM victims despite the comment.
Therefore, this patch preserves TIF_MEMDIE check.
Fixes: 9a67f6488eca926f ("mm: consolidate GFP_NOFAIL checks in the allocator slowpath")
Link: http://lkml.kernel.org/r/201705192112.IAF69238.OQOHSJLFOFFMtV@I-love.SAKURA.ne.jp
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reported-by: Roman Gushchin <guro@fb.com>
Tested-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: <stable@vger.kernel.org> [4.11]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-06-02 21:46:31 +00:00
|
|
|
(gfp_mask & __GFP_NOMEMALLOC)))
|
2017-02-22 23:46:19 +00:00
|
|
|
goto nopage;
|
|
|
|
|
2015-06-24 23:57:21 +00:00
|
|
|
/* Retry as long as the OOM killer is making progress */
|
2016-05-20 23:57:00 +00:00
|
|
|
if (did_some_progress) {
|
|
|
|
no_progress_loops = 0;
|
2015-06-24 23:57:21 +00:00
|
|
|
goto retry;
|
2016-05-20 23:57:00 +00:00
|
|
|
}
|
2015-06-24 23:57:21 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
nopage:
|
mm, page_alloc: fix more premature OOM due to race with cpuset update
I would like to stress that this patchset aims to fix issues and cleanup
the code *within the existing documented semantics*, i.e. patch 1
ignores mempolicy restrictions if the set of allowed nodes has no
intersection with set of nodes allowed by cpuset. I believe discussing
potential changes of the semantics can be better done once we have a
baseline with no known bugs of the current semantics.
I've recently summarized the cpuset/mempolicy issues in a LSF/MM
proposal [1] and the discussion itself [2]. I've been trying to rewrite
the handling as proposed, with the idea that changing semantics to make
all mempolicies static wrt cpuset updates (and discarding the relative
and default modes) can be tried on top, as there's a high risk of being
rejected/reverted because somebody might still care about the removed
modes.
However I haven't yet figured out how to properly:
1) make mempolicies swappable instead of rebinding in place. I thought
mbind() already works that way and uses refcounting to avoid
use-after-free of the old policy by a parallel allocation, but turns
out true refcounting is only done for shared (shmem) mempolicies, and
the actual protection for mbind() comes from mmap_sem. Extending the
refcounting means more overhead in allocator hot path. Also swapping
whole mempolicies means that we have to allocate the new ones, which
can fail, and reverting of the partially done work also means
allocating (note that mbind() doesn't care and will just leave part
of the range updated and part not updated when returning -ENOMEM...).
2) make cpuset's task->mems_allowed also swappable (after converting it
from nodemask to zonelist, which is the easy part) for mostly the
same reasons.
The good news is that while trying to do the above, I've at least
figured out how to hopefully close the remaining premature OOM's, and do
a buch of cleanups on top, removing quite some of the code that was also
supposed to prevent the cpuset update races, but doesn't work anymore
nowadays. This should fix the most pressing concerns with this topic
and give us a better baseline before either proceeding with the original
proposal, or pushing a change of semantics that removes the problem 1)
above. I'd be then fine with trying to change the semantic first and
rewrite later.
Patchset has been tested with the LTP cpuset01 stress test.
[1] https://lkml.kernel.org/r/4c44a589-5fd8-08d0-892c-e893bb525b71@suse.cz
[2] https://lwn.net/Articles/717797/
[3] https://marc.info/?l=linux-mm&m=149191957922828&w=2
This patch (of 6):
Commit e47483bca2cc ("mm, page_alloc: fix premature OOM when racing with
cpuset mems update") has fixed known recent regressions found by LTP's
cpuset01 testcase. I have however found that by modifying the testcase
to use per-vma mempolicies via bind(2) instead of per-task mempolicies
via set_mempolicy(2), the premature OOM still happens and the issue is
much older.
The root of the problem is that the cpuset's mems_allowed and
mempolicy's nodemask can temporarily have no intersection, thus
get_page_from_freelist() cannot find any usable zone. The current
semantic for empty intersection is to ignore mempolicy's nodemask and
honour cpuset restrictions. This is checked in node_zonelist(), but the
racy update can happen after we already passed the check. Such races
should be protected by the seqlock task->mems_allowed_seq, but it
doesn't work here, because 1) mpol_rebind_mm() does not happen under
seqlock for write, and doing so would lead to deadlock, as it takes
mmap_sem for write, while the allocation can have mmap_sem for read when
it's taking the seqlock for read. And 2) the seqlock cookie of callers
of node_zonelist() (alloc_pages_vma() and alloc_pages_current()) is
different than the one of __alloc_pages_slowpath(), so there's still a
potential race window.
This patch fixes the issue by having __alloc_pages_slowpath() check for
empty intersection of cpuset and ac->nodemask before OOM or allocation
failure. If it's indeed empty, the nodemask is ignored and allocation
retried, which mimics node_zonelist(). This works fine, because almost
all callers of __alloc_pages_nodemask are obtaining the nodemask via
node_zonelist(). The only exception is new_node_page() from hotplug,
where the potential violation of nodemask isn't an issue, as there's
already a fallback allocation attempt without any nodemask. If there's
a future caller that needs to have its specific nodemask honoured over
task's cpuset restrictions, we'll have to e.g. add a gfp flag for that.
Link: http://lkml.kernel.org/r/20170517081140.30654-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-06 22:39:56 +00:00
|
|
|
/* Deal with possible cpuset update races before we fail */
|
|
|
|
if (check_retry_cpuset(cpuset_mems_cookie, ac))
|
2017-01-24 23:18:38 +00:00
|
|
|
goto retry_cpuset;
|
|
|
|
|
2017-02-22 23:46:19 +00:00
|
|
|
/*
|
|
|
|
* Make sure that __GFP_NOFAIL request doesn't leak out and make sure
|
|
|
|
* we always retry
|
|
|
|
*/
|
|
|
|
if (gfp_mask & __GFP_NOFAIL) {
|
|
|
|
/*
|
|
|
|
* All existing users of the __GFP_NOFAIL are blockable, so warn
|
|
|
|
* of any new users that actually require GFP_NOWAIT
|
|
|
|
*/
|
|
|
|
if (WARN_ON_ONCE(!can_direct_reclaim))
|
|
|
|
goto fail;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* PF_MEMALLOC request from this context is rather bizarre
|
|
|
|
* because we cannot reclaim anything and only can loop waiting
|
|
|
|
* for somebody to do a work for us
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(current->flags & PF_MEMALLOC);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* non failing costly orders are a hard requirement which we
|
|
|
|
* are not prepared for much so let's warn about these users
|
|
|
|
* so that we can identify them and convert them to something
|
|
|
|
* else.
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(order > PAGE_ALLOC_COSTLY_ORDER);
|
|
|
|
|
2017-02-22 23:46:25 +00:00
|
|
|
/*
|
|
|
|
* Help non-failing allocations by giving them access to memory
|
|
|
|
* reserves but do not use ALLOC_NO_WATERMARKS because this
|
|
|
|
* could deplete whole memory reserves which would just make
|
|
|
|
* the situation worse
|
|
|
|
*/
|
|
|
|
page = __alloc_pages_cpuset_fallback(gfp_mask, order, ALLOC_HARDER, ac);
|
|
|
|
if (page)
|
|
|
|
goto got_pg;
|
|
|
|
|
2017-02-22 23:46:19 +00:00
|
|
|
cond_resched();
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
fail:
|
2017-02-22 23:46:10 +00:00
|
|
|
warn_alloc(gfp_mask, ac->nodemask,
|
2016-10-08 00:01:55 +00:00
|
|
|
"page allocation failure: order:%u", order);
|
2005-04-16 22:20:36 +00:00
|
|
|
got_pg:
|
mm: sl[au]b: add knowledge of PFMEMALLOC reserve pages
When a user or administrator requires swap for their application, they
create a swap partition and file, format it with mkswap and activate it
with swapon. Swap over the network is considered as an option in diskless
systems. The two likely scenarios are when blade servers are used as part
of a cluster where the form factor or maintenance costs do not allow the
use of disks and thin clients.
The Linux Terminal Server Project recommends the use of the Network Block
Device (NBD) for swap according to the manual at
https://sourceforge.net/projects/ltsp/files/Docs-Admin-Guide/LTSPManual.pdf/download
There is also documentation and tutorials on how to setup swap over NBD at
places like https://help.ubuntu.com/community/UbuntuLTSP/EnableNBDSWAP The
nbd-client also documents the use of NBD as swap. Despite this, the fact
is that a machine using NBD for swap can deadlock within minutes if swap
is used intensively. This patch series addresses the problem.
The core issue is that network block devices do not use mempools like
normal block devices do. As the host cannot control where they receive
packets from, they cannot reliably work out in advance how much memory
they might need. Some years ago, Peter Zijlstra developed a series of
patches that supported swap over an NFS that at least one distribution is
carrying within their kernels. This patch series borrows very heavily
from Peter's work to support swapping over NBD as a pre-requisite to
supporting swap-over-NFS. The bulk of the complexity is concerned with
preserving memory that is allocated from the PFMEMALLOC reserves for use
by the network layer which is needed for both NBD and NFS.
Patch 1 adds knowledge of the PFMEMALLOC reserves to SLAB and SLUB to
preserve access to pages allocated under low memory situations
to callers that are freeing memory.
Patch 2 optimises the SLUB fast path to avoid pfmemalloc checks
Patch 3 introduces __GFP_MEMALLOC to allow access to the PFMEMALLOC
reserves without setting PFMEMALLOC.
Patch 4 opens the possibility for softirqs to use PFMEMALLOC reserves
for later use by network packet processing.
Patch 5 only sets page->pfmemalloc when ALLOC_NO_WATERMARKS was required
Patch 6 ignores memory policies when ALLOC_NO_WATERMARKS is set.
Patches 7-12 allows network processing to use PFMEMALLOC reserves when
the socket has been marked as being used by the VM to clean pages. If
packets are received and stored in pages that were allocated under
low-memory situations and are unrelated to the VM, the packets
are dropped.
Patch 11 reintroduces __skb_alloc_page which the networking
folk may object to but is needed in some cases to propogate
pfmemalloc from a newly allocated page to an skb. If there is a
strong objection, this patch can be dropped with the impact being
that swap-over-network will be slower in some cases but it should
not fail.
Patch 13 is a micro-optimisation to avoid a function call in the
common case.
Patch 14 tags NBD sockets as being SOCK_MEMALLOC so they can use
PFMEMALLOC if necessary.
Patch 15 notes that it is still possible for the PFMEMALLOC reserve
to be depleted. To prevent this, direct reclaimers get throttled on
a waitqueue if 50% of the PFMEMALLOC reserves are depleted. It is
expected that kswapd and the direct reclaimers already running
will clean enough pages for the low watermark to be reached and
the throttled processes are woken up.
Patch 16 adds a statistic to track how often processes get throttled
Some basic performance testing was run using kernel builds, netperf on
loopback for UDP and TCP, hackbench (pipes and sockets), iozone and
sysbench. Each of them were expected to use the sl*b allocators
reasonably heavily but there did not appear to be significant performance
variances.
For testing swap-over-NBD, a machine was booted with 2G of RAM with a
swapfile backed by NBD. 8*NUM_CPU processes were started that create
anonymous memory mappings and read them linearly in a loop. The total
size of the mappings were 4*PHYSICAL_MEMORY to use swap heavily under
memory pressure.
Without the patches and using SLUB, the machine locks up within minutes
and runs to completion with them applied. With SLAB, the story is
different as an unpatched kernel run to completion. However, the patched
kernel completed the test 45% faster.
MICRO
3.5.0-rc2 3.5.0-rc2
vanilla swapnbd
Unrecognised test vmscan-anon-mmap-write
MMTests Statistics: duration
Sys Time Running Test (seconds) 197.80 173.07
User+Sys Time Running Test (seconds) 206.96 182.03
Total Elapsed Time (seconds) 3240.70 1762.09
This patch: mm: sl[au]b: add knowledge of PFMEMALLOC reserve pages
Allocations of pages below the min watermark run a risk of the machine
hanging due to a lack of memory. To prevent this, only callers who have
PF_MEMALLOC or TIF_MEMDIE set and are not processing an interrupt are
allowed to allocate with ALLOC_NO_WATERMARKS. Once they are allocated to
a slab though, nothing prevents other callers consuming free objects
within those slabs. This patch limits access to slab pages that were
alloced from the PFMEMALLOC reserves.
When this patch is applied, pages allocated from below the low watermark
are returned with page->pfmemalloc set and it is up to the caller to
determine how the page should be protected. SLAB restricts access to any
page with page->pfmemalloc set to callers which are known to able to
access the PFMEMALLOC reserve. If one is not available, an attempt is
made to allocate a new page rather than use a reserve. SLUB is a bit more
relaxed in that it only records if the current per-CPU page was allocated
from PFMEMALLOC reserve and uses another partial slab if the caller does
not have the necessary GFP or process flags. This was found to be
sufficient in tests to avoid hangs due to SLUB generally maintaining
smaller lists than SLAB.
In low-memory conditions it does mean that !PFMEMALLOC allocators can fail
a slab allocation even though free objects are available because they are
being preserved for callers that are freeing pages.
[a.p.zijlstra@chello.nl: Original implementation]
[sebastian@breakpoint.cc: Correct order of page flag clearing]
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: David Miller <davem@davemloft.net>
Cc: Neil Brown <neilb@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-07-31 23:43:58 +00:00
|
|
|
return page;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2009-06-16 22:31:57 +00:00
|
|
|
|
2017-02-24 22:56:29 +00:00
|
|
|
static inline bool prepare_alloc_pages(gfp_t gfp_mask, unsigned int order,
|
2017-07-06 22:40:03 +00:00
|
|
|
int preferred_nid, nodemask_t *nodemask,
|
2017-02-24 22:56:29 +00:00
|
|
|
struct alloc_context *ac, gfp_t *alloc_mask,
|
|
|
|
unsigned int *alloc_flags)
|
2009-06-16 22:31:57 +00:00
|
|
|
{
|
2020-06-03 22:59:01 +00:00
|
|
|
ac->highest_zoneidx = gfp_zone(gfp_mask);
|
2017-07-06 22:40:03 +00:00
|
|
|
ac->zonelist = node_zonelist(preferred_nid, gfp_mask);
|
2017-02-24 22:56:29 +00:00
|
|
|
ac->nodemask = nodemask;
|
2020-06-03 22:59:08 +00:00
|
|
|
ac->migratetype = gfp_migratetype(gfp_mask);
|
2009-06-16 22:31:57 +00:00
|
|
|
|
2016-05-20 00:13:30 +00:00
|
|
|
if (cpusets_enabled()) {
|
2017-02-24 22:56:29 +00:00
|
|
|
*alloc_mask |= __GFP_HARDWALL;
|
2020-08-07 06:26:01 +00:00
|
|
|
/*
|
|
|
|
* When we are in the interrupt context, it is irrelevant
|
|
|
|
* to the current task context. It means that any node ok.
|
|
|
|
*/
|
|
|
|
if (!in_interrupt() && !ac->nodemask)
|
2017-02-24 22:56:29 +00:00
|
|
|
ac->nodemask = &cpuset_current_mems_allowed;
|
2017-02-24 22:56:53 +00:00
|
|
|
else
|
|
|
|
*alloc_flags |= ALLOC_CPUSET;
|
2016-05-20 00:13:30 +00:00
|
|
|
}
|
|
|
|
|
2017-03-03 09:13:38 +00:00
|
|
|
fs_reclaim_acquire(gfp_mask);
|
|
|
|
fs_reclaim_release(gfp_mask);
|
2009-06-16 22:31:57 +00:00
|
|
|
|
2015-11-07 00:28:21 +00:00
|
|
|
might_sleep_if(gfp_mask & __GFP_DIRECT_RECLAIM);
|
2009-06-16 22:31:57 +00:00
|
|
|
|
|
|
|
if (should_fail_alloc_page(gfp_mask, order))
|
2017-02-24 22:56:29 +00:00
|
|
|
return false;
|
2009-06-16 22:31:57 +00:00
|
|
|
|
2020-08-07 06:26:04 +00:00
|
|
|
*alloc_flags = current_alloc_flags(gfp_mask, *alloc_flags);
|
2018-05-23 01:18:21 +00:00
|
|
|
|
2015-11-07 00:28:12 +00:00
|
|
|
/* Dirty zone balancing only done in the fast path */
|
2017-02-24 22:56:29 +00:00
|
|
|
ac->spread_dirty_pages = (gfp_mask & __GFP_WRITE);
|
2015-11-07 00:28:12 +00:00
|
|
|
|
2016-06-03 21:56:01 +00:00
|
|
|
/*
|
|
|
|
* The preferred zone is used for statistics but crucially it is
|
|
|
|
* also used as the starting point for the zonelist iterator. It
|
|
|
|
* may get reset for allocations that ignore memory policies.
|
|
|
|
*/
|
2017-02-24 22:56:29 +00:00
|
|
|
ac->preferred_zoneref = first_zones_zonelist(ac->zonelist,
|
2020-06-03 22:59:01 +00:00
|
|
|
ac->highest_zoneidx, ac->nodemask);
|
2020-10-13 23:55:51 +00:00
|
|
|
|
|
|
|
return true;
|
2017-02-24 22:56:29 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This is the 'heart' of the zoned buddy allocator.
|
|
|
|
*/
|
|
|
|
struct page *
|
2017-07-06 22:40:03 +00:00
|
|
|
__alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, int preferred_nid,
|
|
|
|
nodemask_t *nodemask)
|
2017-02-24 22:56:29 +00:00
|
|
|
{
|
|
|
|
struct page *page;
|
|
|
|
unsigned int alloc_flags = ALLOC_WMARK_LOW;
|
2017-09-08 23:13:22 +00:00
|
|
|
gfp_t alloc_mask; /* The gfp_t that was actually used for allocation */
|
2017-02-24 22:56:29 +00:00
|
|
|
struct alloc_context ac = { };
|
|
|
|
|
2018-11-16 23:08:53 +00:00
|
|
|
/*
|
|
|
|
* There are several places where we assume that the order value is sane
|
|
|
|
* so bail out early if the request is out of bound.
|
|
|
|
*/
|
|
|
|
if (unlikely(order >= MAX_ORDER)) {
|
|
|
|
WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN));
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2017-02-24 22:56:29 +00:00
|
|
|
gfp_mask &= gfp_allowed_mask;
|
2017-09-08 23:13:22 +00:00
|
|
|
alloc_mask = gfp_mask;
|
2017-07-06 22:40:03 +00:00
|
|
|
if (!prepare_alloc_pages(gfp_mask, order, preferred_nid, nodemask, &ac, &alloc_mask, &alloc_flags))
|
2017-02-24 22:56:29 +00:00
|
|
|
return NULL;
|
|
|
|
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
/*
|
|
|
|
* Forbid the first pass from falling back to types that fragment
|
|
|
|
* memory until all local zones are considered.
|
|
|
|
*/
|
2018-12-28 08:35:48 +00:00
|
|
|
alloc_flags |= alloc_flags_nofragment(ac.preferred_zoneref->zone, gfp_mask);
|
mm, page_alloc: spread allocations across zones before introducing fragmentation
Patch series "Fragmentation avoidance improvements", v5.
It has been noted before that fragmentation avoidance (aka
anti-fragmentation) is not perfect. Given sufficient time or an adverse
workload, memory gets fragmented and the long-term success of high-order
allocations degrades. This series defines an adverse workload, a definition
of external fragmentation events (including serious) ones and a series
that reduces the level of those fragmentation events.
The details of the workload and the consequences are described in more
detail in the changelogs. However, from patch 1, this is a high-level
summary of the adverse workload. The exact details are found in the
mmtests implementation.
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch)
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameterr create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed
3. Warm up a number of fio read-only threads accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll fault back in old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup
Overall the series reduces external fragmentation causing events by over 94%
on 1 and 2 socket machines, which in turn impacts high-order allocation
success rates over the long term. There are differences in latencies and
high-order allocation success rates. Latencies are a mixed bag as they
are vulnerable to exact system state and whether allocations succeeded
so they are treated as a secondary metric.
Patch 1 uses lower zones if they are populated and have free memory
instead of fragmenting a higher zone. It's special cased to
handle a Normal->DMA32 fallback with the reasons explained
in the changelog.
Patch 2-4 boosts watermarks temporarily when an external fragmentation
event occurs. kswapd wakes to reclaim a small amount of old memory
and then wakes kcompactd on completion to recover the system
slightly. This introduces some overhead in the slowpath. The level
of boosting can be tuned or disabled depending on the tolerance
for fragmentation vs allocation latency.
Patch 5 stalls some movable allocation requests to let kswapd from patch 4
make some progress. The duration of the stalls is very low but it
is possible to tune the system to avoid fragmentation events if
larger stalls can be tolerated.
The bulk of the improvement in fragmentation avoidance is from patches
1-4 but patch 5 can deal with a rare corner case and provides the option
of tuning a system for THP allocation success rates in exchange for
some stalls to control fragmentation.
This patch (of 5):
The page allocator zone lists are iterated based on the watermarks of each
zone which does not take anti-fragmentation into account. On x86, node 0
may have multiple zones while other nodes have one zone. A consequence is
that tasks running on node 0 may fragment ZONE_NORMAL even though
ZONE_DMA32 has plenty of free memory. This patch special cases the
allocator fast path such that it'll try an allocation from a lower local
zone before fragmenting a higher zone. In this case, stealing of
pageblocks or orders larger than a pageblock are still allowed in the fast
path as they are uninteresting from a fragmentation point of view.
This was evaluated using a benchmark designed to fragment memory before
attempting THP allocations. It's implemented in mmtests as the following
configurations
configs/config-global-dhp__workload_thpfioscale
configs/config-global-dhp__workload_thpfioscale-defrag
configs/config-global-dhp__workload_thpfioscale-madvhugepage
e.g. from mmtests
./run-mmtests.sh --run-monitor --config configs/config-global-dhp__workload_thpfioscale test-run-1
The broad details of the workload are as follows;
1. Create an XFS filesystem (not specified in the configuration but done
as part of the testing for this patch).
2. Start 4 fio threads that write a number of 64K files inefficiently.
Inefficiently means that files are created on first access and not
created in advance (fio parameter create_on_open=1) and fallocate
is not used (fallocate=none). With multiple IO issuers this creates
a mix of slab and page cache allocations over time. The total size
of the files is 150% physical memory so that the slabs and page cache
pages get mixed.
3. Warm up a number of fio read-only processes accessing the same files
created in step 2. This part runs for the same length of time it
took to create the files. It'll refault old data and further
interleave slab and page cache allocations. As it's now low on
memory due to step 2, fragmentation occurs as pageblocks get
stolen.
4. While step 3 is still running, start a process that tries to allocate
75% of memory as huge pages with a number of threads. The number of
threads is based on a (NR_CPUS_SOCKET - NR_FIO_THREADS)/4 to avoid THP
threads contending with fio, any other threads or forcing cross-NUMA
scheduling. Note that the test has not been used on a machine with less
than 8 cores. The benchmark records whether huge pages were allocated
and what the fault latency was in microseconds.
5. Measure the number of events potentially causing external fragmentation,
the fault latency and the huge page allocation success rate.
6. Cleanup the test files.
Note that due to the use of IO and page cache that this benchmark is not
suitable for running on large machines where the time to fragment memory
may be excessive. Also note that while this is one mix that generates
fragmentation that it's not the only mix that generates fragmentation.
Differences in workload that are more slab-intensive or whether SLUB is
used with high-order pages may yield different results.
When the page allocator fragments memory, it records the event using the
mm_page_alloc_extfrag ftrace event. If the fallback_order is smaller than
a pageblock order (order-9 on 64-bit x86) then it's considered to be an
"external fragmentation event" that may cause issues in the future.
Hence, the primary metric here is the number of external fragmentation
events that occur with order < 9. The secondary metric is allocation
latency and huge page allocation success rates but note that differences
in latencies and what the success rate also can affect the number of
external fragmentation event which is why it's a secondary metric.
1-socket Skylake machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 1 THP allocating thread
--------------------------------------
4.20-rc3 extfrag events < order 9: 804694
4.20-rc3+patch: 408912 (49% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 662.92 ( 0.00%) 653.58 * 1.41%*
Amean fault-huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 0.00 ( 0.00%) 0.00 ( 0.00%)
Fault latencies are slightly reduced while allocation success rates remain
at zero as this configuration does not make any special effort to allocate
THP and fio is heavily active at the time and either filling memory or
keeping pages resident. However, a 49% reduction of serious fragmentation
events reduces the changes of external fragmentation being a problem in
the future.
Vlastimil asked during review for a breakdown of the allocation types
that are falling back.
vanilla
3816 MIGRATE_UNMOVABLE
800845 MIGRATE_MOVABLE
33 MIGRATE_UNRECLAIMABLE
patch
735 MIGRATE_UNMOVABLE
408135 MIGRATE_MOVABLE
42 MIGRATE_UNRECLAIMABLE
The majority of the fallbacks are due to movable allocations and this is
consistent for the workload throughout the series so will not be presented
again as the primary source of fallbacks are movable allocations.
Movable fallbacks are sometimes considered "ok" to fallback because they
can be migrated. The problem is that they can fill an
unmovable/reclaimable pageblock causing those allocations to fallback
later and polluting pageblocks with pages that cannot move. If there is a
movable fallback, it is pretty much guaranteed to affect an
unmovable/reclaimable pageblock and while it might not be enough to
actually cause a unmovable/reclaimable fallback in the future, we cannot
know that in advance so the patch takes the only option available to it.
Hence, it's important to control them. This point is also consistent
throughout the series and will not be repeated.
1-socket Skylake machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 291392
4.20-rc3+patch: 191187 (34% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-1 1495.14 ( 0.00%) 1467.55 ( 1.85%)
Amean fault-huge-1 1098.48 ( 0.00%) 1127.11 ( -2.61%)
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-1 78.57 ( 0.00%) 77.64 ( -1.18%)
Fragmentation events were reduced quite a bit although this is known
to be a little variable. The latencies and allocation success rates
are similar but they were already quite high.
2-socket Haswell machine
config-global-dhp__workload_thpfioscale XFS (no special madvise)
4 fio threads, 5 THP allocating threads
----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 215698
4.20-rc3+patch: 200210 (7% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 1350.05 ( 0.00%) 1346.45 ( 0.27%)
Amean fault-huge-5 4181.01 ( 0.00%) 3418.60 ( 18.24%)
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 1.15 ( 0.00%) 0.78 ( -31.88%)
The reduction of external fragmentation events is slight and this is
partially due to the removal of __GFP_THISNODE in commit ac5b2c18911f
("mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings") as THP
allocations can now spill over to remote nodes instead of fragmenting
local memory.
2-socket Haswell machine
global-dhp__workload_thpfioscale-madvhugepage-xfs (MADV_HUGEPAGE)
-----------------------------------------------------------------
4.20-rc3 extfrag events < order 9: 166352
4.20-rc3+patch: 147463 (11% reduction)
thpfioscale Fault Latencies
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Amean fault-base-5 6138.97 ( 0.00%) 6217.43 ( -1.28%)
Amean fault-huge-5 2294.28 ( 0.00%) 3163.33 * -37.88%*
thpfioscale Percentage Faults Huge
4.20.0-rc3 4.20.0-rc3
vanilla lowzone-v5r8
Percentage huge-5 96.82 ( 0.00%) 95.14 ( -1.74%)
There was a slight reduction in external fragmentation events although the
latencies were higher. The allocation success rate is high enough that
the system is struggling and there is quite a lot of parallel reclaim and
compaction activity. There is also a certain degree of luck on whether
processes start on node 0 or not for this patch but the relevance is
reduced later in the series.
Overall, the patch reduces the number of external fragmentation causing
events so the success of THP over long periods of time would be improved
for this adverse workload.
Link: http://lkml.kernel.org/r/20181123114528.28802-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:35:41 +00:00
|
|
|
|
2009-06-16 22:31:59 +00:00
|
|
|
/* First allocation attempt */
|
2015-02-11 23:25:41 +00:00
|
|
|
page = get_page_from_freelist(alloc_mask, order, alloc_flags, &ac);
|
2016-05-20 00:14:01 +00:00
|
|
|
if (likely(page))
|
|
|
|
goto out;
|
2009-06-16 22:31:57 +00:00
|
|
|
|
2016-05-20 00:14:01 +00:00
|
|
|
/*
|
mm: introduce memalloc_nofs_{save,restore} API
GFP_NOFS context is used for the following 5 reasons currently:
- to prevent from deadlocks when the lock held by the allocation
context would be needed during the memory reclaim
- to prevent from stack overflows during the reclaim because the
allocation is performed from a deep context already
- to prevent lockups when the allocation context depends on other
reclaimers to make a forward progress indirectly
- just in case because this would be safe from the fs POV
- silence lockdep false positives
Unfortunately overuse of this allocation context brings some problems to
the MM. Memory reclaim is much weaker (especially during heavy FS
metadata workloads), OOM killer cannot be invoked because the MM layer
doesn't have enough information about how much memory is freeable by the
FS layer.
In many cases it is far from clear why the weaker context is even used
and so it might be used unnecessarily. We would like to get rid of
those as much as possible. One way to do that is to use the flag in
scopes rather than isolated cases. Such a scope is declared when really
necessary, tracked per task and all the allocation requests from within
the context will simply inherit the GFP_NOFS semantic.
Not only this is easier to understand and maintain because there are
much less problematic contexts than specific allocation requests, this
also helps code paths where FS layer interacts with other layers (e.g.
crypto, security modules, MM etc...) and there is no easy way to convey
the allocation context between the layers.
Introduce memalloc_nofs_{save,restore} API to control the scope of
GFP_NOFS allocation context. This is basically copying
memalloc_noio_{save,restore} API we have for other restricted allocation
context GFP_NOIO. The PF_MEMALLOC_NOFS flag already exists and it is
just an alias for PF_FSTRANS which has been xfs specific until recently.
There are no more PF_FSTRANS users anymore so let's just drop it.
PF_MEMALLOC_NOFS is now checked in the MM layer and drops __GFP_FS
implicitly same as PF_MEMALLOC_NOIO drops __GFP_IO. memalloc_noio_flags
is renamed to current_gfp_context because it now cares about both
PF_MEMALLOC_NOFS and PF_MEMALLOC_NOIO contexts. Xfs code paths preserve
their semantic. kmem_flags_convert() doesn't need to evaluate the flag
anymore.
This patch shouldn't introduce any functional changes.
Let's hope that filesystems will drop direct GFP_NOFS (resp. ~__GFP_FS)
usage as much as possible and only use a properly documented
memalloc_nofs_{save,restore} checkpoints where they are appropriate.
[akpm@linux-foundation.org: fix comment typo, reflow comment]
Link: http://lkml.kernel.org/r/20170306131408.9828-5-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-03 21:53:15 +00:00
|
|
|
* Apply scoped allocation constraints. This is mainly about GFP_NOFS
|
|
|
|
* resp. GFP_NOIO which has to be inherited for all allocation requests
|
|
|
|
* from a particular context which has been marked by
|
|
|
|
* memalloc_no{fs,io}_{save,restore}.
|
2016-05-20 00:14:01 +00:00
|
|
|
*/
|
mm: introduce memalloc_nofs_{save,restore} API
GFP_NOFS context is used for the following 5 reasons currently:
- to prevent from deadlocks when the lock held by the allocation
context would be needed during the memory reclaim
- to prevent from stack overflows during the reclaim because the
allocation is performed from a deep context already
- to prevent lockups when the allocation context depends on other
reclaimers to make a forward progress indirectly
- just in case because this would be safe from the fs POV
- silence lockdep false positives
Unfortunately overuse of this allocation context brings some problems to
the MM. Memory reclaim is much weaker (especially during heavy FS
metadata workloads), OOM killer cannot be invoked because the MM layer
doesn't have enough information about how much memory is freeable by the
FS layer.
In many cases it is far from clear why the weaker context is even used
and so it might be used unnecessarily. We would like to get rid of
those as much as possible. One way to do that is to use the flag in
scopes rather than isolated cases. Such a scope is declared when really
necessary, tracked per task and all the allocation requests from within
the context will simply inherit the GFP_NOFS semantic.
Not only this is easier to understand and maintain because there are
much less problematic contexts than specific allocation requests, this
also helps code paths where FS layer interacts with other layers (e.g.
crypto, security modules, MM etc...) and there is no easy way to convey
the allocation context between the layers.
Introduce memalloc_nofs_{save,restore} API to control the scope of
GFP_NOFS allocation context. This is basically copying
memalloc_noio_{save,restore} API we have for other restricted allocation
context GFP_NOIO. The PF_MEMALLOC_NOFS flag already exists and it is
just an alias for PF_FSTRANS which has been xfs specific until recently.
There are no more PF_FSTRANS users anymore so let's just drop it.
PF_MEMALLOC_NOFS is now checked in the MM layer and drops __GFP_FS
implicitly same as PF_MEMALLOC_NOIO drops __GFP_IO. memalloc_noio_flags
is renamed to current_gfp_context because it now cares about both
PF_MEMALLOC_NOFS and PF_MEMALLOC_NOIO contexts. Xfs code paths preserve
their semantic. kmem_flags_convert() doesn't need to evaluate the flag
anymore.
This patch shouldn't introduce any functional changes.
Let's hope that filesystems will drop direct GFP_NOFS (resp. ~__GFP_FS)
usage as much as possible and only use a properly documented
memalloc_nofs_{save,restore} checkpoints where they are appropriate.
[akpm@linux-foundation.org: fix comment typo, reflow comment]
Link: http://lkml.kernel.org/r/20170306131408.9828-5-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-03 21:53:15 +00:00
|
|
|
alloc_mask = current_gfp_context(gfp_mask);
|
2016-05-20 00:14:01 +00:00
|
|
|
ac.spread_dirty_pages = false;
|
2015-02-11 23:25:07 +00:00
|
|
|
|
2016-05-20 00:14:44 +00:00
|
|
|
/*
|
|
|
|
* Restore the original nodemask if it was potentially replaced with
|
|
|
|
* &cpuset_current_mems_allowed to optimize the fast-path attempt.
|
|
|
|
*/
|
2020-04-02 04:09:53 +00:00
|
|
|
ac.nodemask = nodemask;
|
mm, page_alloc: fix fast-path race with cpuset update or removal
Ganapatrao Kulkarni reported that the LTP test cpuset01 in stress mode
triggers OOM killer in few seconds, despite lots of free memory. The
test attempts to repeatedly fault in memory in one process in a cpuset,
while changing allowed nodes of the cpuset between 0 and 1 in another
process.
One possible cause is that in the fast path we find the preferred
zoneref according to current mems_allowed, so that it points to the
middle of the zonelist, skipping e.g. zones of node 1 completely. If
the mems_allowed is updated to contain only node 1, we never reach it in
the zonelist, and trigger OOM before checking the cpuset_mems_cookie.
This patch fixes the particular case by redoing the preferred zoneref
search if we switch back to the original nodemask. The condition is
also slightly changed so that when the last non-root cpuset is removed,
we don't miss it.
Note that this is not a full fix, and more patches will follow.
Link: http://lkml.kernel.org/r/20170120103843.24587-3-vbabka@suse.cz
Fixes: 682a3385e773 ("mm, page_alloc: inline the fast path of the zonelist iterator")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Ganapatrao Kulkarni <gpkulkarni@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-24 23:18:35 +00:00
|
|
|
|
2016-05-20 00:14:01 +00:00
|
|
|
page = __alloc_pages_slowpath(alloc_mask, order, &ac);
|
cpuset: mm: reduce large amounts of memory barrier related damage v3
Commit c0ff7453bb5c ("cpuset,mm: fix no node to alloc memory when
changing cpuset's mems") wins a super prize for the largest number of
memory barriers entered into fast paths for one commit.
[get|put]_mems_allowed is incredibly heavy with pairs of full memory
barriers inserted into a number of hot paths. This was detected while
investigating at large page allocator slowdown introduced some time
after 2.6.32. The largest portion of this overhead was shown by
oprofile to be at an mfence introduced by this commit into the page
allocator hot path.
For extra style points, the commit introduced the use of yield() in an
implementation of what looks like a spinning mutex.
This patch replaces the full memory barriers on both read and write
sides with a sequence counter with just read barriers on the fast path
side. This is much cheaper on some architectures, including x86. The
main bulk of the patch is the retry logic if the nodemask changes in a
manner that can cause a false failure.
While updating the nodemask, a check is made to see if a false failure
is a risk. If it is, the sequence number gets bumped and parallel
allocators will briefly stall while the nodemask update takes place.
In a page fault test microbenchmark, oprofile samples from
__alloc_pages_nodemask went from 4.53% of all samples to 1.15%. The
actual results were
3.3.0-rc3 3.3.0-rc3
rc3-vanilla nobarrier-v2r1
Clients 1 UserTime 0.07 ( 0.00%) 0.08 (-14.19%)
Clients 2 UserTime 0.07 ( 0.00%) 0.07 ( 2.72%)
Clients 4 UserTime 0.08 ( 0.00%) 0.07 ( 3.29%)
Clients 1 SysTime 0.70 ( 0.00%) 0.65 ( 6.65%)
Clients 2 SysTime 0.85 ( 0.00%) 0.82 ( 3.65%)
Clients 4 SysTime 1.41 ( 0.00%) 1.41 ( 0.32%)
Clients 1 WallTime 0.77 ( 0.00%) 0.74 ( 4.19%)
Clients 2 WallTime 0.47 ( 0.00%) 0.45 ( 3.73%)
Clients 4 WallTime 0.38 ( 0.00%) 0.37 ( 1.58%)
Clients 1 Flt/sec/cpu 497620.28 ( 0.00%) 520294.53 ( 4.56%)
Clients 2 Flt/sec/cpu 414639.05 ( 0.00%) 429882.01 ( 3.68%)
Clients 4 Flt/sec/cpu 257959.16 ( 0.00%) 258761.48 ( 0.31%)
Clients 1 Flt/sec 495161.39 ( 0.00%) 517292.87 ( 4.47%)
Clients 2 Flt/sec 820325.95 ( 0.00%) 850289.77 ( 3.65%)
Clients 4 Flt/sec 1020068.93 ( 0.00%) 1022674.06 ( 0.26%)
MMTests Statistics: duration
Sys Time Running Test (seconds) 135.68 132.17
User+Sys Time Running Test (seconds) 164.2 160.13
Total Elapsed Time (seconds) 123.46 120.87
The overall improvement is small but the System CPU time is much
improved and roughly in correlation to what oprofile reported (these
performance figures are without profiling so skew is expected). The
actual number of page faults is noticeably improved.
For benchmarks like kernel builds, the overall benefit is marginal but
the system CPU time is slightly reduced.
To test the actual bug the commit fixed I opened two terminals. The
first ran within a cpuset and continually ran a small program that
faulted 100M of anonymous data. In a second window, the nodemask of the
cpuset was continually randomised in a loop.
Without the commit, the program would fail every so often (usually
within 10 seconds) and obviously with the commit everything worked fine.
With this patch applied, it also worked fine so the fix should be
functionally equivalent.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-03-21 23:34:11 +00:00
|
|
|
|
2016-05-20 00:14:01 +00:00
|
|
|
out:
|
mm: memcontrol: only mark charged pages with PageKmemcg
To distinguish non-slab pages charged to kmemcg we mark them PageKmemcg,
which sets page->_mapcount to -512. Currently, we set/clear PageKmemcg
in __alloc_pages_nodemask()/free_pages_prepare() for any page allocated
with __GFP_ACCOUNT, including those that aren't actually charged to any
cgroup, i.e. allocated from the root cgroup context. To avoid overhead
in case cgroups are not used, we only do that if memcg_kmem_enabled() is
true. The latter is set iff there are kmem-enabled memory cgroups
(online or offline). The root cgroup is not considered kmem-enabled.
As a result, if a page is allocated with __GFP_ACCOUNT for the root
cgroup when there are kmem-enabled memory cgroups and is freed after all
kmem-enabled memory cgroups were removed, e.g.
# no memory cgroups has been created yet, create one
mkdir /sys/fs/cgroup/memory/test
# run something allocating pages with __GFP_ACCOUNT, e.g.
# a program using pipe
dmesg | tail
# remove the memory cgroup
rmdir /sys/fs/cgroup/memory/test
we'll get bad page state bug complaining about page->_mapcount != -1:
BUG: Bad page state in process swapper/0 pfn:1fd945c
page:ffffea007f651700 count:0 mapcount:-511 mapping: (null) index:0x0
flags: 0x1000000000000000()
To avoid that, let's mark with PageKmemcg only those pages that are
actually charged to and hence pin a non-root memory cgroup.
Fixes: 4949148ad433 ("mm: charge/uncharge kmemcg from generic page allocator paths")
Reported-and-tested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-08 20:03:12 +00:00
|
|
|
if (memcg_kmem_enabled() && (gfp_mask & __GFP_ACCOUNT) && page &&
|
2020-04-02 04:06:46 +00:00
|
|
|
unlikely(__memcg_kmem_charge_page(page, gfp_mask, order) != 0)) {
|
mm: memcontrol: only mark charged pages with PageKmemcg
To distinguish non-slab pages charged to kmemcg we mark them PageKmemcg,
which sets page->_mapcount to -512. Currently, we set/clear PageKmemcg
in __alloc_pages_nodemask()/free_pages_prepare() for any page allocated
with __GFP_ACCOUNT, including those that aren't actually charged to any
cgroup, i.e. allocated from the root cgroup context. To avoid overhead
in case cgroups are not used, we only do that if memcg_kmem_enabled() is
true. The latter is set iff there are kmem-enabled memory cgroups
(online or offline). The root cgroup is not considered kmem-enabled.
As a result, if a page is allocated with __GFP_ACCOUNT for the root
cgroup when there are kmem-enabled memory cgroups and is freed after all
kmem-enabled memory cgroups were removed, e.g.
# no memory cgroups has been created yet, create one
mkdir /sys/fs/cgroup/memory/test
# run something allocating pages with __GFP_ACCOUNT, e.g.
# a program using pipe
dmesg | tail
# remove the memory cgroup
rmdir /sys/fs/cgroup/memory/test
we'll get bad page state bug complaining about page->_mapcount != -1:
BUG: Bad page state in process swapper/0 pfn:1fd945c
page:ffffea007f651700 count:0 mapcount:-511 mapping: (null) index:0x0
flags: 0x1000000000000000()
To avoid that, let's mark with PageKmemcg only those pages that are
actually charged to and hence pin a non-root memory cgroup.
Fixes: 4949148ad433 ("mm: charge/uncharge kmemcg from generic page allocator paths")
Reported-and-tested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-08 20:03:12 +00:00
|
|
|
__free_pages(page, order);
|
|
|
|
page = NULL;
|
mm: charge/uncharge kmemcg from generic page allocator paths
Currently, to charge a non-slab allocation to kmemcg one has to use
alloc_kmem_pages helper with __GFP_ACCOUNT flag. A page allocated with
this helper should finally be freed using free_kmem_pages, otherwise it
won't be uncharged.
This API suits its current users fine, but it turns out to be impossible
to use along with page reference counting, i.e. when an allocation is
supposed to be freed with put_page, as it is the case with pipe or unix
socket buffers.
To overcome this limitation, this patch moves charging/uncharging to
generic page allocator paths, i.e. to __alloc_pages_nodemask and
free_pages_prepare, and zaps alloc/free_kmem_pages helpers. This way,
one can use any of the available page allocation functions to get the
allocated page charged to kmemcg - it's enough to pass __GFP_ACCOUNT,
just like in case of kmalloc and friends. A charged page will be
automatically uncharged on free.
To make it possible, we need to mark pages charged to kmemcg somehow.
To avoid introducing a new page flag, we make use of page->_mapcount for
marking such pages. Since pages charged to kmemcg are not supposed to
be mapped to userspace, it should work just fine. There are other
(ab)users of page->_mapcount - buddy and balloon pages - but we don't
conflict with them.
In case kmemcg is compiled out or not used at runtime, this patch
introduces no overhead to generic page allocator paths. If kmemcg is
used, it will be plus one gfp flags check on alloc and plus one
page->_mapcount check on free, which shouldn't hurt performance, because
the data accessed are hot.
Link: http://lkml.kernel.org/r/a9736d856f895bcb465d9f257b54efe32eda6f99.1464079538.git.vdavydov@virtuozzo.com
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:24:24 +00:00
|
|
|
}
|
|
|
|
|
2016-05-20 00:14:01 +00:00
|
|
|
trace_mm_page_alloc(page, order, alloc_mask, ac.migratetype);
|
|
|
|
|
2009-06-16 22:31:57 +00:00
|
|
|
return page;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2009-06-16 22:31:52 +00:00
|
|
|
EXPORT_SYMBOL(__alloc_pages_nodemask);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/*
|
2018-08-17 22:46:01 +00:00
|
|
|
* Common helper functions. Never use with __GFP_HIGHMEM because the returned
|
|
|
|
* address cannot represent highmem pages. Use alloc_pages and then kmap if
|
|
|
|
* you need to access high mem.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2008-02-05 06:29:26 +00:00
|
|
|
unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2009-09-22 00:01:47 +00:00
|
|
|
struct page *page;
|
|
|
|
|
2018-08-17 22:46:01 +00:00
|
|
|
page = alloc_pages(gfp_mask & ~__GFP_HIGHMEM, order);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (!page)
|
|
|
|
return 0;
|
|
|
|
return (unsigned long) page_address(page);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(__get_free_pages);
|
|
|
|
|
2008-02-05 06:29:26 +00:00
|
|
|
unsigned long get_zeroed_page(gfp_t gfp_mask)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2009-09-22 00:01:47 +00:00
|
|
|
return __get_free_pages(gfp_mask | __GFP_ZERO, 0);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(get_zeroed_page);
|
|
|
|
|
2018-12-28 08:35:22 +00:00
|
|
|
static inline void free_the_page(struct page *page, unsigned int order)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2018-12-28 08:35:22 +00:00
|
|
|
if (order == 0) /* Via pcp? */
|
|
|
|
free_unref_page(page);
|
|
|
|
else
|
2020-10-16 03:09:35 +00:00
|
|
|
__free_pages_ok(page, order, FPI_NONE);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2020-12-15 03:11:09 +00:00
|
|
|
/**
|
|
|
|
* __free_pages - Free pages allocated with alloc_pages().
|
|
|
|
* @page: The page pointer returned from alloc_pages().
|
|
|
|
* @order: The order of the allocation.
|
|
|
|
*
|
|
|
|
* This function can free multi-page allocations that are not compound
|
|
|
|
* pages. It does not check that the @order passed in matches that of
|
|
|
|
* the allocation, so it is easy to leak memory. Freeing more memory
|
|
|
|
* than was allocated will probably emit a warning.
|
|
|
|
*
|
|
|
|
* If the last reference to this page is speculative, it will be released
|
|
|
|
* by put_page() which only frees the first page of a non-compound
|
|
|
|
* allocation. To prevent the remaining pages from being leaked, we free
|
|
|
|
* the subsequent pages here. If you want to use the page's reference
|
|
|
|
* count to decide when to free the allocation, you should allocate a
|
|
|
|
* compound page, and use put_page() instead of __free_pages().
|
|
|
|
*
|
|
|
|
* Context: May be called in interrupt context or while holding a normal
|
|
|
|
* spinlock, but not in NMI context or while holding a raw spinlock.
|
|
|
|
*/
|
2018-12-28 08:35:22 +00:00
|
|
|
void __free_pages(struct page *page, unsigned int order)
|
|
|
|
{
|
|
|
|
if (put_page_testzero(page))
|
|
|
|
free_the_page(page, order);
|
2020-10-13 23:56:04 +00:00
|
|
|
else if (!PageHead(page))
|
|
|
|
while (order-- > 0)
|
|
|
|
free_the_page(page + (1 << order), order);
|
2018-12-28 08:35:22 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
EXPORT_SYMBOL(__free_pages);
|
|
|
|
|
2008-02-05 06:29:26 +00:00
|
|
|
void free_pages(unsigned long addr, unsigned int order)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
if (addr != 0) {
|
2006-09-26 06:30:55 +00:00
|
|
|
VM_BUG_ON(!virt_addr_valid((void *)addr));
|
2005-04-16 22:20:36 +00:00
|
|
|
__free_pages(virt_to_page((void *)addr), order);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
EXPORT_SYMBOL(free_pages);
|
|
|
|
|
2015-05-07 04:11:57 +00:00
|
|
|
/*
|
|
|
|
* Page Fragment:
|
|
|
|
* An arbitrary-length arbitrary-offset area of memory which resides
|
|
|
|
* within a 0 or higher order page. Multiple fragments within that page
|
|
|
|
* are individually refcounted, in the page's reference counter.
|
|
|
|
*
|
|
|
|
* The page_frag functions below provide a simple allocation framework for
|
|
|
|
* page fragments. This is used by the network stack and network device
|
|
|
|
* drivers to provide a backing region of memory for use as either an
|
|
|
|
* sk_buff->head, or to be used in the "frags" portion of skb_shared_info.
|
|
|
|
*/
|
2017-01-11 00:58:09 +00:00
|
|
|
static struct page *__page_frag_cache_refill(struct page_frag_cache *nc,
|
|
|
|
gfp_t gfp_mask)
|
2015-05-07 04:11:57 +00:00
|
|
|
{
|
|
|
|
struct page *page = NULL;
|
|
|
|
gfp_t gfp = gfp_mask;
|
|
|
|
|
|
|
|
#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
|
|
|
|
gfp_mask |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY |
|
|
|
|
__GFP_NOMEMALLOC;
|
|
|
|
page = alloc_pages_node(NUMA_NO_NODE, gfp_mask,
|
|
|
|
PAGE_FRAG_CACHE_MAX_ORDER);
|
|
|
|
nc->size = page ? PAGE_FRAG_CACHE_MAX_SIZE : PAGE_SIZE;
|
|
|
|
#endif
|
|
|
|
if (unlikely(!page))
|
|
|
|
page = alloc_pages_node(NUMA_NO_NODE, gfp, 0);
|
|
|
|
|
|
|
|
nc->va = page ? page_address(page) : NULL;
|
|
|
|
|
|
|
|
return page;
|
|
|
|
}
|
|
|
|
|
2017-01-11 00:58:09 +00:00
|
|
|
void __page_frag_cache_drain(struct page *page, unsigned int count)
|
2016-12-14 23:05:26 +00:00
|
|
|
{
|
|
|
|
VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
|
|
|
|
|
2018-12-28 08:35:22 +00:00
|
|
|
if (page_ref_sub_and_test(page, count))
|
|
|
|
free_the_page(page, compound_order(page));
|
2016-12-14 23:05:26 +00:00
|
|
|
}
|
2017-01-11 00:58:09 +00:00
|
|
|
EXPORT_SYMBOL(__page_frag_cache_drain);
|
2016-12-14 23:05:26 +00:00
|
|
|
|
2021-02-04 10:56:35 +00:00
|
|
|
void *page_frag_alloc_align(struct page_frag_cache *nc,
|
|
|
|
unsigned int fragsz, gfp_t gfp_mask,
|
|
|
|
unsigned int align_mask)
|
2015-05-07 04:11:57 +00:00
|
|
|
{
|
|
|
|
unsigned int size = PAGE_SIZE;
|
|
|
|
struct page *page;
|
|
|
|
int offset;
|
|
|
|
|
|
|
|
if (unlikely(!nc->va)) {
|
|
|
|
refill:
|
2017-01-11 00:58:09 +00:00
|
|
|
page = __page_frag_cache_refill(nc, gfp_mask);
|
2015-05-07 04:11:57 +00:00
|
|
|
if (!page)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
|
|
|
|
/* if size can vary use size else just use PAGE_SIZE */
|
|
|
|
size = nc->size;
|
|
|
|
#endif
|
|
|
|
/* Even if we own the page, we do not use atomic_set().
|
|
|
|
* This would break get_page_unless_zero() users.
|
|
|
|
*/
|
2019-02-15 22:44:12 +00:00
|
|
|
page_ref_add(page, PAGE_FRAG_CACHE_MAX_SIZE);
|
2015-05-07 04:11:57 +00:00
|
|
|
|
|
|
|
/* reset page count bias and offset to start of new frag */
|
2015-08-21 21:11:51 +00:00
|
|
|
nc->pfmemalloc = page_is_pfmemalloc(page);
|
2019-02-15 22:44:12 +00:00
|
|
|
nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
|
2015-05-07 04:11:57 +00:00
|
|
|
nc->offset = size;
|
|
|
|
}
|
|
|
|
|
|
|
|
offset = nc->offset - fragsz;
|
|
|
|
if (unlikely(offset < 0)) {
|
|
|
|
page = virt_to_page(nc->va);
|
|
|
|
|
2016-03-17 21:19:26 +00:00
|
|
|
if (!page_ref_sub_and_test(page, nc->pagecnt_bias))
|
2015-05-07 04:11:57 +00:00
|
|
|
goto refill;
|
|
|
|
|
2020-11-15 20:10:29 +00:00
|
|
|
if (unlikely(nc->pfmemalloc)) {
|
|
|
|
free_the_page(page, compound_order(page));
|
|
|
|
goto refill;
|
|
|
|
}
|
|
|
|
|
2015-05-07 04:11:57 +00:00
|
|
|
#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
|
|
|
|
/* if size can vary use size else just use PAGE_SIZE */
|
|
|
|
size = nc->size;
|
|
|
|
#endif
|
|
|
|
/* OK, page count is 0, we can safely set it */
|
2019-02-15 22:44:12 +00:00
|
|
|
set_page_count(page, PAGE_FRAG_CACHE_MAX_SIZE + 1);
|
2015-05-07 04:11:57 +00:00
|
|
|
|
|
|
|
/* reset page count bias and offset to start of new frag */
|
2019-02-15 22:44:12 +00:00
|
|
|
nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
|
2015-05-07 04:11:57 +00:00
|
|
|
offset = size - fragsz;
|
|
|
|
}
|
|
|
|
|
|
|
|
nc->pagecnt_bias--;
|
2021-02-04 10:56:35 +00:00
|
|
|
offset &= align_mask;
|
2015-05-07 04:11:57 +00:00
|
|
|
nc->offset = offset;
|
|
|
|
|
|
|
|
return nc->va + offset;
|
|
|
|
}
|
2021-02-04 10:56:35 +00:00
|
|
|
EXPORT_SYMBOL(page_frag_alloc_align);
|
2015-05-07 04:11:57 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Frees a page fragment allocated out of either a compound or order 0 page.
|
|
|
|
*/
|
2017-01-11 00:58:06 +00:00
|
|
|
void page_frag_free(void *addr)
|
2015-05-07 04:11:57 +00:00
|
|
|
{
|
|
|
|
struct page *page = virt_to_head_page(addr);
|
|
|
|
|
2018-12-28 08:35:22 +00:00
|
|
|
if (unlikely(put_page_testzero(page)))
|
|
|
|
free_the_page(page, compound_order(page));
|
2015-05-07 04:11:57 +00:00
|
|
|
}
|
2017-01-11 00:58:06 +00:00
|
|
|
EXPORT_SYMBOL(page_frag_free);
|
2015-05-07 04:11:57 +00:00
|
|
|
|
2015-11-07 00:29:57 +00:00
|
|
|
static void *make_alloc_exact(unsigned long addr, unsigned int order,
|
|
|
|
size_t size)
|
2011-05-11 22:13:34 +00:00
|
|
|
{
|
|
|
|
if (addr) {
|
|
|
|
unsigned long alloc_end = addr + (PAGE_SIZE << order);
|
|
|
|
unsigned long used = addr + PAGE_ALIGN(size);
|
|
|
|
|
|
|
|
split_page(virt_to_page((void *)addr), order);
|
|
|
|
while (used < alloc_end) {
|
|
|
|
free_page(used);
|
|
|
|
used += PAGE_SIZE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return (void *)addr;
|
|
|
|
}
|
|
|
|
|
2008-07-24 04:28:11 +00:00
|
|
|
/**
|
|
|
|
* alloc_pages_exact - allocate an exact number physically-contiguous pages.
|
|
|
|
* @size: the number of bytes to allocate
|
mm, page_alloc: disallow __GFP_COMP in alloc_pages_exact()
alloc_pages_exact*() allocates a page of sufficient order and then splits
it to return only the number of pages requested. That makes it
incompatible with __GFP_COMP, because compound pages cannot be split.
As shown by [1] things may silently work until the requested size
(possibly depending on user) stops being power of two. Then for
CONFIG_DEBUG_VM, BUG_ON() triggers in split_page(). Without
CONFIG_DEBUG_VM, consequences are unclear.
There are several options here, none of them great:
1) Don't do the splitting when __GFP_COMP is passed, and return the
whole compound page. However if caller then returns it via
free_pages_exact(), that will be unexpected and the freeing actions
there will be wrong.
2) Warn and remove __GFP_COMP from the flags. But the caller may have
really wanted it, so things may break later somewhere.
3) Warn and return NULL. However NULL may be unexpected, especially
for small sizes.
This patch picks option 2, because as Michal Hocko put it: "callers wanted
it" is much less probable than "caller is simply confused and more gfp
flags is surely better than fewer".
[1] https://lore.kernel.org/lkml/20181126002805.GI18977@shao2-debian/T/#u
Link: http://lkml.kernel.org/r/0c6393eb-b28d-4607-c386-862a71f09de6@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Takashi Iwai <tiwai@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 00:16:47 +00:00
|
|
|
* @gfp_mask: GFP flags for the allocation, must not contain __GFP_COMP
|
2008-07-24 04:28:11 +00:00
|
|
|
*
|
|
|
|
* This function is similar to alloc_pages(), except that it allocates the
|
|
|
|
* minimum number of pages to satisfy the request. alloc_pages() can only
|
|
|
|
* allocate memory in power-of-two pages.
|
|
|
|
*
|
|
|
|
* This function is also limited by MAX_ORDER.
|
|
|
|
*
|
|
|
|
* Memory allocated by this function must be released by free_pages_exact().
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: pointer to the allocated area or %NULL in case of error.
|
2008-07-24 04:28:11 +00:00
|
|
|
*/
|
|
|
|
void *alloc_pages_exact(size_t size, gfp_t gfp_mask)
|
|
|
|
{
|
|
|
|
unsigned int order = get_order(size);
|
|
|
|
unsigned long addr;
|
|
|
|
|
mm, page_alloc: disallow __GFP_COMP in alloc_pages_exact()
alloc_pages_exact*() allocates a page of sufficient order and then splits
it to return only the number of pages requested. That makes it
incompatible with __GFP_COMP, because compound pages cannot be split.
As shown by [1] things may silently work until the requested size
(possibly depending on user) stops being power of two. Then for
CONFIG_DEBUG_VM, BUG_ON() triggers in split_page(). Without
CONFIG_DEBUG_VM, consequences are unclear.
There are several options here, none of them great:
1) Don't do the splitting when __GFP_COMP is passed, and return the
whole compound page. However if caller then returns it via
free_pages_exact(), that will be unexpected and the freeing actions
there will be wrong.
2) Warn and remove __GFP_COMP from the flags. But the caller may have
really wanted it, so things may break later somewhere.
3) Warn and return NULL. However NULL may be unexpected, especially
for small sizes.
This patch picks option 2, because as Michal Hocko put it: "callers wanted
it" is much less probable than "caller is simply confused and more gfp
flags is surely better than fewer".
[1] https://lore.kernel.org/lkml/20181126002805.GI18977@shao2-debian/T/#u
Link: http://lkml.kernel.org/r/0c6393eb-b28d-4607-c386-862a71f09de6@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Takashi Iwai <tiwai@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 00:16:47 +00:00
|
|
|
if (WARN_ON_ONCE(gfp_mask & __GFP_COMP))
|
|
|
|
gfp_mask &= ~__GFP_COMP;
|
|
|
|
|
2008-07-24 04:28:11 +00:00
|
|
|
addr = __get_free_pages(gfp_mask, order);
|
2011-05-11 22:13:34 +00:00
|
|
|
return make_alloc_exact(addr, order, size);
|
2008-07-24 04:28:11 +00:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(alloc_pages_exact);
|
|
|
|
|
2011-05-11 22:13:34 +00:00
|
|
|
/**
|
|
|
|
* alloc_pages_exact_nid - allocate an exact number of physically-contiguous
|
|
|
|
* pages on a node.
|
2011-05-16 20:16:54 +00:00
|
|
|
* @nid: the preferred node ID where memory should be allocated
|
2011-05-11 22:13:34 +00:00
|
|
|
* @size: the number of bytes to allocate
|
mm, page_alloc: disallow __GFP_COMP in alloc_pages_exact()
alloc_pages_exact*() allocates a page of sufficient order and then splits
it to return only the number of pages requested. That makes it
incompatible with __GFP_COMP, because compound pages cannot be split.
As shown by [1] things may silently work until the requested size
(possibly depending on user) stops being power of two. Then for
CONFIG_DEBUG_VM, BUG_ON() triggers in split_page(). Without
CONFIG_DEBUG_VM, consequences are unclear.
There are several options here, none of them great:
1) Don't do the splitting when __GFP_COMP is passed, and return the
whole compound page. However if caller then returns it via
free_pages_exact(), that will be unexpected and the freeing actions
there will be wrong.
2) Warn and remove __GFP_COMP from the flags. But the caller may have
really wanted it, so things may break later somewhere.
3) Warn and return NULL. However NULL may be unexpected, especially
for small sizes.
This patch picks option 2, because as Michal Hocko put it: "callers wanted
it" is much less probable than "caller is simply confused and more gfp
flags is surely better than fewer".
[1] https://lore.kernel.org/lkml/20181126002805.GI18977@shao2-debian/T/#u
Link: http://lkml.kernel.org/r/0c6393eb-b28d-4607-c386-862a71f09de6@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Takashi Iwai <tiwai@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 00:16:47 +00:00
|
|
|
* @gfp_mask: GFP flags for the allocation, must not contain __GFP_COMP
|
2011-05-11 22:13:34 +00:00
|
|
|
*
|
|
|
|
* Like alloc_pages_exact(), but try to allocate on node nid first before falling
|
|
|
|
* back.
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: pointer to the allocated area or %NULL in case of error.
|
2011-05-11 22:13:34 +00:00
|
|
|
*/
|
2014-08-06 23:04:59 +00:00
|
|
|
void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask)
|
2011-05-11 22:13:34 +00:00
|
|
|
{
|
2015-11-07 00:29:57 +00:00
|
|
|
unsigned int order = get_order(size);
|
mm, page_alloc: disallow __GFP_COMP in alloc_pages_exact()
alloc_pages_exact*() allocates a page of sufficient order and then splits
it to return only the number of pages requested. That makes it
incompatible with __GFP_COMP, because compound pages cannot be split.
As shown by [1] things may silently work until the requested size
(possibly depending on user) stops being power of two. Then for
CONFIG_DEBUG_VM, BUG_ON() triggers in split_page(). Without
CONFIG_DEBUG_VM, consequences are unclear.
There are several options here, none of them great:
1) Don't do the splitting when __GFP_COMP is passed, and return the
whole compound page. However if caller then returns it via
free_pages_exact(), that will be unexpected and the freeing actions
there will be wrong.
2) Warn and remove __GFP_COMP from the flags. But the caller may have
really wanted it, so things may break later somewhere.
3) Warn and return NULL. However NULL may be unexpected, especially
for small sizes.
This patch picks option 2, because as Michal Hocko put it: "callers wanted
it" is much less probable than "caller is simply confused and more gfp
flags is surely better than fewer".
[1] https://lore.kernel.org/lkml/20181126002805.GI18977@shao2-debian/T/#u
Link: http://lkml.kernel.org/r/0c6393eb-b28d-4607-c386-862a71f09de6@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Takashi Iwai <tiwai@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 00:16:47 +00:00
|
|
|
struct page *p;
|
|
|
|
|
|
|
|
if (WARN_ON_ONCE(gfp_mask & __GFP_COMP))
|
|
|
|
gfp_mask &= ~__GFP_COMP;
|
|
|
|
|
|
|
|
p = alloc_pages_node(nid, gfp_mask, order);
|
2011-05-11 22:13:34 +00:00
|
|
|
if (!p)
|
|
|
|
return NULL;
|
|
|
|
return make_alloc_exact((unsigned long)page_address(p), order, size);
|
|
|
|
}
|
|
|
|
|
2008-07-24 04:28:11 +00:00
|
|
|
/**
|
|
|
|
* free_pages_exact - release memory allocated via alloc_pages_exact()
|
|
|
|
* @virt: the value returned by alloc_pages_exact.
|
|
|
|
* @size: size of allocation, same value as passed to alloc_pages_exact().
|
|
|
|
*
|
|
|
|
* Release the memory allocated by a previous call to alloc_pages_exact.
|
|
|
|
*/
|
|
|
|
void free_pages_exact(void *virt, size_t size)
|
|
|
|
{
|
|
|
|
unsigned long addr = (unsigned long)virt;
|
|
|
|
unsigned long end = addr + PAGE_ALIGN(size);
|
|
|
|
|
|
|
|
while (addr < end) {
|
|
|
|
free_page(addr);
|
|
|
|
addr += PAGE_SIZE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(free_pages_exact);
|
|
|
|
|
2013-02-23 00:35:54 +00:00
|
|
|
/**
|
|
|
|
* nr_free_zone_pages - count number of pages beyond high watermark
|
|
|
|
* @offset: The zone index of the highest zone
|
|
|
|
*
|
2019-03-05 23:48:42 +00:00
|
|
|
* nr_free_zone_pages() counts the number of pages which are beyond the
|
2013-02-23 00:35:54 +00:00
|
|
|
* high watermark within all zones at or below a given zone index. For each
|
|
|
|
* zone, the number of pages is calculated as:
|
2017-03-30 20:11:36 +00:00
|
|
|
*
|
|
|
|
* nr_free_zone_pages = managed_pages - high_pages
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: number of pages beyond high watermark.
|
2013-02-23 00:35:54 +00:00
|
|
|
*/
|
2013-02-23 00:35:43 +00:00
|
|
|
static unsigned long nr_free_zone_pages(int offset)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2008-04-28 09:12:17 +00:00
|
|
|
struct zoneref *z;
|
2008-04-28 09:12:16 +00:00
|
|
|
struct zone *zone;
|
|
|
|
|
2005-07-30 05:59:18 +00:00
|
|
|
/* Just pick one node, since fallback list is circular */
|
2013-02-23 00:35:43 +00:00
|
|
|
unsigned long sum = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-04-28 09:12:14 +00:00
|
|
|
struct zonelist *zonelist = node_zonelist(numa_node_id(), GFP_KERNEL);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-04-28 09:12:16 +00:00
|
|
|
for_each_zone_zonelist(zone, z, zonelist, offset) {
|
2018-12-28 08:34:24 +00:00
|
|
|
unsigned long size = zone_managed_pages(zone);
|
2009-06-16 22:32:12 +00:00
|
|
|
unsigned long high = high_wmark_pages(zone);
|
2005-07-30 05:59:18 +00:00
|
|
|
if (size > high)
|
|
|
|
sum += size - high;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return sum;
|
|
|
|
}
|
|
|
|
|
2013-02-23 00:35:54 +00:00
|
|
|
/**
|
|
|
|
* nr_free_buffer_pages - count number of pages beyond high watermark
|
|
|
|
*
|
|
|
|
* nr_free_buffer_pages() counts the number of pages which are beyond the high
|
|
|
|
* watermark within ZONE_DMA and ZONE_NORMAL.
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: number of pages beyond high watermark within ZONE_DMA and
|
|
|
|
* ZONE_NORMAL.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2013-02-23 00:35:43 +00:00
|
|
|
unsigned long nr_free_buffer_pages(void)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2005-10-21 06:55:38 +00:00
|
|
|
return nr_free_zone_pages(gfp_zone(GFP_USER));
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2007-07-17 11:04:39 +00:00
|
|
|
EXPORT_SYMBOL_GPL(nr_free_buffer_pages);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-09-27 08:50:06 +00:00
|
|
|
static inline void show_node(struct zone *zone)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2012-12-12 00:00:29 +00:00
|
|
|
if (IS_ENABLED(CONFIG_NUMA))
|
2006-12-07 04:33:03 +00:00
|
|
|
printk("Node %d ", zone_to_nid(zone));
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2016-03-17 21:19:05 +00:00
|
|
|
long si_mem_available(void)
|
|
|
|
{
|
|
|
|
long available;
|
|
|
|
unsigned long pagecache;
|
|
|
|
unsigned long wmark_low = 0;
|
|
|
|
unsigned long pages[NR_LRU_LISTS];
|
2018-10-26 22:05:46 +00:00
|
|
|
unsigned long reclaimable;
|
2016-03-17 21:19:05 +00:00
|
|
|
struct zone *zone;
|
|
|
|
int lru;
|
|
|
|
|
|
|
|
for (lru = LRU_BASE; lru < NR_LRU_LISTS; lru++)
|
2016-08-11 22:32:57 +00:00
|
|
|
pages[lru] = global_node_page_state(NR_LRU_BASE + lru);
|
2016-03-17 21:19:05 +00:00
|
|
|
|
|
|
|
for_each_zone(zone)
|
2018-12-28 08:35:44 +00:00
|
|
|
wmark_low += low_wmark_pages(zone);
|
2016-03-17 21:19:05 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Estimate the amount of memory available for userspace allocations,
|
|
|
|
* without causing swapping.
|
|
|
|
*/
|
2017-09-06 23:23:36 +00:00
|
|
|
available = global_zone_page_state(NR_FREE_PAGES) - totalreserve_pages;
|
2016-03-17 21:19:05 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Not all the page cache can be freed, otherwise the system will
|
|
|
|
* start swapping. Assume at least half of the page cache, or the
|
|
|
|
* low watermark worth of cache, needs to stay.
|
|
|
|
*/
|
|
|
|
pagecache = pages[LRU_ACTIVE_FILE] + pages[LRU_INACTIVE_FILE];
|
|
|
|
pagecache -= min(pagecache / 2, wmark_low);
|
|
|
|
available += pagecache;
|
|
|
|
|
|
|
|
/*
|
2018-10-26 22:05:46 +00:00
|
|
|
* Part of the reclaimable slab and other kernel memory consists of
|
|
|
|
* items that are in use, and cannot be freed. Cap this estimate at the
|
|
|
|
* low watermark.
|
2016-03-17 21:19:05 +00:00
|
|
|
*/
|
2020-08-07 06:20:39 +00:00
|
|
|
reclaimable = global_node_page_state_pages(NR_SLAB_RECLAIMABLE_B) +
|
|
|
|
global_node_page_state(NR_KERNEL_MISC_RECLAIMABLE);
|
2018-10-26 22:05:46 +00:00
|
|
|
available += reclaimable - min(reclaimable / 2, wmark_low);
|
2018-04-10 23:27:40 +00:00
|
|
|
|
2016-03-17 21:19:05 +00:00
|
|
|
if (available < 0)
|
|
|
|
available = 0;
|
|
|
|
return available;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(si_mem_available);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
void si_meminfo(struct sysinfo *val)
|
|
|
|
{
|
2018-12-28 08:34:29 +00:00
|
|
|
val->totalram = totalram_pages();
|
2016-07-28 22:46:20 +00:00
|
|
|
val->sharedram = global_node_page_state(NR_SHMEM);
|
2017-09-06 23:23:36 +00:00
|
|
|
val->freeram = global_zone_page_state(NR_FREE_PAGES);
|
2005-04-16 22:20:36 +00:00
|
|
|
val->bufferram = nr_blockdev_pages();
|
2018-12-28 08:34:29 +00:00
|
|
|
val->totalhigh = totalhigh_pages();
|
2005-04-16 22:20:36 +00:00
|
|
|
val->freehigh = nr_free_highpages();
|
|
|
|
val->mem_unit = PAGE_SIZE;
|
|
|
|
}
|
|
|
|
|
|
|
|
EXPORT_SYMBOL(si_meminfo);
|
|
|
|
|
|
|
|
#ifdef CONFIG_NUMA
|
|
|
|
void si_meminfo_node(struct sysinfo *val, int nid)
|
|
|
|
{
|
2013-07-03 22:03:27 +00:00
|
|
|
int zone_type; /* needs to be signed */
|
|
|
|
unsigned long managed_pages = 0;
|
2016-05-20 00:12:23 +00:00
|
|
|
unsigned long managed_highpages = 0;
|
|
|
|
unsigned long free_highpages = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
pg_data_t *pgdat = NODE_DATA(nid);
|
|
|
|
|
2013-07-03 22:03:27 +00:00
|
|
|
for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++)
|
2018-12-28 08:34:24 +00:00
|
|
|
managed_pages += zone_managed_pages(&pgdat->node_zones[zone_type]);
|
2013-07-03 22:03:27 +00:00
|
|
|
val->totalram = managed_pages;
|
2016-07-28 22:46:20 +00:00
|
|
|
val->sharedram = node_page_state(pgdat, NR_SHMEM);
|
mm, vmstat: add infrastructure for per-node vmstats
Patchset: "Move LRU page reclaim from zones to nodes v9"
This series moves LRUs from the zones to the node. While this is a
current rebase, the test results were based on mmotm as of June 23rd.
Conceptually, this series is simple but there are a lot of details.
Some of the broad motivations for this are;
1. The residency of a page partially depends on what zone the page was
allocated from. This is partially combatted by the fair zone allocation
policy but that is a partial solution that introduces overhead in the
page allocator paths.
2. Currently, reclaim on node 0 behaves slightly different to node 1. For
example, direct reclaim scans in zonelist order and reclaims even if
the zone is over the high watermark regardless of the age of pages
in that LRU. Kswapd on the other hand starts reclaim on the highest
unbalanced zone. A difference in distribution of file/anon pages due
to when they were allocated results can result in a difference in
again. While the fair zone allocation policy mitigates some of the
problems here, the page reclaim results on a multi-zone node will
always be different to a single-zone node.
it was scheduled on as a result.
3. kswapd and the page allocator scan zones in the opposite order to
avoid interfering with each other but it's sensitive to timing. This
mitigates the page allocator using pages that were allocated very recently
in the ideal case but it's sensitive to timing. When kswapd is allocating
from lower zones then it's great but during the rebalancing of the highest
zone, the page allocator and kswapd interfere with each other. It's worse
if the highest zone is small and difficult to balance.
4. slab shrinkers are node-based which makes it harder to identify the exact
relationship between slab reclaim and LRU reclaim.
The reason we have zone-based reclaim is that we used to have
large highmem zones in common configurations and it was necessary
to quickly find ZONE_NORMAL pages for reclaim. Today, this is much
less of a concern as machines with lots of memory will (or should) use
64-bit kernels. Combinations of 32-bit hardware and 64-bit hardware are
rare. Machines that do use highmem should have relatively low highmem:lowmem
ratios than we worried about in the past.
Conceptually, moving to node LRUs should be easier to understand. The
page allocator plays fewer tricks to game reclaim and reclaim behaves
similarly on all nodes.
The series has been tested on a 16 core UMA machine and a 2-socket 48
core NUMA machine. The UMA results are presented in most cases as the NUMA
machine behaved similarly.
pagealloc
---------
This is a microbenchmark that shows the benefit of removing the fair zone
allocation policy. It was tested uip to order-4 but only orders 0 and 1 are
shown as the other orders were comparable.
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v9
Min total-odr0-1 490.00 ( 0.00%) 457.00 ( 6.73%)
Min total-odr0-2 347.00 ( 0.00%) 329.00 ( 5.19%)
Min total-odr0-4 288.00 ( 0.00%) 273.00 ( 5.21%)
Min total-odr0-8 251.00 ( 0.00%) 239.00 ( 4.78%)
Min total-odr0-16 234.00 ( 0.00%) 222.00 ( 5.13%)
Min total-odr0-32 223.00 ( 0.00%) 211.00 ( 5.38%)
Min total-odr0-64 217.00 ( 0.00%) 208.00 ( 4.15%)
Min total-odr0-128 214.00 ( 0.00%) 204.00 ( 4.67%)
Min total-odr0-256 250.00 ( 0.00%) 230.00 ( 8.00%)
Min total-odr0-512 271.00 ( 0.00%) 269.00 ( 0.74%)
Min total-odr0-1024 291.00 ( 0.00%) 282.00 ( 3.09%)
Min total-odr0-2048 303.00 ( 0.00%) 296.00 ( 2.31%)
Min total-odr0-4096 311.00 ( 0.00%) 309.00 ( 0.64%)
Min total-odr0-8192 316.00 ( 0.00%) 314.00 ( 0.63%)
Min total-odr0-16384 317.00 ( 0.00%) 315.00 ( 0.63%)
Min total-odr1-1 742.00 ( 0.00%) 712.00 ( 4.04%)
Min total-odr1-2 562.00 ( 0.00%) 530.00 ( 5.69%)
Min total-odr1-4 457.00 ( 0.00%) 433.00 ( 5.25%)
Min total-odr1-8 411.00 ( 0.00%) 381.00 ( 7.30%)
Min total-odr1-16 381.00 ( 0.00%) 356.00 ( 6.56%)
Min total-odr1-32 372.00 ( 0.00%) 346.00 ( 6.99%)
Min total-odr1-64 372.00 ( 0.00%) 343.00 ( 7.80%)
Min total-odr1-128 375.00 ( 0.00%) 351.00 ( 6.40%)
Min total-odr1-256 379.00 ( 0.00%) 351.00 ( 7.39%)
Min total-odr1-512 385.00 ( 0.00%) 355.00 ( 7.79%)
Min total-odr1-1024 386.00 ( 0.00%) 358.00 ( 7.25%)
Min total-odr1-2048 390.00 ( 0.00%) 362.00 ( 7.18%)
Min total-odr1-4096 390.00 ( 0.00%) 362.00 ( 7.18%)
Min total-odr1-8192 388.00 ( 0.00%) 363.00 ( 6.44%)
This shows a steady improvement throughout. The primary benefit is from
reduced system CPU usage which is obvious from the overall times;
4.7.0-rc4 4.7.0-rc4
mmotm-20160623nodelru-v8
User 189.19 191.80
System 2604.45 2533.56
Elapsed 2855.30 2786.39
The vmstats also showed that the fair zone allocation policy was definitely
removed as can be seen here;
4.7.0-rc3 4.7.0-rc3
mmotm-20160623 nodelru-v8
DMA32 allocs 28794729769 0
Normal allocs 48432501431 77227309877
Movable allocs 0 0
tiobench on ext4
----------------
tiobench is a benchmark that artifically benefits if old pages remain resident
while new pages get reclaimed. The fair zone allocation policy mitigates this
problem so pages age fairly. While the benchmark has problems, it is important
that tiobench performance remains constant as it implies that page aging
problems that the fair zone allocation policy fixes are not re-introduced.
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v9
Min PotentialReadSpeed 89.65 ( 0.00%) 90.21 ( 0.62%)
Min SeqRead-MB/sec-1 82.68 ( 0.00%) 82.01 ( -0.81%)
Min SeqRead-MB/sec-2 72.76 ( 0.00%) 72.07 ( -0.95%)
Min SeqRead-MB/sec-4 75.13 ( 0.00%) 74.92 ( -0.28%)
Min SeqRead-MB/sec-8 64.91 ( 0.00%) 65.19 ( 0.43%)
Min SeqRead-MB/sec-16 62.24 ( 0.00%) 62.22 ( -0.03%)
Min RandRead-MB/sec-1 0.88 ( 0.00%) 0.88 ( 0.00%)
Min RandRead-MB/sec-2 0.95 ( 0.00%) 0.92 ( -3.16%)
Min RandRead-MB/sec-4 1.43 ( 0.00%) 1.34 ( -6.29%)
Min RandRead-MB/sec-8 1.61 ( 0.00%) 1.60 ( -0.62%)
Min RandRead-MB/sec-16 1.80 ( 0.00%) 1.90 ( 5.56%)
Min SeqWrite-MB/sec-1 76.41 ( 0.00%) 76.85 ( 0.58%)
Min SeqWrite-MB/sec-2 74.11 ( 0.00%) 73.54 ( -0.77%)
Min SeqWrite-MB/sec-4 80.05 ( 0.00%) 80.13 ( 0.10%)
Min SeqWrite-MB/sec-8 72.88 ( 0.00%) 73.20 ( 0.44%)
Min SeqWrite-MB/sec-16 75.91 ( 0.00%) 76.44 ( 0.70%)
Min RandWrite-MB/sec-1 1.18 ( 0.00%) 1.14 ( -3.39%)
Min RandWrite-MB/sec-2 1.02 ( 0.00%) 1.03 ( 0.98%)
Min RandWrite-MB/sec-4 1.05 ( 0.00%) 0.98 ( -6.67%)
Min RandWrite-MB/sec-8 0.89 ( 0.00%) 0.92 ( 3.37%)
Min RandWrite-MB/sec-16 0.92 ( 0.00%) 0.93 ( 1.09%)
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 approx-v9
User 645.72 525.90
System 403.85 331.75
Elapsed 6795.36 6783.67
This shows that the series has little or not impact on tiobench which is
desirable and a reduction in system CPU usage. It indicates that the fair
zone allocation policy was removed in a manner that didn't reintroduce
one class of page aging bug. There were only minor differences in overall
reclaim activity
4.7.0-rc4 4.7.0-rc4
mmotm-20160623nodelru-v8
Minor Faults 645838 647465
Major Faults 573 640
Swap Ins 0 0
Swap Outs 0 0
DMA allocs 0 0
DMA32 allocs 46041453 44190646
Normal allocs 78053072 79887245
Movable allocs 0 0
Allocation stalls 24 67
Stall zone DMA 0 0
Stall zone DMA32 0 0
Stall zone Normal 0 2
Stall zone HighMem 0 0
Stall zone Movable 0 65
Direct pages scanned 10969 30609
Kswapd pages scanned 93375144 93492094
Kswapd pages reclaimed 93372243 93489370
Direct pages reclaimed 10969 30609
Kswapd efficiency 99% 99%
Kswapd velocity 13741.015 13781.934
Direct efficiency 100% 100%
Direct velocity 1.614 4.512
Percentage direct scans 0% 0%
kswapd activity was roughly comparable. There were differences in direct
reclaim activity but negligible in the context of the overall workload
(velocity of 4 pages per second with the patches applied, 1.6 pages per
second in the baseline kernel).
pgbench read-only large configuration on ext4
---------------------------------------------
pgbench is a database benchmark that can be sensitive to page reclaim
decisions. This also checks if removing the fair zone allocation policy
is safe
pgbench Transactions
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v8
Hmean 1 188.26 ( 0.00%) 189.78 ( 0.81%)
Hmean 5 330.66 ( 0.00%) 328.69 ( -0.59%)
Hmean 12 370.32 ( 0.00%) 380.72 ( 2.81%)
Hmean 21 368.89 ( 0.00%) 369.00 ( 0.03%)
Hmean 30 382.14 ( 0.00%) 360.89 ( -5.56%)
Hmean 32 428.87 ( 0.00%) 432.96 ( 0.95%)
Negligible differences again. As with tiobench, overall reclaim activity
was comparable.
bonnie++ on ext4
----------------
No interesting performance difference, negligible differences on reclaim
stats.
paralleldd on ext4
------------------
This workload uses varying numbers of dd instances to read large amounts of
data from disk.
4.7.0-rc3 4.7.0-rc3
mmotm-20160623 nodelru-v9
Amean Elapsd-1 186.04 ( 0.00%) 189.41 ( -1.82%)
Amean Elapsd-3 192.27 ( 0.00%) 191.38 ( 0.46%)
Amean Elapsd-5 185.21 ( 0.00%) 182.75 ( 1.33%)
Amean Elapsd-7 183.71 ( 0.00%) 182.11 ( 0.87%)
Amean Elapsd-12 180.96 ( 0.00%) 181.58 ( -0.35%)
Amean Elapsd-16 181.36 ( 0.00%) 183.72 ( -1.30%)
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v9
User 1548.01 1552.44
System 8609.71 8515.08
Elapsed 3587.10 3594.54
There is little or no change in performance but some drop in system CPU usage.
4.7.0-rc3 4.7.0-rc3
mmotm-20160623 nodelru-v9
Minor Faults 362662 367360
Major Faults 1204 1143
Swap Ins 22 0
Swap Outs 2855 1029
DMA allocs 0 0
DMA32 allocs 31409797 28837521
Normal allocs 46611853 49231282
Movable allocs 0 0
Direct pages scanned 0 0
Kswapd pages scanned 40845270 40869088
Kswapd pages reclaimed 40830976 40855294
Direct pages reclaimed 0 0
Kswapd efficiency 99% 99%
Kswapd velocity 11386.711 11369.769
Direct efficiency 100% 100%
Direct velocity 0.000 0.000
Percentage direct scans 0% 0%
Page writes by reclaim 2855 1029
Page writes file 0 0
Page writes anon 2855 1029
Page reclaim immediate 771 1628
Sector Reads 293312636 293536360
Sector Writes 18213568 18186480
Page rescued immediate 0 0
Slabs scanned 128257 132747
Direct inode steals 181 56
Kswapd inode steals 59 1131
It basically shows that kswapd was active at roughly the same rate in
both kernels. There was also comparable slab scanning activity and direct
reclaim was avoided in both cases. There appears to be a large difference
in numbers of inodes reclaimed but the workload has few active inodes and
is likely a timing artifact.
stutter
-------
stutter simulates a simple workload. One part uses a lot of anonymous
memory, a second measures mmap latency and a third copies a large file.
The primary metric is checking for mmap latency.
stutter
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v8
Min mmap 16.6283 ( 0.00%) 13.4258 ( 19.26%)
1st-qrtle mmap 54.7570 ( 0.00%) 34.9121 ( 36.24%)
2nd-qrtle mmap 57.3163 ( 0.00%) 46.1147 ( 19.54%)
3rd-qrtle mmap 58.9976 ( 0.00%) 47.1882 ( 20.02%)
Max-90% mmap 59.7433 ( 0.00%) 47.4453 ( 20.58%)
Max-93% mmap 60.1298 ( 0.00%) 47.6037 ( 20.83%)
Max-95% mmap 73.4112 ( 0.00%) 82.8719 (-12.89%)
Max-99% mmap 92.8542 ( 0.00%) 88.8870 ( 4.27%)
Max mmap 1440.6569 ( 0.00%) 121.4201 ( 91.57%)
Mean mmap 59.3493 ( 0.00%) 42.2991 ( 28.73%)
Best99%Mean mmap 57.2121 ( 0.00%) 41.8207 ( 26.90%)
Best95%Mean mmap 55.9113 ( 0.00%) 39.9620 ( 28.53%)
Best90%Mean mmap 55.6199 ( 0.00%) 39.3124 ( 29.32%)
Best50%Mean mmap 53.2183 ( 0.00%) 33.1307 ( 37.75%)
Best10%Mean mmap 45.9842 ( 0.00%) 20.4040 ( 55.63%)
Best5%Mean mmap 43.2256 ( 0.00%) 17.9654 ( 58.44%)
Best1%Mean mmap 32.9388 ( 0.00%) 16.6875 ( 49.34%)
This shows a number of improvements with the worst-case outlier greatly
improved.
Some of the vmstats are interesting
4.7.0-rc4 4.7.0-rc4
mmotm-20160623nodelru-v8
Swap Ins 163 502
Swap Outs 0 0
DMA allocs 0 0
DMA32 allocs 618719206 1381662383
Normal allocs 891235743 564138421
Movable allocs 0 0
Allocation stalls 2603 1
Direct pages scanned 216787 2
Kswapd pages scanned 50719775 41778378
Kswapd pages reclaimed 41541765 41777639
Direct pages reclaimed 209159 0
Kswapd efficiency 81% 99%
Kswapd velocity 16859.554 14329.059
Direct efficiency 96% 0%
Direct velocity 72.061 0.001
Percentage direct scans 0% 0%
Page writes by reclaim 6215049 0
Page writes file 6215049 0
Page writes anon 0 0
Page reclaim immediate 70673 90
Sector Reads 81940800 81680456
Sector Writes 100158984 98816036
Page rescued immediate 0 0
Slabs scanned 1366954 22683
While this is not guaranteed in all cases, this particular test showed
a large reduction in direct reclaim activity. It's also worth noting
that no page writes were issued from reclaim context.
This series is not without its hazards. There are at least three areas
that I'm concerned with even though I could not reproduce any problems in
that area.
1. Reclaim/compaction is going to be affected because the amount of reclaim is
no longer targetted at a specific zone. Compaction works on a per-zone basis
so there is no guarantee that reclaiming a few THP's worth page pages will
have a positive impact on compaction success rates.
2. The Slab/LRU reclaim ratio is affected because the frequency the shrinkers
are called is now different. This may or may not be a problem but if it
is, it'll be because shrinkers are not called enough and some balancing
is required.
3. The anon/file reclaim ratio may be affected. Pages about to be dirtied are
distributed between zones and the fair zone allocation policy used to do
something very similar for anon. The distribution is now different but not
necessarily in any way that matters but it's still worth bearing in mind.
VM statistic counters for reclaim decisions are zone-based. If the kernel
is to reclaim on a per-node basis then we need to track per-node
statistics but there is no infrastructure for that. The most notable
change is that the old node_page_state is renamed to
sum_zone_node_page_state. The new node_page_state takes a pglist_data and
uses per-node stats but none exist yet. There is some renaming such as
vm_stat to vm_zone_stat and the addition of vm_node_stat and the renaming
of mod_state to mod_zone_state. Otherwise, this is mostly a mechanical
patch with no functional change. There is a lot of similarity between the
node and zone helpers which is unfortunate but there was no obvious way of
reusing the code and maintaining type safety.
Link: http://lkml.kernel.org/r/1467970510-21195-2-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Rik van Riel <riel@surriel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:45:24 +00:00
|
|
|
val->freeram = sum_zone_node_page_state(nid, NR_FREE_PAGES);
|
2006-09-26 06:31:12 +00:00
|
|
|
#ifdef CONFIG_HIGHMEM
|
2016-05-20 00:12:23 +00:00
|
|
|
for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++) {
|
|
|
|
struct zone *zone = &pgdat->node_zones[zone_type];
|
|
|
|
|
|
|
|
if (is_highmem(zone)) {
|
2018-12-28 08:34:24 +00:00
|
|
|
managed_highpages += zone_managed_pages(zone);
|
2016-05-20 00:12:23 +00:00
|
|
|
free_highpages += zone_page_state(zone, NR_FREE_PAGES);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
val->totalhigh = managed_highpages;
|
|
|
|
val->freehigh = free_highpages;
|
2006-09-26 06:31:12 +00:00
|
|
|
#else
|
2016-05-20 00:12:23 +00:00
|
|
|
val->totalhigh = managed_highpages;
|
|
|
|
val->freehigh = free_highpages;
|
2006-09-26 06:31:12 +00:00
|
|
|
#endif
|
2005-04-16 22:20:36 +00:00
|
|
|
val->mem_unit = PAGE_SIZE;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2011-03-22 23:30:46 +00:00
|
|
|
/*
|
2011-05-25 00:11:16 +00:00
|
|
|
* Determine whether the node should be displayed or not, depending on whether
|
|
|
|
* SHOW_MEM_FILTER_NODES was passed to show_free_areas().
|
2011-03-22 23:30:46 +00:00
|
|
|
*/
|
2017-02-22 23:46:16 +00:00
|
|
|
static bool show_mem_node_skip(unsigned int flags, int nid, nodemask_t *nodemask)
|
2011-03-22 23:30:46 +00:00
|
|
|
{
|
|
|
|
if (!(flags & SHOW_MEM_FILTER_NODES))
|
2017-02-22 23:46:16 +00:00
|
|
|
return false;
|
2011-03-22 23:30:46 +00:00
|
|
|
|
2017-02-22 23:46:16 +00:00
|
|
|
/*
|
|
|
|
* no node mask - aka implicit memory numa policy. Do not bother with
|
|
|
|
* the synchronization - read_mems_allowed_begin - because we do not
|
|
|
|
* have to be precise here.
|
|
|
|
*/
|
|
|
|
if (!nodemask)
|
|
|
|
nodemask = &cpuset_current_mems_allowed;
|
|
|
|
|
|
|
|
return !node_isset(nid, *nodemask);
|
2011-03-22 23:30:46 +00:00
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
#define K(x) ((x) << (PAGE_SHIFT-10))
|
|
|
|
|
2012-12-12 00:00:24 +00:00
|
|
|
static void show_migration_types(unsigned char type)
|
|
|
|
{
|
|
|
|
static const char types[MIGRATE_TYPES] = {
|
|
|
|
[MIGRATE_UNMOVABLE] = 'U',
|
|
|
|
[MIGRATE_MOVABLE] = 'M',
|
2015-12-11 21:40:29 +00:00
|
|
|
[MIGRATE_RECLAIMABLE] = 'E',
|
|
|
|
[MIGRATE_HIGHATOMIC] = 'H',
|
2012-12-12 00:00:24 +00:00
|
|
|
#ifdef CONFIG_CMA
|
|
|
|
[MIGRATE_CMA] = 'C',
|
|
|
|
#endif
|
2013-02-23 00:33:58 +00:00
|
|
|
#ifdef CONFIG_MEMORY_ISOLATION
|
2012-12-12 00:00:24 +00:00
|
|
|
[MIGRATE_ISOLATE] = 'I',
|
2013-02-23 00:33:58 +00:00
|
|
|
#endif
|
2012-12-12 00:00:24 +00:00
|
|
|
};
|
|
|
|
char tmp[MIGRATE_TYPES + 1];
|
|
|
|
char *p = tmp;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < MIGRATE_TYPES; i++) {
|
|
|
|
if (type & (1 << i))
|
|
|
|
*p++ = types[i];
|
|
|
|
}
|
|
|
|
|
|
|
|
*p = '\0';
|
2016-10-28 00:46:29 +00:00
|
|
|
printk(KERN_CONT "(%s) ", tmp);
|
2012-12-12 00:00:24 +00:00
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Show free area list (used inside shift_scroll-lock stuff)
|
|
|
|
* We also calculate the percentage fragmentation. We do this by counting the
|
|
|
|
* memory on each free list with the exception of the first item on the list.
|
2015-04-14 22:45:30 +00:00
|
|
|
*
|
|
|
|
* Bits in @filter:
|
|
|
|
* SHOW_MEM_FILTER_NODES: suppress nodes that are not allowed by current's
|
|
|
|
* cpuset.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2017-02-22 23:46:16 +00:00
|
|
|
void show_free_areas(unsigned int filter, nodemask_t *nodemask)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2015-04-14 22:45:30 +00:00
|
|
|
unsigned long free_pcp = 0;
|
[PATCH] Condense output of show_free_areas()
On larger systems, the amount of output dumped on the console when you do
SysRq-M is beyond insane. This patch is trying to reduce it somewhat as
even with the smaller NUMA systems that have hit the desktop this seems to
be a fair thing to do.
The philosophy I have taken is as follows:
1) If a zone is empty, don't tell, we don't need yet another line
telling us so. The information is available since one can look up
the fact how many zones were initialized in the first place.
2) Put as much information on a line is possible, if it can be done
in one line, rahter than two, then do it in one. I tried to format
the temperature stuff for easy reading.
Change show_free_areas() to not print lines for empty zones. If no zone
output is printed, the zone is empty. This reduces the number of lines
dumped to the console in sysrq on a large system by several thousand lines.
Change the zone temperature printouts to use one line per CPU instead of
two lines (one hot, one cold). On a 1024 CPU, 1024 node system, this
reduces the console output by over a million lines of output.
While this is a bigger problem on large NUMA systems, it is also applicable
to smaller desktop sized and mid range NUMA systems.
Old format:
Mem-info:
Node 0 DMA per-cpu:
cpu 0 hot: high 42, batch 7 used:24
cpu 0 cold: high 14, batch 3 used:1
cpu 1 hot: high 42, batch 7 used:34
cpu 1 cold: high 14, batch 3 used:0
cpu 2 hot: high 42, batch 7 used:0
cpu 2 cold: high 14, batch 3 used:0
cpu 3 hot: high 42, batch 7 used:0
cpu 3 cold: high 14, batch 3 used:0
cpu 4 hot: high 42, batch 7 used:0
cpu 4 cold: high 14, batch 3 used:0
cpu 5 hot: high 42, batch 7 used:0
cpu 5 cold: high 14, batch 3 used:0
cpu 6 hot: high 42, batch 7 used:0
cpu 6 cold: high 14, batch 3 used:0
cpu 7 hot: high 42, batch 7 used:0
cpu 7 cold: high 14, batch 3 used:0
Node 0 DMA32 per-cpu: empty
Node 0 Normal per-cpu: empty
Node 0 HighMem per-cpu: empty
Node 1 DMA per-cpu:
[snip]
Free pages: 5410688kB (0kB HighMem)
Active:9536 inactive:4261 dirty:6 writeback:0 unstable:0 free:338168 slab:1931 mapped:1900 pagetables:208
Node 0 DMA free:1676304kB min:3264kB low:4080kB high:4896kB active:128048kB inactive:61568kB present:1970880kB pages_scanned:0 all_unreclaimable? no
lowmem_reserve[]: 0 0 0 0
Node 0 DMA32 free:0kB min:0kB low:0kB high:0kB active:0kB inactive:0kB present:0kB pages_scanned:0 all_unreclaimable? no
lowmem_reserve[]: 0 0 0 0
Node 0 Normal free:0kB min:0kB low:0kB high:0kB active:0kB inactive:0kB present:0kB pages_scanned:0 all_unreclaimable? no
lowmem_reserve[]: 0 0 0 0
Node 0 HighMem free:0kB min:512kB low:512kB high:512kB active:0kB inactive:0kB present:0kB pages_scanned:0 all_unreclaimable? no
lowmem_reserve[]: 0 0 0 0
Node 1 DMA free:1951728kB min:3280kB low:4096kB high:4912kB active:5632kB inactive:1504kB present:1982464kB pages_scanned:0 all_unreclaimable? no
lowmem_reserve[]: 0 0 0 0
....
New format:
Mem-info:
Node 0 DMA per-cpu:
CPU 0: Hot: hi: 42, btch: 7 usd: 41 Cold: hi: 14, btch: 3 usd: 2
CPU 1: Hot: hi: 42, btch: 7 usd: 40 Cold: hi: 14, btch: 3 usd: 1
CPU 2: Hot: hi: 42, btch: 7 usd: 0 Cold: hi: 14, btch: 3 usd: 0
CPU 3: Hot: hi: 42, btch: 7 usd: 0 Cold: hi: 14, btch: 3 usd: 0
CPU 4: Hot: hi: 42, btch: 7 usd: 0 Cold: hi: 14, btch: 3 usd: 0
CPU 5: Hot: hi: 42, btch: 7 usd: 0 Cold: hi: 14, btch: 3 usd: 0
CPU 6: Hot: hi: 42, btch: 7 usd: 0 Cold: hi: 14, btch: 3 usd: 0
CPU 7: Hot: hi: 42, btch: 7 usd: 0 Cold: hi: 14, btch: 3 usd: 0
Node 1 DMA per-cpu:
[snip]
Free pages: 5411088kB (0kB HighMem)
Active:9558 inactive:4233 dirty:6 writeback:0 unstable:0 free:338193 slab:1942 mapped:1918 pagetables:208
Node 0 DMA free:1677648kB min:3264kB low:4080kB high:4896kB active:129296kB inactive:58864kB present:1970880kB pages_scanned:0 all_unreclaimable? no
lowmem_reserve[]: 0 0 0 0
Node 1 DMA free:1948448kB min:3280kB low:4096kB high:4912kB active:6864kB inactive:3536kB present:1982464kB pages_scanned:0 all_unreclaimable? no
lowmem_reserve[]: 0 0 0 0
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:50:05 +00:00
|
|
|
int cpu;
|
2005-04-16 22:20:36 +00:00
|
|
|
struct zone *zone;
|
2016-07-28 22:45:31 +00:00
|
|
|
pg_data_t *pgdat;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2009-03-31 22:19:31 +00:00
|
|
|
for_each_populated_zone(zone) {
|
2017-02-22 23:46:16 +00:00
|
|
|
if (show_mem_node_skip(filter, zone_to_nid(zone), nodemask))
|
2011-03-22 23:30:46 +00:00
|
|
|
continue;
|
2015-04-14 22:45:30 +00:00
|
|
|
|
2015-04-14 22:45:32 +00:00
|
|
|
for_each_online_cpu(cpu)
|
|
|
|
free_pcp += per_cpu_ptr(zone->pageset, cpu)->pcp.count;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2009-09-22 00:01:37 +00:00
|
|
|
printk("active_anon:%lu inactive_anon:%lu isolated_anon:%lu\n"
|
|
|
|
" active_file:%lu inactive_file:%lu isolated_file:%lu\n"
|
2020-06-02 04:48:21 +00:00
|
|
|
" unevictable:%lu dirty:%lu writeback:%lu\n"
|
2015-04-14 22:45:30 +00:00
|
|
|
" slab_reclaimable:%lu slab_unreclaimable:%lu\n"
|
2012-10-08 23:32:02 +00:00
|
|
|
" mapped:%lu shmem:%lu pagetables:%lu bounce:%lu\n"
|
2015-04-14 22:45:30 +00:00
|
|
|
" free:%lu free_pcp:%lu free_cma:%lu\n",
|
2016-07-28 22:45:31 +00:00
|
|
|
global_node_page_state(NR_ACTIVE_ANON),
|
|
|
|
global_node_page_state(NR_INACTIVE_ANON),
|
|
|
|
global_node_page_state(NR_ISOLATED_ANON),
|
|
|
|
global_node_page_state(NR_ACTIVE_FILE),
|
|
|
|
global_node_page_state(NR_INACTIVE_FILE),
|
|
|
|
global_node_page_state(NR_ISOLATED_FILE),
|
|
|
|
global_node_page_state(NR_UNEVICTABLE),
|
2016-07-28 22:46:20 +00:00
|
|
|
global_node_page_state(NR_FILE_DIRTY),
|
|
|
|
global_node_page_state(NR_WRITEBACK),
|
2020-08-07 06:20:39 +00:00
|
|
|
global_node_page_state_pages(NR_SLAB_RECLAIMABLE_B),
|
|
|
|
global_node_page_state_pages(NR_SLAB_UNRECLAIMABLE_B),
|
2016-07-28 22:46:14 +00:00
|
|
|
global_node_page_state(NR_FILE_MAPPED),
|
2016-07-28 22:46:20 +00:00
|
|
|
global_node_page_state(NR_SHMEM),
|
2020-12-15 03:07:17 +00:00
|
|
|
global_node_page_state(NR_PAGETABLE),
|
2017-09-06 23:23:36 +00:00
|
|
|
global_zone_page_state(NR_BOUNCE),
|
|
|
|
global_zone_page_state(NR_FREE_PAGES),
|
2015-04-14 22:45:30 +00:00
|
|
|
free_pcp,
|
2017-09-06 23:23:36 +00:00
|
|
|
global_zone_page_state(NR_FREE_CMA_PAGES));
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2016-07-28 22:45:31 +00:00
|
|
|
for_each_online_pgdat(pgdat) {
|
2017-02-22 23:46:16 +00:00
|
|
|
if (show_mem_node_skip(filter, pgdat->node_id, nodemask))
|
2017-02-22 23:46:07 +00:00
|
|
|
continue;
|
|
|
|
|
2016-07-28 22:45:31 +00:00
|
|
|
printk("Node %d"
|
|
|
|
" active_anon:%lukB"
|
|
|
|
" inactive_anon:%lukB"
|
|
|
|
" active_file:%lukB"
|
|
|
|
" inactive_file:%lukB"
|
|
|
|
" unevictable:%lukB"
|
|
|
|
" isolated(anon):%lukB"
|
|
|
|
" isolated(file):%lukB"
|
2016-07-28 22:46:14 +00:00
|
|
|
" mapped:%lukB"
|
2016-07-28 22:46:20 +00:00
|
|
|
" dirty:%lukB"
|
|
|
|
" writeback:%lukB"
|
|
|
|
" shmem:%lukB"
|
|
|
|
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
|
|
|
|
" shmem_thp: %lukB"
|
|
|
|
" shmem_pmdmapped: %lukB"
|
|
|
|
" anon_thp: %lukB"
|
|
|
|
#endif
|
|
|
|
" writeback_tmp:%lukB"
|
2020-08-07 06:21:37 +00:00
|
|
|
" kernel_stack:%lukB"
|
|
|
|
#ifdef CONFIG_SHADOW_CALL_STACK
|
|
|
|
" shadow_call_stack:%lukB"
|
|
|
|
#endif
|
2020-12-15 03:07:17 +00:00
|
|
|
" pagetables:%lukB"
|
2016-07-28 22:45:31 +00:00
|
|
|
" all_unreclaimable? %s"
|
|
|
|
"\n",
|
|
|
|
pgdat->node_id,
|
|
|
|
K(node_page_state(pgdat, NR_ACTIVE_ANON)),
|
|
|
|
K(node_page_state(pgdat, NR_INACTIVE_ANON)),
|
|
|
|
K(node_page_state(pgdat, NR_ACTIVE_FILE)),
|
|
|
|
K(node_page_state(pgdat, NR_INACTIVE_FILE)),
|
|
|
|
K(node_page_state(pgdat, NR_UNEVICTABLE)),
|
|
|
|
K(node_page_state(pgdat, NR_ISOLATED_ANON)),
|
|
|
|
K(node_page_state(pgdat, NR_ISOLATED_FILE)),
|
2016-07-28 22:46:14 +00:00
|
|
|
K(node_page_state(pgdat, NR_FILE_MAPPED)),
|
2016-07-28 22:46:20 +00:00
|
|
|
K(node_page_state(pgdat, NR_FILE_DIRTY)),
|
|
|
|
K(node_page_state(pgdat, NR_WRITEBACK)),
|
2017-04-07 23:04:45 +00:00
|
|
|
K(node_page_state(pgdat, NR_SHMEM)),
|
2016-07-28 22:46:20 +00:00
|
|
|
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
|
2021-02-24 20:03:31 +00:00
|
|
|
K(node_page_state(pgdat, NR_SHMEM_THPS)),
|
2021-02-24 20:03:35 +00:00
|
|
|
K(node_page_state(pgdat, NR_SHMEM_PMDMAPPED)),
|
2021-02-24 20:03:23 +00:00
|
|
|
K(node_page_state(pgdat, NR_ANON_THPS)),
|
2016-07-28 22:46:20 +00:00
|
|
|
#endif
|
|
|
|
K(node_page_state(pgdat, NR_WRITEBACK_TEMP)),
|
2020-08-07 06:21:37 +00:00
|
|
|
node_page_state(pgdat, NR_KERNEL_STACK_KB),
|
|
|
|
#ifdef CONFIG_SHADOW_CALL_STACK
|
|
|
|
node_page_state(pgdat, NR_KERNEL_SCS_KB),
|
|
|
|
#endif
|
2020-12-15 03:07:17 +00:00
|
|
|
K(node_page_state(pgdat, NR_PAGETABLE)),
|
mm: fix 100% CPU kswapd busyloop on unreclaimable nodes
Patch series "mm: kswapd spinning on unreclaimable nodes - fixes and
cleanups".
Jia reported a scenario in which the kswapd of a node indefinitely spins
at 100% CPU usage. We have seen similar cases at Facebook.
The kernel's current method of judging its ability to reclaim a node (or
whether to back off and sleep) is based on the amount of scanned pages
in proportion to the amount of reclaimable pages. In Jia's and our
scenarios, there are no reclaimable pages in the node, however, and the
condition for backing off is never met. Kswapd busyloops in an attempt
to restore the watermarks while having nothing to work with.
This series reworks the definition of an unreclaimable node based not on
scanning but on whether kswapd is able to actually reclaim pages in
MAX_RECLAIM_RETRIES (16) consecutive runs. This is the same criteria
the page allocator uses for giving up on direct reclaim and invoking the
OOM killer. If it cannot free any pages, kswapd will go to sleep and
leave further attempts to direct reclaim invocations, which will either
make progress and re-enable kswapd, or invoke the OOM killer.
Patch #1 fixes the immediate problem Jia reported, the remainder are
smaller fixlets, cleanups, and overall phasing out of the old method.
Patch #6 is the odd one out. It's a nice cleanup to get_scan_count(),
and directly related to #5, but in itself not relevant to the series.
If the whole series is too ambitious for 4.11, I would consider the
first three patches fixes, the rest cleanups.
This patch (of 9):
Jia He reports a problem with kswapd spinning at 100% CPU when
requesting more hugepages than memory available in the system:
$ echo 4000 >/proc/sys/vm/nr_hugepages
top - 13:42:59 up 3:37, 1 user, load average: 1.09, 1.03, 1.01
Tasks: 1 total, 1 running, 0 sleeping, 0 stopped, 0 zombie
%Cpu(s): 0.0 us, 12.5 sy, 0.0 ni, 85.5 id, 2.0 wa, 0.0 hi, 0.0 si, 0.0 st
KiB Mem: 31371520 total, 30915136 used, 456384 free, 320 buffers
KiB Swap: 6284224 total, 115712 used, 6168512 free. 48192 cached Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
76 root 20 0 0 0 0 R 100.0 0.000 217:17.29 kswapd3
At that time, there are no reclaimable pages left in the node, but as
kswapd fails to restore the high watermarks it refuses to go to sleep.
Kswapd needs to back away from nodes that fail to balance. Up until
commit 1d82de618ddd ("mm, vmscan: make kswapd reclaim in terms of
nodes") kswapd had such a mechanism. It considered zones whose
theoretically reclaimable pages it had reclaimed six times over as
unreclaimable and backed away from them. This guard was erroneously
removed as the patch changed the definition of a balanced node.
However, simply restoring this code wouldn't help in the case reported
here: there *are* no reclaimable pages that could be scanned until the
threshold is met. Kswapd would stay awake anyway.
Introduce a new and much simpler way of backing off. If kswapd runs
through MAX_RECLAIM_RETRIES (16) cycles without reclaiming a single
page, make it back off from the node. This is the same number of shots
direct reclaim takes before declaring OOM. Kswapd will go to sleep on
that node until a direct reclaimer manages to reclaim some pages, thus
proving the node reclaimable again.
[hannes@cmpxchg.org: check kswapd failure against the cumulative nr_reclaimed count]
Link: http://lkml.kernel.org/r/20170306162410.GB2090@cmpxchg.org
[shakeelb@google.com: fix condition for throttle_direct_reclaim]
Link: http://lkml.kernel.org/r/20170314183228.20152-1-shakeelb@google.com
Link: http://lkml.kernel.org/r/20170228214007.5621-2-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Shakeel Butt <shakeelb@google.com>
Reported-by: Jia He <hejianet@gmail.com>
Tested-by: Jia He <hejianet@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-03 21:51:51 +00:00
|
|
|
pgdat->kswapd_failures >= MAX_RECLAIM_RETRIES ?
|
|
|
|
"yes" : "no");
|
2016-07-28 22:45:31 +00:00
|
|
|
}
|
|
|
|
|
2009-03-31 22:19:31 +00:00
|
|
|
for_each_populated_zone(zone) {
|
2005-04-16 22:20:36 +00:00
|
|
|
int i;
|
|
|
|
|
2017-02-22 23:46:16 +00:00
|
|
|
if (show_mem_node_skip(filter, zone_to_nid(zone), nodemask))
|
2011-03-22 23:30:46 +00:00
|
|
|
continue;
|
2015-04-14 22:45:30 +00:00
|
|
|
|
|
|
|
free_pcp = 0;
|
|
|
|
for_each_online_cpu(cpu)
|
|
|
|
free_pcp += per_cpu_ptr(zone->pageset, cpu)->pcp.count;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
show_node(zone);
|
2016-10-28 00:46:29 +00:00
|
|
|
printk(KERN_CONT
|
|
|
|
"%s"
|
2005-04-16 22:20:36 +00:00
|
|
|
" free:%lukB"
|
|
|
|
" min:%lukB"
|
|
|
|
" low:%lukB"
|
|
|
|
" high:%lukB"
|
2019-12-01 01:55:21 +00:00
|
|
|
" reserved_highatomic:%luKB"
|
mm: add per-zone lru list stat
When I did stress test with hackbench, I got OOM message frequently
which didn't ever happen in zone-lru.
gfp_mask=0x26004c0(GFP_KERNEL|__GFP_REPEAT|__GFP_NOTRACK), order=0
..
..
__alloc_pages_nodemask+0xe52/0xe60
? new_slab+0x39c/0x3b0
new_slab+0x39c/0x3b0
___slab_alloc.constprop.87+0x6da/0x840
? __alloc_skb+0x3c/0x260
? _raw_spin_unlock_irq+0x27/0x60
? trace_hardirqs_on_caller+0xec/0x1b0
? finish_task_switch+0xa6/0x220
? poll_select_copy_remaining+0x140/0x140
__slab_alloc.isra.81.constprop.86+0x40/0x6d
? __alloc_skb+0x3c/0x260
kmem_cache_alloc+0x22c/0x260
? __alloc_skb+0x3c/0x260
__alloc_skb+0x3c/0x260
alloc_skb_with_frags+0x4e/0x1a0
sock_alloc_send_pskb+0x16a/0x1b0
? wait_for_unix_gc+0x31/0x90
? alloc_set_pte+0x2ad/0x310
unix_stream_sendmsg+0x28d/0x340
sock_sendmsg+0x2d/0x40
sock_write_iter+0x6c/0xc0
__vfs_write+0xc0/0x120
vfs_write+0x9b/0x1a0
? __might_fault+0x49/0xa0
SyS_write+0x44/0x90
do_fast_syscall_32+0xa6/0x1e0
sysenter_past_esp+0x45/0x74
Mem-Info:
active_anon:104698 inactive_anon:105791 isolated_anon:192
active_file:433 inactive_file:283 isolated_file:22
unevictable:0 dirty:0 writeback:296 unstable:0
slab_reclaimable:6389 slab_unreclaimable:78927
mapped:474 shmem:0 pagetables:101426 bounce:0
free:10518 free_pcp:334 free_cma:0
Node 0 active_anon:418792kB inactive_anon:423164kB active_file:1732kB inactive_file:1132kB unevictable:0kB isolated(anon):768kB isolated(file):88kB mapped:1896kB dirty:0kB writeback:1184kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1478632 all_unreclaimable? yes
DMA free:3304kB min:68kB low:84kB high:100kB present:15992kB managed:15916kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:4088kB kernel_stack:0kB pagetables:2480kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 809 1965 1965
Normal free:3436kB min:3604kB low:4504kB high:5404kB present:897016kB managed:858460kB mlocked:0kB slab_reclaimable:25556kB slab_unreclaimable:311712kB kernel_stack:164608kB pagetables:30844kB bounce:0kB free_pcp:620kB local_pcp:104kB free_cma:0kB
lowmem_reserve[]: 0 0 9247 9247
HighMem free:33808kB min:512kB low:1796kB high:3080kB present:1183736kB managed:1183736kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:372252kB bounce:0kB free_pcp:428kB local_pcp:72kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 2*4kB (UM) 2*8kB (UM) 0*16kB 1*32kB (U) 1*64kB (U) 2*128kB (UM) 1*256kB (U) 1*512kB (M) 0*1024kB 1*2048kB (U) 0*4096kB = 3192kB
Normal: 33*4kB (MH) 79*8kB (ME) 11*16kB (M) 4*32kB (M) 2*64kB (ME) 2*128kB (EH) 7*256kB (EH) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 3244kB
HighMem: 2590*4kB (UM) 1568*8kB (UM) 491*16kB (UM) 60*32kB (UM) 6*64kB (M) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 33064kB
Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=2048kB
25121 total pagecache pages
24160 pages in swap cache
Swap cache stats: add 86371, delete 62211, find 42865/60187
Free swap = 4015560kB
Total swap = 4192252kB
524186 pages RAM
295934 pages HighMem/MovableOnly
9658 pages reserved
0 pages cma reserved
The order-0 allocation for normal zone failed while there are a lot of
reclaimable memory(i.e., anonymous memory with free swap). I wanted to
analyze the problem but it was hard because we removed per-zone lru stat
so I couldn't know how many of anonymous memory there are in normal/dma
zone.
When we investigate OOM problem, reclaimable memory count is crucial
stat to find a problem. Without it, it's hard to parse the OOM message
so I believe we should keep it.
With per-zone lru stat,
gfp_mask=0x26004c0(GFP_KERNEL|__GFP_REPEAT|__GFP_NOTRACK), order=0
Mem-Info:
active_anon:101103 inactive_anon:102219 isolated_anon:0
active_file:503 inactive_file:544 isolated_file:0
unevictable:0 dirty:0 writeback:34 unstable:0
slab_reclaimable:6298 slab_unreclaimable:74669
mapped:863 shmem:0 pagetables:100998 bounce:0
free:23573 free_pcp:1861 free_cma:0
Node 0 active_anon:404412kB inactive_anon:409040kB active_file:2012kB inactive_file:2176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:3452kB dirty:0kB writeback:136kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1320845 all_unreclaimable? yes
DMA free:3296kB min:68kB low:84kB high:100kB active_anon:5540kB inactive_anon:0kB active_file:0kB inactive_file:0kB present:15992kB managed:15916kB mlocked:0kB slab_reclaimable:248kB slab_unreclaimable:2628kB kernel_stack:792kB pagetables:2316kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 809 1965 1965
Normal free:3600kB min:3604kB low:4504kB high:5404kB active_anon:86304kB inactive_anon:0kB active_file:160kB inactive_file:376kB present:897016kB managed:858524kB mlocked:0kB slab_reclaimable:24944kB slab_unreclaimable:296048kB kernel_stack:163832kB pagetables:35892kB bounce:0kB free_pcp:3076kB local_pcp:656kB free_cma:0kB
lowmem_reserve[]: 0 0 9247 9247
HighMem free:86156kB min:512kB low:1796kB high:3080kB active_anon:312852kB inactive_anon:410024kB active_file:1924kB inactive_file:2012kB present:1183736kB managed:1183736kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:365784kB bounce:0kB free_pcp:3868kB local_pcp:720kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 8*4kB (UM) 8*8kB (UM) 4*16kB (M) 2*32kB (UM) 2*64kB (UM) 1*128kB (M) 3*256kB (UME) 2*512kB (UE) 1*1024kB (E) 0*2048kB 0*4096kB = 3296kB
Normal: 240*4kB (UME) 160*8kB (UME) 23*16kB (ME) 3*32kB (UE) 3*64kB (UME) 2*128kB (ME) 1*256kB (U) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 3408kB
HighMem: 10942*4kB (UM) 3102*8kB (UM) 866*16kB (UM) 76*32kB (UM) 11*64kB (UM) 4*128kB (UM) 1*256kB (M) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 86344kB
Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=2048kB
54409 total pagecache pages
53215 pages in swap cache
Swap cache stats: add 300982, delete 247765, find 157978/226539
Free swap = 3803244kB
Total swap = 4192252kB
524186 pages RAM
295934 pages HighMem/MovableOnly
9642 pages reserved
0 pages cma reserved
With that, we can see normal zone has a 86M reclaimable memory so we can
know something goes wrong(I will fix the problem in next patch) in
reclaim.
[mgorman@techsingularity.net: rename zone LRU stats in /proc/vmstat]
Link: http://lkml.kernel.org/r/20160725072300.GK10438@techsingularity.net
Link: http://lkml.kernel.org/r/1469110261-7365-2-git-send-email-mgorman@techsingularity.net
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:47:26 +00:00
|
|
|
" active_anon:%lukB"
|
|
|
|
" inactive_anon:%lukB"
|
|
|
|
" active_file:%lukB"
|
|
|
|
" inactive_file:%lukB"
|
|
|
|
" unevictable:%lukB"
|
2016-07-28 22:47:31 +00:00
|
|
|
" writepending:%lukB"
|
2005-04-16 22:20:36 +00:00
|
|
|
" present:%lukB"
|
2012-12-12 21:52:12 +00:00
|
|
|
" managed:%lukB"
|
2009-09-22 00:01:30 +00:00
|
|
|
" mlocked:%lukB"
|
|
|
|
" bounce:%lukB"
|
2015-04-14 22:45:30 +00:00
|
|
|
" free_pcp:%lukB"
|
|
|
|
" local_pcp:%ukB"
|
2012-10-08 23:32:02 +00:00
|
|
|
" free_cma:%lukB"
|
2005-04-16 22:20:36 +00:00
|
|
|
"\n",
|
|
|
|
zone->name,
|
mm: page allocator: adjust the per-cpu counter threshold when memory is low
Commit aa45484 ("calculate a better estimate of NR_FREE_PAGES when memory
is low") noted that watermarks were based on the vmstat NR_FREE_PAGES. To
avoid synchronization overhead, these counters are maintained on a per-cpu
basis and drained both periodically and when a threshold is above a
threshold. On large CPU systems, the difference between the estimate and
real value of NR_FREE_PAGES can be very high. The system can get into a
case where pages are allocated far below the min watermark potentially
causing livelock issues. The commit solved the problem by taking a better
reading of NR_FREE_PAGES when memory was low.
Unfortately, as reported by Shaohua Li this accurate reading can consume a
large amount of CPU time on systems with many sockets due to cache line
bouncing. This patch takes a different approach. For large machines
where counter drift might be unsafe and while kswapd is awake, the per-cpu
thresholds for the target pgdat are reduced to limit the level of drift to
what should be a safe level. This incurs a performance penalty in heavy
memory pressure by a factor that depends on the workload and the machine
but the machine should function correctly without accidentally exhausting
all memory on a node. There is an additional cost when kswapd wakes and
sleeps but the event is not expected to be frequent - in Shaohua's test
case, there was one recorded sleep and wake event at least.
To ensure that kswapd wakes up, a safe version of zone_watermark_ok() is
introduced that takes a more accurate reading of NR_FREE_PAGES when called
from wakeup_kswapd, when deciding whether it is really safe to go back to
sleep in sleeping_prematurely() and when deciding if a zone is really
balanced or not in balance_pgdat(). We are still using an expensive
function but limiting how often it is called.
When the test case is reproduced, the time spent in the watermark
functions is reduced. The following report is on the percentage of time
spent cumulatively spent in the functions zone_nr_free_pages(),
zone_watermark_ok(), __zone_watermark_ok(), zone_watermark_ok_safe(),
zone_page_state_snapshot(), zone_page_state().
vanilla 11.6615%
disable-threshold 0.2584%
David said:
: We had to pull aa454840 "mm: page allocator: calculate a better estimate
: of NR_FREE_PAGES when memory is low and kswapd is awake" from 2.6.36
: internally because tests showed that it would cause the machine to stall
: as the result of heavy kswapd activity. I merged it back with this fix as
: it is pending in the -mm tree and it solves the issue we were seeing, so I
: definitely think this should be pushed to -stable (and I would seriously
: consider it for 2.6.37 inclusion even at this late date).
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reported-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Tested-by: Nicolas Bareil <nico@chdir.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: <stable@kernel.org> [2.6.37.1, 2.6.36.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:45:41 +00:00
|
|
|
K(zone_page_state(zone, NR_FREE_PAGES)),
|
2009-06-16 22:32:12 +00:00
|
|
|
K(min_wmark_pages(zone)),
|
|
|
|
K(low_wmark_pages(zone)),
|
|
|
|
K(high_wmark_pages(zone)),
|
2019-12-01 01:55:21 +00:00
|
|
|
K(zone->nr_reserved_highatomic),
|
mm: add per-zone lru list stat
When I did stress test with hackbench, I got OOM message frequently
which didn't ever happen in zone-lru.
gfp_mask=0x26004c0(GFP_KERNEL|__GFP_REPEAT|__GFP_NOTRACK), order=0
..
..
__alloc_pages_nodemask+0xe52/0xe60
? new_slab+0x39c/0x3b0
new_slab+0x39c/0x3b0
___slab_alloc.constprop.87+0x6da/0x840
? __alloc_skb+0x3c/0x260
? _raw_spin_unlock_irq+0x27/0x60
? trace_hardirqs_on_caller+0xec/0x1b0
? finish_task_switch+0xa6/0x220
? poll_select_copy_remaining+0x140/0x140
__slab_alloc.isra.81.constprop.86+0x40/0x6d
? __alloc_skb+0x3c/0x260
kmem_cache_alloc+0x22c/0x260
? __alloc_skb+0x3c/0x260
__alloc_skb+0x3c/0x260
alloc_skb_with_frags+0x4e/0x1a0
sock_alloc_send_pskb+0x16a/0x1b0
? wait_for_unix_gc+0x31/0x90
? alloc_set_pte+0x2ad/0x310
unix_stream_sendmsg+0x28d/0x340
sock_sendmsg+0x2d/0x40
sock_write_iter+0x6c/0xc0
__vfs_write+0xc0/0x120
vfs_write+0x9b/0x1a0
? __might_fault+0x49/0xa0
SyS_write+0x44/0x90
do_fast_syscall_32+0xa6/0x1e0
sysenter_past_esp+0x45/0x74
Mem-Info:
active_anon:104698 inactive_anon:105791 isolated_anon:192
active_file:433 inactive_file:283 isolated_file:22
unevictable:0 dirty:0 writeback:296 unstable:0
slab_reclaimable:6389 slab_unreclaimable:78927
mapped:474 shmem:0 pagetables:101426 bounce:0
free:10518 free_pcp:334 free_cma:0
Node 0 active_anon:418792kB inactive_anon:423164kB active_file:1732kB inactive_file:1132kB unevictable:0kB isolated(anon):768kB isolated(file):88kB mapped:1896kB dirty:0kB writeback:1184kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1478632 all_unreclaimable? yes
DMA free:3304kB min:68kB low:84kB high:100kB present:15992kB managed:15916kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:4088kB kernel_stack:0kB pagetables:2480kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 809 1965 1965
Normal free:3436kB min:3604kB low:4504kB high:5404kB present:897016kB managed:858460kB mlocked:0kB slab_reclaimable:25556kB slab_unreclaimable:311712kB kernel_stack:164608kB pagetables:30844kB bounce:0kB free_pcp:620kB local_pcp:104kB free_cma:0kB
lowmem_reserve[]: 0 0 9247 9247
HighMem free:33808kB min:512kB low:1796kB high:3080kB present:1183736kB managed:1183736kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:372252kB bounce:0kB free_pcp:428kB local_pcp:72kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 2*4kB (UM) 2*8kB (UM) 0*16kB 1*32kB (U) 1*64kB (U) 2*128kB (UM) 1*256kB (U) 1*512kB (M) 0*1024kB 1*2048kB (U) 0*4096kB = 3192kB
Normal: 33*4kB (MH) 79*8kB (ME) 11*16kB (M) 4*32kB (M) 2*64kB (ME) 2*128kB (EH) 7*256kB (EH) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 3244kB
HighMem: 2590*4kB (UM) 1568*8kB (UM) 491*16kB (UM) 60*32kB (UM) 6*64kB (M) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 33064kB
Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=2048kB
25121 total pagecache pages
24160 pages in swap cache
Swap cache stats: add 86371, delete 62211, find 42865/60187
Free swap = 4015560kB
Total swap = 4192252kB
524186 pages RAM
295934 pages HighMem/MovableOnly
9658 pages reserved
0 pages cma reserved
The order-0 allocation for normal zone failed while there are a lot of
reclaimable memory(i.e., anonymous memory with free swap). I wanted to
analyze the problem but it was hard because we removed per-zone lru stat
so I couldn't know how many of anonymous memory there are in normal/dma
zone.
When we investigate OOM problem, reclaimable memory count is crucial
stat to find a problem. Without it, it's hard to parse the OOM message
so I believe we should keep it.
With per-zone lru stat,
gfp_mask=0x26004c0(GFP_KERNEL|__GFP_REPEAT|__GFP_NOTRACK), order=0
Mem-Info:
active_anon:101103 inactive_anon:102219 isolated_anon:0
active_file:503 inactive_file:544 isolated_file:0
unevictable:0 dirty:0 writeback:34 unstable:0
slab_reclaimable:6298 slab_unreclaimable:74669
mapped:863 shmem:0 pagetables:100998 bounce:0
free:23573 free_pcp:1861 free_cma:0
Node 0 active_anon:404412kB inactive_anon:409040kB active_file:2012kB inactive_file:2176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:3452kB dirty:0kB writeback:136kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1320845 all_unreclaimable? yes
DMA free:3296kB min:68kB low:84kB high:100kB active_anon:5540kB inactive_anon:0kB active_file:0kB inactive_file:0kB present:15992kB managed:15916kB mlocked:0kB slab_reclaimable:248kB slab_unreclaimable:2628kB kernel_stack:792kB pagetables:2316kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
lowmem_reserve[]: 0 809 1965 1965
Normal free:3600kB min:3604kB low:4504kB high:5404kB active_anon:86304kB inactive_anon:0kB active_file:160kB inactive_file:376kB present:897016kB managed:858524kB mlocked:0kB slab_reclaimable:24944kB slab_unreclaimable:296048kB kernel_stack:163832kB pagetables:35892kB bounce:0kB free_pcp:3076kB local_pcp:656kB free_cma:0kB
lowmem_reserve[]: 0 0 9247 9247
HighMem free:86156kB min:512kB low:1796kB high:3080kB active_anon:312852kB inactive_anon:410024kB active_file:1924kB inactive_file:2012kB present:1183736kB managed:1183736kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:365784kB bounce:0kB free_pcp:3868kB local_pcp:720kB free_cma:0kB
lowmem_reserve[]: 0 0 0 0
DMA: 8*4kB (UM) 8*8kB (UM) 4*16kB (M) 2*32kB (UM) 2*64kB (UM) 1*128kB (M) 3*256kB (UME) 2*512kB (UE) 1*1024kB (E) 0*2048kB 0*4096kB = 3296kB
Normal: 240*4kB (UME) 160*8kB (UME) 23*16kB (ME) 3*32kB (UE) 3*64kB (UME) 2*128kB (ME) 1*256kB (U) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 3408kB
HighMem: 10942*4kB (UM) 3102*8kB (UM) 866*16kB (UM) 76*32kB (UM) 11*64kB (UM) 4*128kB (UM) 1*256kB (M) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 86344kB
Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=2048kB
54409 total pagecache pages
53215 pages in swap cache
Swap cache stats: add 300982, delete 247765, find 157978/226539
Free swap = 3803244kB
Total swap = 4192252kB
524186 pages RAM
295934 pages HighMem/MovableOnly
9642 pages reserved
0 pages cma reserved
With that, we can see normal zone has a 86M reclaimable memory so we can
know something goes wrong(I will fix the problem in next patch) in
reclaim.
[mgorman@techsingularity.net: rename zone LRU stats in /proc/vmstat]
Link: http://lkml.kernel.org/r/20160725072300.GK10438@techsingularity.net
Link: http://lkml.kernel.org/r/1469110261-7365-2-git-send-email-mgorman@techsingularity.net
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:47:26 +00:00
|
|
|
K(zone_page_state(zone, NR_ZONE_ACTIVE_ANON)),
|
|
|
|
K(zone_page_state(zone, NR_ZONE_INACTIVE_ANON)),
|
|
|
|
K(zone_page_state(zone, NR_ZONE_ACTIVE_FILE)),
|
|
|
|
K(zone_page_state(zone, NR_ZONE_INACTIVE_FILE)),
|
|
|
|
K(zone_page_state(zone, NR_ZONE_UNEVICTABLE)),
|
2016-07-28 22:47:31 +00:00
|
|
|
K(zone_page_state(zone, NR_ZONE_WRITE_PENDING)),
|
2005-04-16 22:20:36 +00:00
|
|
|
K(zone->present_pages),
|
2018-12-28 08:34:24 +00:00
|
|
|
K(zone_managed_pages(zone)),
|
2009-09-22 00:01:30 +00:00
|
|
|
K(zone_page_state(zone, NR_MLOCK)),
|
|
|
|
K(zone_page_state(zone, NR_BOUNCE)),
|
2015-04-14 22:45:30 +00:00
|
|
|
K(free_pcp),
|
|
|
|
K(this_cpu_read(zone->pageset->pcp.count)),
|
2016-07-28 22:47:14 +00:00
|
|
|
K(zone_page_state(zone, NR_FREE_CMA_PAGES)));
|
2005-04-16 22:20:36 +00:00
|
|
|
printk("lowmem_reserve[]:");
|
|
|
|
for (i = 0; i < MAX_NR_ZONES; i++)
|
2016-10-28 00:46:29 +00:00
|
|
|
printk(KERN_CONT " %ld", zone->lowmem_reserve[i]);
|
|
|
|
printk(KERN_CONT "\n");
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2009-03-31 22:19:31 +00:00
|
|
|
for_each_populated_zone(zone) {
|
2015-11-07 00:29:57 +00:00
|
|
|
unsigned int order;
|
|
|
|
unsigned long nr[MAX_ORDER], flags, total = 0;
|
2012-12-12 00:00:24 +00:00
|
|
|
unsigned char types[MAX_ORDER];
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2017-02-22 23:46:16 +00:00
|
|
|
if (show_mem_node_skip(filter, zone_to_nid(zone), nodemask))
|
2011-03-22 23:30:46 +00:00
|
|
|
continue;
|
2005-04-16 22:20:36 +00:00
|
|
|
show_node(zone);
|
2016-10-28 00:46:29 +00:00
|
|
|
printk(KERN_CONT "%s: ", zone->name);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
|
|
|
for (order = 0; order < MAX_ORDER; order++) {
|
2012-12-12 00:00:24 +00:00
|
|
|
struct free_area *area = &zone->free_area[order];
|
|
|
|
int type;
|
|
|
|
|
|
|
|
nr[order] = area->nr_free;
|
2006-06-23 09:03:50 +00:00
|
|
|
total += nr[order] << order;
|
2012-12-12 00:00:24 +00:00
|
|
|
|
|
|
|
types[order] = 0;
|
|
|
|
for (type = 0; type < MIGRATE_TYPES; type++) {
|
2019-05-14 22:41:32 +00:00
|
|
|
if (!free_area_empty(area, type))
|
2012-12-12 00:00:24 +00:00
|
|
|
types[order] |= 1 << type;
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
2012-12-12 00:00:24 +00:00
|
|
|
for (order = 0; order < MAX_ORDER; order++) {
|
2016-10-28 00:46:29 +00:00
|
|
|
printk(KERN_CONT "%lu*%lukB ",
|
|
|
|
nr[order], K(1UL) << order);
|
2012-12-12 00:00:24 +00:00
|
|
|
if (nr[order])
|
|
|
|
show_migration_types(types[order]);
|
|
|
|
}
|
2016-10-28 00:46:29 +00:00
|
|
|
printk(KERN_CONT "= %lukB\n", K(total));
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-04-29 22:07:48 +00:00
|
|
|
hugetlb_show_meminfo();
|
|
|
|
|
2016-07-28 22:46:20 +00:00
|
|
|
printk("%ld total pagecache pages\n", global_node_page_state(NR_FILE_PAGES));
|
2008-02-05 06:29:30 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
show_swap_cache_info();
|
|
|
|
}
|
|
|
|
|
2008-04-28 09:12:18 +00:00
|
|
|
static void zoneref_set_zone(struct zone *zone, struct zoneref *zoneref)
|
|
|
|
{
|
|
|
|
zoneref->zone = zone;
|
|
|
|
zoneref->zone_idx = zone_idx(zone);
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Builds allocation fallback zone lists.
|
2006-01-06 08:11:16 +00:00
|
|
|
*
|
|
|
|
* Add all populated zones of a node to the zonelist.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2017-09-06 23:20:30 +00:00
|
|
|
static int build_zonerefs_node(pg_data_t *pgdat, struct zoneref *zonerefs)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-01-06 08:11:16 +00:00
|
|
|
struct zone *zone;
|
2013-07-08 23:00:06 +00:00
|
|
|
enum zone_type zone_type = MAX_NR_ZONES;
|
2017-09-06 23:20:30 +00:00
|
|
|
int nr_zones = 0;
|
2006-01-06 08:11:18 +00:00
|
|
|
|
|
|
|
do {
|
2006-09-26 06:31:18 +00:00
|
|
|
zone_type--;
|
2006-01-06 08:11:19 +00:00
|
|
|
zone = pgdat->node_zones + zone_type;
|
2016-09-01 23:14:55 +00:00
|
|
|
if (managed_zone(zone)) {
|
2017-09-06 23:20:30 +00:00
|
|
|
zoneref_set_zone(zone, &zonerefs[nr_zones++]);
|
2006-01-06 08:11:19 +00:00
|
|
|
check_highest_zone(zone_type);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2006-09-26 06:31:18 +00:00
|
|
|
} while (zone_type);
|
2013-07-08 23:00:06 +00:00
|
|
|
|
2006-01-06 08:11:19 +00:00
|
|
|
return nr_zones;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_NUMA
|
2007-07-16 06:38:01 +00:00
|
|
|
|
|
|
|
static int __parse_numa_zonelist_order(char *s)
|
|
|
|
{
|
2017-09-06 23:20:13 +00:00
|
|
|
/*
|
|
|
|
* We used to support different zonlists modes but they turned
|
|
|
|
* out to be just not useful. Let's keep the warning in place
|
|
|
|
* if somebody still use the cmd line parameter so that we do
|
|
|
|
* not fail it silently
|
|
|
|
*/
|
|
|
|
if (!(*s == 'd' || *s == 'D' || *s == 'n' || *s == 'N')) {
|
|
|
|
pr_warn("Ignoring unsupported numa_zonelist_order value: %s\n", s);
|
2007-07-16 06:38:01 +00:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-09-06 23:20:13 +00:00
|
|
|
char numa_zonelist_order[] = "Node";
|
|
|
|
|
2007-07-16 06:38:01 +00:00
|
|
|
/*
|
|
|
|
* sysctl handler for numa_zonelist_order
|
|
|
|
*/
|
2014-06-06 21:38:09 +00:00
|
|
|
int numa_zonelist_order_handler(struct ctl_table *table, int write,
|
2020-04-24 06:43:38 +00:00
|
|
|
void *buffer, size_t *length, loff_t *ppos)
|
2007-07-16 06:38:01 +00:00
|
|
|
{
|
2020-04-24 06:43:38 +00:00
|
|
|
if (write)
|
|
|
|
return __parse_numa_zonelist_order(buffer);
|
|
|
|
return proc_dostring(table, write, buffer, length, ppos);
|
2007-07-16 06:38:01 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2009-06-16 22:32:15 +00:00
|
|
|
#define MAX_NODE_LOAD (nr_online_nodes)
|
2007-07-16 06:38:01 +00:00
|
|
|
static int node_load[MAX_NUMNODES];
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/**
|
2005-05-01 15:59:25 +00:00
|
|
|
* find_next_best_node - find the next node that should appear in a given node's fallback list
|
2005-04-16 22:20:36 +00:00
|
|
|
* @node: node whose fallback list we're appending
|
|
|
|
* @used_node_mask: nodemask_t of already used nodes
|
|
|
|
*
|
|
|
|
* We use a number of factors to determine which is the next node that should
|
|
|
|
* appear on a given node's fallback list. The node should not have appeared
|
|
|
|
* already in @node's fallback list, and it should be the next closest node
|
|
|
|
* according to the distance array (which contains arbitrary distance values
|
|
|
|
* from each node to each node in the system), and should also prefer nodes
|
|
|
|
* with no CPUs, since presumably they'll have very little allocation pressure
|
|
|
|
* on them otherwise.
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: node id of the found node or %NUMA_NO_NODE if no node is found.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2007-07-16 06:38:01 +00:00
|
|
|
static int find_next_best_node(int node, nodemask_t *used_node_mask)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-02-17 19:38:21 +00:00
|
|
|
int n, val;
|
2005-04-16 22:20:36 +00:00
|
|
|
int min_val = INT_MAX;
|
2013-02-23 00:35:36 +00:00
|
|
|
int best_node = NUMA_NO_NODE;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-02-17 19:38:21 +00:00
|
|
|
/* Use the local node if we haven't already */
|
|
|
|
if (!node_isset(node, *used_node_mask)) {
|
|
|
|
node_set(node, *used_node_mask);
|
|
|
|
return node;
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2012-12-12 21:51:46 +00:00
|
|
|
for_each_node_state(n, N_MEMORY) {
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* Don't want a node to appear more than once */
|
|
|
|
if (node_isset(n, *used_node_mask))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/* Use the distance array to find the distance */
|
|
|
|
val = node_distance(node, n);
|
|
|
|
|
2006-02-17 19:38:21 +00:00
|
|
|
/* Penalize nodes under us ("prefer the next node") */
|
|
|
|
val += (n < node);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* Give preference to headless and unused nodes */
|
2020-10-13 23:55:42 +00:00
|
|
|
if (!cpumask_empty(cpumask_of_node(n)))
|
2005-04-16 22:20:36 +00:00
|
|
|
val += PENALTY_FOR_NODE_WITH_CPUS;
|
|
|
|
|
|
|
|
/* Slight preference for less loaded node */
|
|
|
|
val *= (MAX_NODE_LOAD*MAX_NUMNODES);
|
|
|
|
val += node_load[n];
|
|
|
|
|
|
|
|
if (val < min_val) {
|
|
|
|
min_val = val;
|
|
|
|
best_node = n;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (best_node >= 0)
|
|
|
|
node_set(best_node, *used_node_mask);
|
|
|
|
|
|
|
|
return best_node;
|
|
|
|
}
|
|
|
|
|
2007-07-16 06:38:01 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Build zonelists ordered by node and zones within node.
|
|
|
|
* This results in maximum locality--normal zone overflows into local
|
|
|
|
* DMA zone, if any--but risks exhausting DMA zone.
|
|
|
|
*/
|
2017-09-06 23:20:30 +00:00
|
|
|
static void build_zonelists_in_node_order(pg_data_t *pgdat, int *node_order,
|
|
|
|
unsigned nr_nodes)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2017-09-06 23:20:30 +00:00
|
|
|
struct zoneref *zonerefs;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
zonerefs = pgdat->node_zonelists[ZONELIST_FALLBACK]._zonerefs;
|
|
|
|
|
|
|
|
for (i = 0; i < nr_nodes; i++) {
|
|
|
|
int nr_zones;
|
|
|
|
|
|
|
|
pg_data_t *node = NODE_DATA(node_order[i]);
|
2007-07-16 06:38:01 +00:00
|
|
|
|
2017-09-06 23:20:30 +00:00
|
|
|
nr_zones = build_zonerefs_node(node, zonerefs);
|
|
|
|
zonerefs += nr_zones;
|
|
|
|
}
|
|
|
|
zonerefs->zone = NULL;
|
|
|
|
zonerefs->zone_idx = 0;
|
2007-07-16 06:38:01 +00:00
|
|
|
}
|
|
|
|
|
2007-10-16 08:25:37 +00:00
|
|
|
/*
|
|
|
|
* Build gfp_thisnode zonelists
|
|
|
|
*/
|
|
|
|
static void build_thisnode_zonelists(pg_data_t *pgdat)
|
|
|
|
{
|
2017-09-06 23:20:30 +00:00
|
|
|
struct zoneref *zonerefs;
|
|
|
|
int nr_zones;
|
2007-10-16 08:25:37 +00:00
|
|
|
|
2017-09-06 23:20:30 +00:00
|
|
|
zonerefs = pgdat->node_zonelists[ZONELIST_NOFALLBACK]._zonerefs;
|
|
|
|
nr_zones = build_zonerefs_node(pgdat, zonerefs);
|
|
|
|
zonerefs += nr_zones;
|
|
|
|
zonerefs->zone = NULL;
|
|
|
|
zonerefs->zone_idx = 0;
|
2007-10-16 08:25:37 +00:00
|
|
|
}
|
|
|
|
|
2007-07-16 06:38:01 +00:00
|
|
|
/*
|
|
|
|
* Build zonelists ordered by zone and nodes within zones.
|
|
|
|
* This results in conserving DMA zone[s] until all Normal memory is
|
|
|
|
* exhausted, but results in overflowing to remote node while memory
|
|
|
|
* may still exist in local DMA zone.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void build_zonelists(pg_data_t *pgdat)
|
|
|
|
{
|
2017-09-06 23:20:30 +00:00
|
|
|
static int node_order[MAX_NUMNODES];
|
|
|
|
int node, load, nr_nodes = 0;
|
2020-06-03 22:59:05 +00:00
|
|
|
nodemask_t used_mask = NODE_MASK_NONE;
|
2007-07-16 06:38:01 +00:00
|
|
|
int local_node, prev_node;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* NUMA-aware ordering of nodes */
|
|
|
|
local_node = pgdat->node_id;
|
2009-06-16 22:32:15 +00:00
|
|
|
load = nr_online_nodes;
|
2005-04-16 22:20:36 +00:00
|
|
|
prev_node = local_node;
|
2007-07-16 06:38:01 +00:00
|
|
|
|
|
|
|
memset(node_order, 0, sizeof(node_order));
|
2005-04-16 22:20:36 +00:00
|
|
|
while ((node = find_next_best_node(local_node, &used_mask)) >= 0) {
|
|
|
|
/*
|
|
|
|
* We don't want to pressure a particular node.
|
|
|
|
* So adding penalty to the first node in same
|
|
|
|
* distance group to make it round-robin.
|
|
|
|
*/
|
2012-10-08 23:33:24 +00:00
|
|
|
if (node_distance(local_node, node) !=
|
|
|
|
node_distance(local_node, prev_node))
|
2007-07-16 06:38:01 +00:00
|
|
|
node_load[node] = load;
|
|
|
|
|
2017-09-06 23:20:30 +00:00
|
|
|
node_order[nr_nodes++] = node;
|
2005-04-16 22:20:36 +00:00
|
|
|
prev_node = node;
|
|
|
|
load--;
|
|
|
|
}
|
2007-10-16 08:25:37 +00:00
|
|
|
|
2017-09-06 23:20:30 +00:00
|
|
|
build_zonelists_in_node_order(pgdat, node_order, nr_nodes);
|
2007-10-16 08:25:37 +00:00
|
|
|
build_thisnode_zonelists(pgdat);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2010-05-26 21:45:00 +00:00
|
|
|
#ifdef CONFIG_HAVE_MEMORYLESS_NODES
|
|
|
|
/*
|
|
|
|
* Return node id of node used for "local" allocations.
|
|
|
|
* I.e., first node id of first zone in arg node's generic zonelist.
|
|
|
|
* Used for initializing percpu 'numa_mem', which is used primarily
|
|
|
|
* for kernel allocations, so use GFP_KERNEL flags to locate zonelist.
|
|
|
|
*/
|
|
|
|
int local_memory_node(int node)
|
|
|
|
{
|
2016-05-20 00:14:10 +00:00
|
|
|
struct zoneref *z;
|
2010-05-26 21:45:00 +00:00
|
|
|
|
2016-05-20 00:14:10 +00:00
|
|
|
z = first_zones_zonelist(node_zonelist(node, GFP_KERNEL),
|
2010-05-26 21:45:00 +00:00
|
|
|
gfp_zone(GFP_KERNEL),
|
2016-05-20 00:14:10 +00:00
|
|
|
NULL);
|
2018-08-22 04:53:32 +00:00
|
|
|
return zone_to_nid(z->zone);
|
2010-05-26 21:45:00 +00:00
|
|
|
}
|
|
|
|
#endif
|
2007-07-16 06:38:01 +00:00
|
|
|
|
2016-08-10 23:27:49 +00:00
|
|
|
static void setup_min_unmapped_ratio(void);
|
|
|
|
static void setup_min_slab_ratio(void);
|
2005-04-16 22:20:36 +00:00
|
|
|
#else /* CONFIG_NUMA */
|
|
|
|
|
2007-07-16 06:38:01 +00:00
|
|
|
static void build_zonelists(pg_data_t *pgdat)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-09-26 06:31:19 +00:00
|
|
|
int node, local_node;
|
2017-09-06 23:20:30 +00:00
|
|
|
struct zoneref *zonerefs;
|
|
|
|
int nr_zones;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
local_node = pgdat->node_id;
|
|
|
|
|
2017-09-06 23:20:30 +00:00
|
|
|
zonerefs = pgdat->node_zonelists[ZONELIST_FALLBACK]._zonerefs;
|
|
|
|
nr_zones = build_zonerefs_node(pgdat, zonerefs);
|
|
|
|
zonerefs += nr_zones;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-04-28 09:12:16 +00:00
|
|
|
/*
|
|
|
|
* Now we build the zonelist so that it contains the zones
|
|
|
|
* of all the other nodes.
|
|
|
|
* We don't want to pressure a particular node, so when
|
|
|
|
* building the zones for node N, we make sure that the
|
|
|
|
* zones coming right after the local ones are those from
|
|
|
|
* node N+1 (modulo N)
|
|
|
|
*/
|
|
|
|
for (node = local_node + 1; node < MAX_NUMNODES; node++) {
|
|
|
|
if (!node_online(node))
|
|
|
|
continue;
|
2017-09-06 23:20:30 +00:00
|
|
|
nr_zones = build_zonerefs_node(NODE_DATA(node), zonerefs);
|
|
|
|
zonerefs += nr_zones;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2008-04-28 09:12:16 +00:00
|
|
|
for (node = 0; node < local_node; node++) {
|
|
|
|
if (!node_online(node))
|
|
|
|
continue;
|
2017-09-06 23:20:30 +00:00
|
|
|
nr_zones = build_zonerefs_node(NODE_DATA(node), zonerefs);
|
|
|
|
zonerefs += nr_zones;
|
2008-04-28 09:12:16 +00:00
|
|
|
}
|
|
|
|
|
2017-09-06 23:20:30 +00:00
|
|
|
zonerefs->zone = NULL;
|
|
|
|
zonerefs->zone_idx = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* CONFIG_NUMA */
|
|
|
|
|
2010-01-05 06:34:51 +00:00
|
|
|
/*
|
|
|
|
* Boot pageset table. One per cpu which is going to be used for all
|
|
|
|
* zones and all nodes. The parameters will be set in such a way
|
|
|
|
* that an item put on a list will immediately be handed over to
|
|
|
|
* the buddy list. This is safe since pageset manipulation is done
|
|
|
|
* with interrupts disabled.
|
|
|
|
*
|
|
|
|
* The boot_pagesets must be kept even after bootup is complete for
|
|
|
|
* unused processors and/or zones. They do play a role for bootstrapping
|
|
|
|
* hotplugged processors.
|
|
|
|
*
|
|
|
|
* zoneinfo_show() and maybe other functions do
|
|
|
|
* not check if the processor is online before following the pageset pointer.
|
|
|
|
* Other parts of the kernel may not check if the zone is available.
|
|
|
|
*/
|
2020-12-15 03:10:47 +00:00
|
|
|
static void pageset_init(struct per_cpu_pageset *p);
|
2020-12-15 03:10:53 +00:00
|
|
|
/* These effectively disable the pcplists in the boot pageset completely */
|
|
|
|
#define BOOT_PAGESET_HIGH 0
|
|
|
|
#define BOOT_PAGESET_BATCH 1
|
2010-01-05 06:34:51 +00:00
|
|
|
static DEFINE_PER_CPU(struct per_cpu_pageset, boot_pageset);
|
mm: vmstat: move slab statistics from zone to node counters
Patch series "mm: per-lruvec slab stats"
Josef is working on a new approach to balancing slab caches and the page
cache. For this to work, he needs slab cache statistics on the lruvec
level. These patches implement that by adding infrastructure that
allows updating and reading generic VM stat items per lruvec, then
switches some existing VM accounting sites, including the slab
accounting ones, to this new cgroup-aware API.
I'll follow up with more patches on this, because there is actually
substantial simplification that can be done to the memory controller
when we replace private memcg accounting with making the existing VM
accounting sites cgroup-aware. But this is enough for Josef to base his
slab reclaim work on, so here goes.
This patch (of 5):
To re-implement slab cache vs. page cache balancing, we'll need the
slab counters at the lruvec level, which, ever since lru reclaim was
moved from the zone to the node, is the intersection of the node, not
the zone, and the memcg.
We could retain the per-zone counters for when the page allocator dumps
its memory information on failures, and have counters on both levels -
which on all but NUMA node 0 is usually redundant. But let's keep it
simple for now and just move them. If anybody complains we can restore
the per-zone counters.
[hannes@cmpxchg.org: fix oops]
Link: http://lkml.kernel.org/r/20170605183511.GA8915@cmpxchg.org
Link: http://lkml.kernel.org/r/20170530181724.27197-3-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-06 22:40:43 +00:00
|
|
|
static DEFINE_PER_CPU(struct per_cpu_nodestat, boot_nodestats);
|
2010-01-05 06:34:51 +00:00
|
|
|
|
mm, page_alloc: remove stop_machine from build_all_zonelists
build_all_zonelists has been (ab)using stop_machine to make sure that
zonelists do not change while somebody is looking at them. This is is
just a gross hack because a) it complicates the context from which we
can call build_all_zonelists (see 3f906ba23689 ("mm/memory-hotplug:
switch locking to a percpu rwsem")) and b) is is not really necessary
especially after "mm, page_alloc: simplify zonelist initialization" and
c) it doesn't really provide the protection it claims (see below).
Updates of the zonelists happen very seldom, basically only when a zone
becomes populated during memory online or when it loses all the memory
during offline. A racing iteration over zonelists could either miss a
zone or try to work on one zone twice. Both of these are something we
can live with occasionally because there will always be at least one
zone visible so we are not likely to fail allocation too easily for
example.
Please note that the original stop_machine approach doesn't really
provide a better exclusion because the iteration might be interrupted
half way (unless the whole iteration is preempt disabled which is not
the case in most cases) so the some zones could still be seen twice or a
zone missed.
I have run the pathological online/offline of the single memblock in the
movable zone while stressing the same small node with some memory
pressure.
Node 1, zone DMA
pages free 0
min 0
low 0
high 0
spanned 0
present 0
managed 0
protection: (0, 943, 943, 943)
Node 1, zone DMA32
pages free 227310
min 8294
low 10367
high 12440
spanned 262112
present 262112
managed 241436
protection: (0, 0, 0, 0)
Node 1, zone Normal
pages free 0
min 0
low 0
high 0
spanned 0
present 0
managed 0
protection: (0, 0, 0, 1024)
Node 1, zone Movable
pages free 32722
min 85
low 117
high 149
spanned 32768
present 32768
managed 32768
protection: (0, 0, 0, 0)
root@test1:/sys/devices/system/node/node1# while true
do
echo offline > memory34/state
echo online_movable > memory34/state
done
root@test1:/mnt/data/test/linux-3.7-rc5# numactl --preferred=1 make -j4
and it survived without any unexpected behavior. While this is not
really a great testing coverage it should exercise the allocation path
quite a lot.
Link: http://lkml.kernel.org/r/20170721143915.14161-8-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:20:34 +00:00
|
|
|
static void __build_all_zonelists(void *data)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-06-23 09:03:11 +00:00
|
|
|
int nid;
|
2017-09-06 23:20:17 +00:00
|
|
|
int __maybe_unused cpu;
|
2012-07-31 23:43:28 +00:00
|
|
|
pg_data_t *self = data;
|
2017-09-06 23:20:37 +00:00
|
|
|
static DEFINE_SPINLOCK(lock);
|
|
|
|
|
|
|
|
spin_lock(&lock);
|
[PATCH] memory page_alloc zonelist caching speedup
Optimize the critical zonelist scanning for free pages in the kernel memory
allocator by caching the zones that were found to be full recently, and
skipping them.
Remembers the zones in a zonelist that were short of free memory in the
last second. And it stashes a zone-to-node table in the zonelist struct,
to optimize that conversion (minimize its cache footprint.)
Recent changes:
This differs in a significant way from a similar patch that I
posted a week ago. Now, instead of having a nodemask_t of
recently full nodes, I have a bitmask of recently full zones.
This solves a problem that last weeks patch had, which on
systems with multiple zones per node (such as DMA zone) would
take seeing any of these zones full as meaning that all zones
on that node were full.
Also I changed names - from "zonelist faster" to "zonelist cache",
as that seemed to better convey what we're doing here - caching
some of the key zonelist state (for faster access.)
See below for some performance benchmark results. After all that
discussion with David on why I didn't need them, I went and got
some ;). I wanted to verify that I had not hurt the normal case
of memory allocation noticeably. At least for my one little
microbenchmark, I found (1) the normal case wasn't affected, and
(2) workloads that forced scanning across multiple nodes for
memory improved up to 10% fewer System CPU cycles and lower
elapsed clock time ('sys' and 'real'). Good. See details, below.
I didn't have the logic in get_page_from_freelist() for various
full nodes and zone reclaim failures correct. That should be
fixed up now - notice the new goto labels zonelist_scan,
this_zone_full, and try_next_zone, in get_page_from_freelist().
There are two reasons I persued this alternative, over some earlier
proposals that would have focused on optimizing the fake numa
emulation case by caching the last useful zone:
1) Contrary to what I said before, we (SGI, on large ia64 sn2 systems)
have seen real customer loads where the cost to scan the zonelist
was a problem, due to many nodes being full of memory before
we got to a node we could use. Or at least, I think we have.
This was related to me by another engineer, based on experiences
from some time past. So this is not guaranteed. Most likely, though.
The following approach should help such real numa systems just as
much as it helps fake numa systems, or any combination thereof.
2) The effort to distinguish fake from real numa, using node_distance,
so that we could cache a fake numa node and optimize choosing
it over equivalent distance fake nodes, while continuing to
properly scan all real nodes in distance order, was going to
require a nasty blob of zonelist and node distance munging.
The following approach has no new dependency on node distances or
zone sorting.
See comment in the patch below for a description of what it actually does.
Technical details of note (or controversy):
- See the use of "zlc_active" and "did_zlc_setup" below, to delay
adding any work for this new mechanism until we've looked at the
first zone in zonelist. I figured the odds of the first zone
having the memory we needed were high enough that we should just
look there, first, then get fancy only if we need to keep looking.
- Some odd hackery was needed to add items to struct zonelist, while
not tripping up the custom zonelists built by the mm/mempolicy.c
code for MPOL_BIND. My usual wordy comments below explain this.
Search for "MPOL_BIND".
- Some per-node data in the struct zonelist is now modified frequently,
with no locking. Multiple CPU cores on a node could hit and mangle
this data. The theory is that this is just performance hint data,
and the memory allocator will work just fine despite any such mangling.
The fields at risk are the struct 'zonelist_cache' fields 'fullzones'
(a bitmask) and 'last_full_zap' (unsigned long jiffies). It should
all be self correcting after at most a one second delay.
- This still does a linear scan of the same lengths as before. All
I've optimized is making the scan faster, not algorithmically
shorter. It is now able to scan a compact array of 'unsigned
short' in the case of many full nodes, so one cache line should
cover quite a few nodes, rather than each node hitting another
one or two new and distinct cache lines.
- If both Andi and Nick don't find this too complicated, I will be
(pleasantly) flabbergasted.
- I removed the comment claiming we only use one cachline's worth of
zonelist. We seem, at least in the fake numa case, to have put the
lie to that claim.
- I pay no attention to the various watermarks and such in this performance
hint. A node could be marked full for one watermark, and then skipped
over when searching for a page using a different watermark. I think
that's actually quite ok, as it will tend to slightly increase the
spreading of memory over other nodes, away from a memory stressed node.
===============
Performance - some benchmark results and analysis:
This benchmark runs a memory hog program that uses multiple
threads to touch alot of memory as quickly as it can.
Multiple runs were made, touching 12, 38, 64 or 90 GBytes out of
the total 96 GBytes on the system, and using 1, 19, 37, or 55
threads (on a 56 CPU system.) System, user and real (elapsed)
timings were recorded for each run, shown in units of seconds,
in the table below.
Two kernels were tested - 2.6.18-mm3 and the same kernel with
this zonelist caching patch added. The table also shows the
percentage improvement the zonelist caching sys time is over
(lower than) the stock *-mm kernel.
number 2.6.18-mm3 zonelist-cache delta (< 0 good) percent
GBs N ------------ -------------- ---------------- systime
mem threads sys user real sys user real sys user real better
12 1 153 24 177 151 24 176 -2 0 -1 1%
12 19 99 22 8 99 22 8 0 0 0 0%
12 37 111 25 6 112 25 6 1 0 0 -0%
12 55 115 25 5 110 23 5 -5 -2 0 4%
38 1 502 74 576 497 73 570 -5 -1 -6 0%
38 19 426 78 48 373 76 39 -53 -2 -9 12%
38 37 544 83 36 547 82 36 3 -1 0 -0%
38 55 501 77 23 511 80 24 10 3 1 -1%
64 1 917 125 1042 890 124 1014 -27 -1 -28 2%
64 19 1118 138 119 965 141 103 -153 3 -16 13%
64 37 1202 151 94 1136 150 81 -66 -1 -13 5%
64 55 1118 141 61 1072 140 58 -46 -1 -3 4%
90 1 1342 177 1519 1275 174 1450 -67 -3 -69 4%
90 19 2392 199 192 2116 189 176 -276 -10 -16 11%
90 37 3313 238 175 2972 225 145 -341 -13 -30 10%
90 55 1948 210 104 1843 213 100 -105 3 -4 5%
Notes:
1) This test ran a memory hog program that started a specified number N of
threads, and had each thread allocate and touch 1/N'th of
the total memory to be used in the test run in a single loop,
writing a constant word to memory, one store every 4096 bytes.
Watching this test during some earlier trial runs, I would see
each of these threads sit down on one CPU and stay there, for
the remainder of the pass, a different CPU for each thread.
2) The 'real' column is not comparable to the 'sys' or 'user' columns.
The 'real' column is seconds wall clock time elapsed, from beginning
to end of that test pass. The 'sys' and 'user' columns are total
CPU seconds spent on that test pass. For a 19 thread test run,
for example, the sum of 'sys' and 'user' could be up to 19 times the
number of 'real' elapsed wall clock seconds.
3) Tests were run on a fresh, single-user boot, to minimize the amount
of memory already in use at the start of the test, and to minimize
the amount of background activity that might interfere.
4) Tests were done on a 56 CPU, 28 Node system with 96 GBytes of RAM.
5) Notice that the 'real' time gets large for the single thread runs, even
though the measured 'sys' and 'user' times are modest. I'm not sure what
that means - probably something to do with it being slow for one thread to
be accessing memory along ways away. Perhaps the fake numa system, running
ostensibly the same workload, would not show this substantial degradation
of 'real' time for one thread on many nodes -- lets hope not.
6) The high thread count passes (one thread per CPU - on 55 of 56 CPUs)
ran quite efficiently, as one might expect. Each pair of threads needed
to allocate and touch the memory on the node the two threads shared, a
pleasantly parallizable workload.
7) The intermediate thread count passes, when asking for alot of memory forcing
them to go to a few neighboring nodes, improved the most with this zonelist
caching patch.
Conclusions:
* This zonelist cache patch probably makes little difference one way or the
other for most workloads on real numa hardware, if those workloads avoid
heavy off node allocations.
* For memory intensive workloads requiring substantial off-node allocations
on real numa hardware, this patch improves both kernel and elapsed timings
up to ten per-cent.
* For fake numa systems, I'm optimistic, but will have to leave that up to
Rohit Seth to actually test (once I get him a 2.6.18 backport.)
Signed-off-by: Paul Jackson <pj@sgi.com>
Cc: Rohit Seth <rohitseth@google.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Cc: David Rientjes <rientjes@cs.washington.edu>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:31:48 +00:00
|
|
|
|
2009-08-18 21:11:19 +00:00
|
|
|
#ifdef CONFIG_NUMA
|
|
|
|
memset(node_load, 0, sizeof(node_load));
|
|
|
|
#endif
|
2012-07-31 23:43:28 +00:00
|
|
|
|
2017-09-06 23:19:33 +00:00
|
|
|
/*
|
|
|
|
* This node is hotadded and no memory is yet present. So just
|
|
|
|
* building zonelists is fine - no need to touch other nodes.
|
|
|
|
*/
|
2012-07-31 23:43:28 +00:00
|
|
|
if (self && !node_online(self->node_id)) {
|
|
|
|
build_zonelists(self);
|
2017-09-06 23:19:33 +00:00
|
|
|
} else {
|
|
|
|
for_each_online_node(nid) {
|
|
|
|
pg_data_t *pgdat = NODE_DATA(nid);
|
2007-10-16 08:25:29 +00:00
|
|
|
|
2017-09-06 23:19:33 +00:00
|
|
|
build_zonelists(pgdat);
|
|
|
|
}
|
2010-01-05 06:34:51 +00:00
|
|
|
|
2010-05-26 21:45:00 +00:00
|
|
|
#ifdef CONFIG_HAVE_MEMORYLESS_NODES
|
|
|
|
/*
|
|
|
|
* We now know the "local memory node" for each node--
|
|
|
|
* i.e., the node of the first zone in the generic zonelist.
|
|
|
|
* Set up numa_mem percpu variable for on-line cpus. During
|
|
|
|
* boot, only the boot cpu should be on-line; we'll init the
|
|
|
|
* secondary cpus' numa_mem as they come on-line. During
|
|
|
|
* node/memory hotplug, we'll fixup all on-line cpus.
|
|
|
|
*/
|
2017-09-06 23:20:20 +00:00
|
|
|
for_each_online_cpu(cpu)
|
2010-05-26 21:45:00 +00:00
|
|
|
set_cpu_numa_mem(cpu, local_memory_node(cpu_to_node(cpu)));
|
2017-09-06 23:20:17 +00:00
|
|
|
#endif
|
2017-09-06 23:20:20 +00:00
|
|
|
}
|
2017-09-06 23:20:37 +00:00
|
|
|
|
|
|
|
spin_unlock(&lock);
|
2006-06-23 09:03:11 +00:00
|
|
|
}
|
|
|
|
|
2015-02-12 23:00:06 +00:00
|
|
|
static noinline void __init
|
|
|
|
build_all_zonelists_init(void)
|
|
|
|
{
|
2017-09-06 23:20:17 +00:00
|
|
|
int cpu;
|
|
|
|
|
2015-02-12 23:00:06 +00:00
|
|
|
__build_all_zonelists(NULL);
|
2017-09-06 23:20:17 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Initialize the boot_pagesets that are going to be used
|
|
|
|
* for bootstrapping processors. The real pagesets for
|
|
|
|
* each zone will be allocated later when the per cpu
|
|
|
|
* allocator is available.
|
|
|
|
*
|
|
|
|
* boot_pagesets are used also for bootstrapping offline
|
|
|
|
* cpus if the system is already booted because the pagesets
|
|
|
|
* are needed to initialize allocators on a specific cpu too.
|
|
|
|
* F.e. the percpu allocator needs the page allocator which
|
|
|
|
* needs the percpu allocator in order to allocate its pagesets
|
|
|
|
* (a chicken-egg dilemma).
|
|
|
|
*/
|
|
|
|
for_each_possible_cpu(cpu)
|
2020-12-15 03:10:47 +00:00
|
|
|
pageset_init(&per_cpu(boot_pageset, cpu));
|
2017-09-06 23:20:17 +00:00
|
|
|
|
2015-02-12 23:00:06 +00:00
|
|
|
mminit_verify_zonelist();
|
|
|
|
cpuset_init_current_mems_allowed();
|
|
|
|
}
|
|
|
|
|
2010-05-24 21:32:52 +00:00
|
|
|
/*
|
|
|
|
* unless system_state == SYSTEM_BOOTING.
|
2015-02-12 23:00:06 +00:00
|
|
|
*
|
2017-09-06 23:20:24 +00:00
|
|
|
* __ref due to call of __init annotated helper build_all_zonelists_init
|
2015-02-12 23:00:06 +00:00
|
|
|
* [protected by SYSTEM_BOOTING].
|
2010-05-24 21:32:52 +00:00
|
|
|
*/
|
2017-09-06 23:20:24 +00:00
|
|
|
void __ref build_all_zonelists(pg_data_t *pgdat)
|
2006-06-23 09:03:11 +00:00
|
|
|
{
|
2020-08-07 06:25:27 +00:00
|
|
|
unsigned long vm_total_pages;
|
|
|
|
|
2006-06-23 09:03:11 +00:00
|
|
|
if (system_state == SYSTEM_BOOTING) {
|
2015-02-12 23:00:06 +00:00
|
|
|
build_all_zonelists_init();
|
2006-06-23 09:03:11 +00:00
|
|
|
} else {
|
mm, page_alloc: remove stop_machine from build_all_zonelists
build_all_zonelists has been (ab)using stop_machine to make sure that
zonelists do not change while somebody is looking at them. This is is
just a gross hack because a) it complicates the context from which we
can call build_all_zonelists (see 3f906ba23689 ("mm/memory-hotplug:
switch locking to a percpu rwsem")) and b) is is not really necessary
especially after "mm, page_alloc: simplify zonelist initialization" and
c) it doesn't really provide the protection it claims (see below).
Updates of the zonelists happen very seldom, basically only when a zone
becomes populated during memory online or when it loses all the memory
during offline. A racing iteration over zonelists could either miss a
zone or try to work on one zone twice. Both of these are something we
can live with occasionally because there will always be at least one
zone visible so we are not likely to fail allocation too easily for
example.
Please note that the original stop_machine approach doesn't really
provide a better exclusion because the iteration might be interrupted
half way (unless the whole iteration is preempt disabled which is not
the case in most cases) so the some zones could still be seen twice or a
zone missed.
I have run the pathological online/offline of the single memblock in the
movable zone while stressing the same small node with some memory
pressure.
Node 1, zone DMA
pages free 0
min 0
low 0
high 0
spanned 0
present 0
managed 0
protection: (0, 943, 943, 943)
Node 1, zone DMA32
pages free 227310
min 8294
low 10367
high 12440
spanned 262112
present 262112
managed 241436
protection: (0, 0, 0, 0)
Node 1, zone Normal
pages free 0
min 0
low 0
high 0
spanned 0
present 0
managed 0
protection: (0, 0, 0, 1024)
Node 1, zone Movable
pages free 32722
min 85
low 117
high 149
spanned 32768
present 32768
managed 32768
protection: (0, 0, 0, 0)
root@test1:/sys/devices/system/node/node1# while true
do
echo offline > memory34/state
echo online_movable > memory34/state
done
root@test1:/mnt/data/test/linux-3.7-rc5# numactl --preferred=1 make -j4
and it survived without any unexpected behavior. While this is not
really a great testing coverage it should exercise the allocation path
quite a lot.
Link: http://lkml.kernel.org/r/20170721143915.14161-8-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:20:34 +00:00
|
|
|
__build_all_zonelists(pgdat);
|
2006-06-23 09:03:11 +00:00
|
|
|
/* cpuset refresh routine should be here */
|
|
|
|
}
|
2020-08-07 06:25:30 +00:00
|
|
|
/* Get the number of free pages beyond high watermark in all zones. */
|
|
|
|
vm_total_pages = nr_free_zone_pages(gfp_zone(GFP_HIGHUSER_MOVABLE));
|
2007-10-16 08:25:54 +00:00
|
|
|
/*
|
|
|
|
* Disable grouping by mobility if the number of pages in the
|
|
|
|
* system is too low to allow the mechanism to work. It would be
|
|
|
|
* more accurate, but expensive to check per-zone. This check is
|
|
|
|
* made on memory-hotadd so a system can start with mobility
|
|
|
|
* disabled and enable it later
|
|
|
|
*/
|
2007-10-16 08:26:01 +00:00
|
|
|
if (vm_total_pages < (pageblock_nr_pages * MIGRATE_TYPES))
|
2007-10-16 08:25:54 +00:00
|
|
|
page_group_by_mobility_disabled = 1;
|
|
|
|
else
|
|
|
|
page_group_by_mobility_disabled = 0;
|
|
|
|
|
2019-03-05 23:48:29 +00:00
|
|
|
pr_info("Built %u zonelists, mobility grouping %s. Total pages: %ld\n",
|
2016-03-17 21:19:47 +00:00
|
|
|
nr_online_nodes,
|
|
|
|
page_group_by_mobility_disabled ? "off" : "on",
|
|
|
|
vm_total_pages);
|
2007-07-16 06:38:01 +00:00
|
|
|
#ifdef CONFIG_NUMA
|
2014-12-10 23:42:53 +00:00
|
|
|
pr_info("Policy zone: %s\n", zone_names[policy_zone]);
|
2007-07-16 06:38:01 +00:00
|
|
|
#endif
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2018-10-26 22:09:40 +00:00
|
|
|
/* If zone is ZONE_MOVABLE but memory is mirrored, it is an overlapped init */
|
|
|
|
static bool __meminit
|
|
|
|
overlap_memmap_init(unsigned long zone, unsigned long *pfn)
|
|
|
|
{
|
|
|
|
static struct memblock_region *r;
|
|
|
|
|
|
|
|
if (mirrored_kernelcore && zone == ZONE_MOVABLE) {
|
|
|
|
if (!r || *pfn >= memblock_region_memory_end_pfn(r)) {
|
2020-10-13 23:58:30 +00:00
|
|
|
for_each_mem_region(r) {
|
2018-10-26 22:09:40 +00:00
|
|
|
if (*pfn < memblock_region_memory_end_pfn(r))
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (*pfn >= memblock_region_memory_base_pfn(r) &&
|
|
|
|
memblock_is_mirror(r)) {
|
|
|
|
*pfn = memblock_region_memory_end_pfn(r);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Initially all pages are reserved - free ones are freed
|
2018-10-30 22:09:30 +00:00
|
|
|
* up by memblock_free_all() once the early boot process is
|
2005-04-16 22:20:36 +00:00
|
|
|
* done. Non-atomic initialization, single-pass.
|
2020-10-16 03:08:19 +00:00
|
|
|
*
|
|
|
|
* All aligned pageblocks are initialized to the specified migratetype
|
|
|
|
* (usually MIGRATE_MOVABLE). Besides setting the migratetype, no related
|
|
|
|
* zone stats (e.g., nr_isolate_pageblock) are touched.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2021-02-24 20:06:14 +00:00
|
|
|
void __meminit memmap_init_range(unsigned long size, int nid, unsigned long zone,
|
mm: memmap defer init doesn't work as expected
VMware observed a performance regression during memmap init on their
platform, and bisected to commit 73a6e474cb376 ("mm: memmap_init:
iterate over memblock regions rather that check each PFN") causing it.
Before the commit:
[0.033176] Normal zone: 1445888 pages used for memmap
[0.033176] Normal zone: 89391104 pages, LIFO batch:63
[0.035851] ACPI: PM-Timer IO Port: 0x448
With commit
[0.026874] Normal zone: 1445888 pages used for memmap
[0.026875] Normal zone: 89391104 pages, LIFO batch:63
[2.028450] ACPI: PM-Timer IO Port: 0x448
The root cause is the current memmap defer init doesn't work as expected.
Before, memmap_init_zone() was used to do memmap init of one whole zone,
to initialize all low zones of one numa node, but defer memmap init of
the last zone in that numa node. However, since commit 73a6e474cb376,
function memmap_init() is adapted to iterater over memblock regions
inside one zone, then call memmap_init_zone() to do memmap init for each
region.
E.g, on VMware's system, the memory layout is as below, there are two
memory regions in node 2. The current code will mistakenly initialize the
whole 1st region [mem 0xab00000000-0xfcffffffff], then do memmap defer to
iniatialize only one memmory section on the 2nd region [mem
0x10000000000-0x1033fffffff]. In fact, we only expect to see that there's
only one memory section's memmap initialized. That's why more time is
costed at the time.
[ 0.008842] ACPI: SRAT: Node 0 PXM 0 [mem 0x00000000-0x0009ffff]
[ 0.008842] ACPI: SRAT: Node 0 PXM 0 [mem 0x00100000-0xbfffffff]
[ 0.008843] ACPI: SRAT: Node 0 PXM 0 [mem 0x100000000-0x55ffffffff]
[ 0.008844] ACPI: SRAT: Node 1 PXM 1 [mem 0x5600000000-0xaaffffffff]
[ 0.008844] ACPI: SRAT: Node 2 PXM 2 [mem 0xab00000000-0xfcffffffff]
[ 0.008845] ACPI: SRAT: Node 2 PXM 2 [mem 0x10000000000-0x1033fffffff]
Now, let's add a parameter 'zone_end_pfn' to memmap_init_zone() to pass
down the real zone end pfn so that defer_init() can use it to judge
whether defer need be taken in zone wide.
Link: https://lkml.kernel.org/r/20201223080811.16211-1-bhe@redhat.com
Link: https://lkml.kernel.org/r/20201223080811.16211-2-bhe@redhat.com
Fixes: commit 73a6e474cb376 ("mm: memmap_init: iterate over memblock regions rather that check each PFN")
Signed-off-by: Baoquan He <bhe@redhat.com>
Reported-by: Rahul Gopakumar <gopakumarr@vmware.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-29 23:14:37 +00:00
|
|
|
unsigned long start_pfn, unsigned long zone_end_pfn,
|
2020-10-16 03:08:19 +00:00
|
|
|
enum meminit_context context,
|
|
|
|
struct vmem_altmap *altmap, int migratetype)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2018-10-26 22:09:40 +00:00
|
|
|
unsigned long pfn, end_pfn = start_pfn + size;
|
2018-04-05 23:23:00 +00:00
|
|
|
struct page *page;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2009-01-06 22:40:09 +00:00
|
|
|
if (highest_memmap_pfn < end_pfn - 1)
|
|
|
|
highest_memmap_pfn = end_pfn - 1;
|
|
|
|
|
2018-10-26 22:07:52 +00:00
|
|
|
#ifdef CONFIG_ZONE_DEVICE
|
2016-01-16 00:56:22 +00:00
|
|
|
/*
|
|
|
|
* Honor reservation requested by the driver for this ZONE_DEVICE
|
2018-10-26 22:07:52 +00:00
|
|
|
* memory. We limit the total number of pages to initialize to just
|
|
|
|
* those that might contain the memory mapping. We will defer the
|
|
|
|
* ZONE_DEVICE page initialization until after we have released
|
|
|
|
* the hotplug lock.
|
2016-01-16 00:56:22 +00:00
|
|
|
*/
|
2018-10-26 22:07:52 +00:00
|
|
|
if (zone == ZONE_DEVICE) {
|
|
|
|
if (!altmap)
|
|
|
|
return;
|
|
|
|
|
|
|
|
if (start_pfn == altmap->base_pfn)
|
|
|
|
start_pfn += altmap->reserve;
|
|
|
|
end_pfn = altmap->base_pfn + vmem_altmap_offset(altmap);
|
|
|
|
}
|
|
|
|
#endif
|
2016-01-16 00:56:22 +00:00
|
|
|
|
2020-02-04 01:33:59 +00:00
|
|
|
for (pfn = start_pfn; pfn < end_pfn; ) {
|
2007-01-11 07:15:30 +00:00
|
|
|
/*
|
2016-03-15 21:55:25 +00:00
|
|
|
* There can be holes in boot-time mem_map[]s handed to this
|
|
|
|
* function. They do not exist on hotplugged memory.
|
2007-01-11 07:15:30 +00:00
|
|
|
*/
|
2020-09-26 04:19:28 +00:00
|
|
|
if (context == MEMINIT_EARLY) {
|
2018-10-26 22:09:40 +00:00
|
|
|
if (overlap_memmap_init(zone, &pfn))
|
|
|
|
continue;
|
mm: memmap defer init doesn't work as expected
VMware observed a performance regression during memmap init on their
platform, and bisected to commit 73a6e474cb376 ("mm: memmap_init:
iterate over memblock regions rather that check each PFN") causing it.
Before the commit:
[0.033176] Normal zone: 1445888 pages used for memmap
[0.033176] Normal zone: 89391104 pages, LIFO batch:63
[0.035851] ACPI: PM-Timer IO Port: 0x448
With commit
[0.026874] Normal zone: 1445888 pages used for memmap
[0.026875] Normal zone: 89391104 pages, LIFO batch:63
[2.028450] ACPI: PM-Timer IO Port: 0x448
The root cause is the current memmap defer init doesn't work as expected.
Before, memmap_init_zone() was used to do memmap init of one whole zone,
to initialize all low zones of one numa node, but defer memmap init of
the last zone in that numa node. However, since commit 73a6e474cb376,
function memmap_init() is adapted to iterater over memblock regions
inside one zone, then call memmap_init_zone() to do memmap init for each
region.
E.g, on VMware's system, the memory layout is as below, there are two
memory regions in node 2. The current code will mistakenly initialize the
whole 1st region [mem 0xab00000000-0xfcffffffff], then do memmap defer to
iniatialize only one memmory section on the 2nd region [mem
0x10000000000-0x1033fffffff]. In fact, we only expect to see that there's
only one memory section's memmap initialized. That's why more time is
costed at the time.
[ 0.008842] ACPI: SRAT: Node 0 PXM 0 [mem 0x00000000-0x0009ffff]
[ 0.008842] ACPI: SRAT: Node 0 PXM 0 [mem 0x00100000-0xbfffffff]
[ 0.008843] ACPI: SRAT: Node 0 PXM 0 [mem 0x100000000-0x55ffffffff]
[ 0.008844] ACPI: SRAT: Node 1 PXM 1 [mem 0x5600000000-0xaaffffffff]
[ 0.008844] ACPI: SRAT: Node 2 PXM 2 [mem 0xab00000000-0xfcffffffff]
[ 0.008845] ACPI: SRAT: Node 2 PXM 2 [mem 0x10000000000-0x1033fffffff]
Now, let's add a parameter 'zone_end_pfn' to memmap_init_zone() to pass
down the real zone end pfn so that defer_init() can use it to judge
whether defer need be taken in zone wide.
Link: https://lkml.kernel.org/r/20201223080811.16211-1-bhe@redhat.com
Link: https://lkml.kernel.org/r/20201223080811.16211-2-bhe@redhat.com
Fixes: commit 73a6e474cb376 ("mm: memmap_init: iterate over memblock regions rather that check each PFN")
Signed-off-by: Baoquan He <bhe@redhat.com>
Reported-by: Rahul Gopakumar <gopakumarr@vmware.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-29 23:14:37 +00:00
|
|
|
if (defer_init(nid, pfn, zone_end_pfn))
|
2018-10-26 22:09:40 +00:00
|
|
|
break;
|
2007-01-11 07:15:30 +00:00
|
|
|
}
|
2015-06-30 21:57:20 +00:00
|
|
|
|
2018-04-05 23:23:00 +00:00
|
|
|
page = pfn_to_page(pfn);
|
|
|
|
__init_single_page(page, pfn, zone, nid);
|
2020-09-26 04:19:28 +00:00
|
|
|
if (context == MEMINIT_HOTPLUG)
|
2018-10-26 22:07:48 +00:00
|
|
|
__SetPageReserved(page);
|
2018-04-05 23:23:00 +00:00
|
|
|
|
2015-06-30 21:57:20 +00:00
|
|
|
/*
|
2020-10-16 03:08:19 +00:00
|
|
|
* Usually, we want to mark the pageblock MIGRATE_MOVABLE,
|
|
|
|
* such that unmovable allocations won't be scattered all
|
|
|
|
* over the place during system boot.
|
2015-06-30 21:57:20 +00:00
|
|
|
*/
|
2020-10-16 03:08:15 +00:00
|
|
|
if (IS_ALIGNED(pfn, pageblock_nr_pages)) {
|
2020-10-16 03:08:19 +00:00
|
|
|
set_pageblock_migratetype(page, migratetype);
|
2017-10-03 23:16:19 +00:00
|
|
|
cond_resched();
|
2015-06-30 21:57:20 +00:00
|
|
|
}
|
2020-02-04 01:33:59 +00:00
|
|
|
pfn++;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-10-26 22:07:52 +00:00
|
|
|
#ifdef CONFIG_ZONE_DEVICE
|
|
|
|
void __ref memmap_init_zone_device(struct zone *zone,
|
|
|
|
unsigned long start_pfn,
|
mm/memmap_init: update variable name in memmap_init_zone
Patch series "mm/memory_hotplug: Shrink zones before removing memory", v6.
This series fixes the access of uninitialized memmaps when shrinking
zones/nodes and when removing memory. Also, it contains all fixes for
crashes that can be triggered when removing certain namespace using
memunmap_pages() - ZONE_DEVICE, reported by Aneesh.
We stop trying to shrink ZONE_DEVICE, as it's buggy, fixing it would be
more involved (we don't have SECTION_IS_ONLINE as an indicator), and
shrinking is only of limited use (set_zone_contiguous() cannot detect the
ZONE_DEVICE as contiguous).
We continue shrinking !ZONE_DEVICE zones, however, I reduced the amount of
code to a minimum. Shrinking is especially necessary to keep
zone->contiguous set where possible, especially, on memory unplug of DIMMs
at zone boundaries.
--------------------------------------------------------------------------
Zones are now properly shrunk when offlining memory blocks or when
onlining failed. This allows to properly shrink zones on memory unplug
even if the separate memory blocks of a DIMM were onlined to different
zones or re-onlined to a different zone after offlining.
Example:
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 0
present 0
managed 0
:/# echo "online_movable" > /sys/devices/system/memory/memory41/state
:/# echo "online_movable" > /sys/devices/system/memory/memory43/state
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 98304
present 65536
managed 65536
:/# echo 0 > /sys/devices/system/memory/memory43/online
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 32768
present 32768
managed 32768
:/# echo 0 > /sys/devices/system/memory/memory41/online
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 0
present 0
managed 0
This patch (of 6):
The third argument is actually number of pages. Change the variable name
from size to nr_pages to indicate this better.
No functional change in this patch.
Link: http://lkml.kernel.org/r/20191006085646.5768-3-david@redhat.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Pankaj Gupta <pagupta@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-02-04 01:34:06 +00:00
|
|
|
unsigned long nr_pages,
|
2018-10-26 22:07:52 +00:00
|
|
|
struct dev_pagemap *pgmap)
|
|
|
|
{
|
mm/memmap_init: update variable name in memmap_init_zone
Patch series "mm/memory_hotplug: Shrink zones before removing memory", v6.
This series fixes the access of uninitialized memmaps when shrinking
zones/nodes and when removing memory. Also, it contains all fixes for
crashes that can be triggered when removing certain namespace using
memunmap_pages() - ZONE_DEVICE, reported by Aneesh.
We stop trying to shrink ZONE_DEVICE, as it's buggy, fixing it would be
more involved (we don't have SECTION_IS_ONLINE as an indicator), and
shrinking is only of limited use (set_zone_contiguous() cannot detect the
ZONE_DEVICE as contiguous).
We continue shrinking !ZONE_DEVICE zones, however, I reduced the amount of
code to a minimum. Shrinking is especially necessary to keep
zone->contiguous set where possible, especially, on memory unplug of DIMMs
at zone boundaries.
--------------------------------------------------------------------------
Zones are now properly shrunk when offlining memory blocks or when
onlining failed. This allows to properly shrink zones on memory unplug
even if the separate memory blocks of a DIMM were onlined to different
zones or re-onlined to a different zone after offlining.
Example:
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 0
present 0
managed 0
:/# echo "online_movable" > /sys/devices/system/memory/memory41/state
:/# echo "online_movable" > /sys/devices/system/memory/memory43/state
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 98304
present 65536
managed 65536
:/# echo 0 > /sys/devices/system/memory/memory43/online
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 32768
present 32768
managed 32768
:/# echo 0 > /sys/devices/system/memory/memory41/online
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 0
present 0
managed 0
This patch (of 6):
The third argument is actually number of pages. Change the variable name
from size to nr_pages to indicate this better.
No functional change in this patch.
Link: http://lkml.kernel.org/r/20191006085646.5768-3-david@redhat.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Pankaj Gupta <pagupta@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-02-04 01:34:06 +00:00
|
|
|
unsigned long pfn, end_pfn = start_pfn + nr_pages;
|
2018-10-26 22:07:52 +00:00
|
|
|
struct pglist_data *pgdat = zone->zone_pgdat;
|
2019-06-26 12:27:13 +00:00
|
|
|
struct vmem_altmap *altmap = pgmap_altmap(pgmap);
|
2018-10-26 22:07:52 +00:00
|
|
|
unsigned long zone_idx = zone_idx(zone);
|
|
|
|
unsigned long start = jiffies;
|
|
|
|
int nid = pgdat->node_id;
|
|
|
|
|
2019-07-18 22:58:18 +00:00
|
|
|
if (WARN_ON_ONCE(!pgmap || zone_idx(zone) != ZONE_DEVICE))
|
2018-10-26 22:07:52 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The call to memmap_init_zone should have already taken care
|
|
|
|
* of the pages reserved for the memmap, so we can just jump to
|
|
|
|
* the end of that region and start processing the device pages.
|
|
|
|
*/
|
2019-06-26 12:27:13 +00:00
|
|
|
if (altmap) {
|
2018-10-26 22:07:52 +00:00
|
|
|
start_pfn = altmap->base_pfn + vmem_altmap_offset(altmap);
|
mm/memmap_init: update variable name in memmap_init_zone
Patch series "mm/memory_hotplug: Shrink zones before removing memory", v6.
This series fixes the access of uninitialized memmaps when shrinking
zones/nodes and when removing memory. Also, it contains all fixes for
crashes that can be triggered when removing certain namespace using
memunmap_pages() - ZONE_DEVICE, reported by Aneesh.
We stop trying to shrink ZONE_DEVICE, as it's buggy, fixing it would be
more involved (we don't have SECTION_IS_ONLINE as an indicator), and
shrinking is only of limited use (set_zone_contiguous() cannot detect the
ZONE_DEVICE as contiguous).
We continue shrinking !ZONE_DEVICE zones, however, I reduced the amount of
code to a minimum. Shrinking is especially necessary to keep
zone->contiguous set where possible, especially, on memory unplug of DIMMs
at zone boundaries.
--------------------------------------------------------------------------
Zones are now properly shrunk when offlining memory blocks or when
onlining failed. This allows to properly shrink zones on memory unplug
even if the separate memory blocks of a DIMM were onlined to different
zones or re-onlined to a different zone after offlining.
Example:
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 0
present 0
managed 0
:/# echo "online_movable" > /sys/devices/system/memory/memory41/state
:/# echo "online_movable" > /sys/devices/system/memory/memory43/state
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 98304
present 65536
managed 65536
:/# echo 0 > /sys/devices/system/memory/memory43/online
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 32768
present 32768
managed 32768
:/# echo 0 > /sys/devices/system/memory/memory41/online
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 0
present 0
managed 0
This patch (of 6):
The third argument is actually number of pages. Change the variable name
from size to nr_pages to indicate this better.
No functional change in this patch.
Link: http://lkml.kernel.org/r/20191006085646.5768-3-david@redhat.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Pankaj Gupta <pagupta@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-02-04 01:34:06 +00:00
|
|
|
nr_pages = end_pfn - start_pfn;
|
2018-10-26 22:07:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
for (pfn = start_pfn; pfn < end_pfn; pfn++) {
|
|
|
|
struct page *page = pfn_to_page(pfn);
|
|
|
|
|
|
|
|
__init_single_page(page, pfn, zone_idx, nid);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Mark page reserved as it will need to wait for onlining
|
|
|
|
* phase for it to be fully associated with a zone.
|
|
|
|
*
|
|
|
|
* We can use the non-atomic __set_bit operation for setting
|
|
|
|
* the flag as we are still initializing the pages.
|
|
|
|
*/
|
|
|
|
__SetPageReserved(page);
|
|
|
|
|
|
|
|
/*
|
2019-06-26 12:27:21 +00:00
|
|
|
* ZONE_DEVICE pages union ->lru with a ->pgmap back pointer
|
|
|
|
* and zone_device_data. It is a bug if a ZONE_DEVICE page is
|
|
|
|
* ever freed or placed on a driver-private list.
|
2018-10-26 22:07:52 +00:00
|
|
|
*/
|
|
|
|
page->pgmap = pgmap;
|
2019-06-26 12:27:21 +00:00
|
|
|
page->zone_device_data = NULL;
|
2018-10-26 22:07:52 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Mark the block movable so that blocks are reserved for
|
|
|
|
* movable at startup. This will force kernel allocations
|
|
|
|
* to reserve their blocks rather than leaking throughout
|
|
|
|
* the address space during boot when many long-lived
|
|
|
|
* kernel allocations are made.
|
|
|
|
*
|
2020-09-26 04:19:28 +00:00
|
|
|
* Please note that MEMINIT_HOTPLUG path doesn't clear memmap
|
2019-07-18 22:58:26 +00:00
|
|
|
* because this is done early in section_activate()
|
2018-10-26 22:07:52 +00:00
|
|
|
*/
|
2020-10-16 03:08:15 +00:00
|
|
|
if (IS_ALIGNED(pfn, pageblock_nr_pages)) {
|
2018-10-26 22:07:52 +00:00
|
|
|
set_pageblock_migratetype(page, MIGRATE_MOVABLE);
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-08-18 09:05:55 +00:00
|
|
|
pr_info("%s initialised %lu pages in %ums\n", __func__,
|
mm/memmap_init: update variable name in memmap_init_zone
Patch series "mm/memory_hotplug: Shrink zones before removing memory", v6.
This series fixes the access of uninitialized memmaps when shrinking
zones/nodes and when removing memory. Also, it contains all fixes for
crashes that can be triggered when removing certain namespace using
memunmap_pages() - ZONE_DEVICE, reported by Aneesh.
We stop trying to shrink ZONE_DEVICE, as it's buggy, fixing it would be
more involved (we don't have SECTION_IS_ONLINE as an indicator), and
shrinking is only of limited use (set_zone_contiguous() cannot detect the
ZONE_DEVICE as contiguous).
We continue shrinking !ZONE_DEVICE zones, however, I reduced the amount of
code to a minimum. Shrinking is especially necessary to keep
zone->contiguous set where possible, especially, on memory unplug of DIMMs
at zone boundaries.
--------------------------------------------------------------------------
Zones are now properly shrunk when offlining memory blocks or when
onlining failed. This allows to properly shrink zones on memory unplug
even if the separate memory blocks of a DIMM were onlined to different
zones or re-onlined to a different zone after offlining.
Example:
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 0
present 0
managed 0
:/# echo "online_movable" > /sys/devices/system/memory/memory41/state
:/# echo "online_movable" > /sys/devices/system/memory/memory43/state
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 98304
present 65536
managed 65536
:/# echo 0 > /sys/devices/system/memory/memory43/online
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 32768
present 32768
managed 32768
:/# echo 0 > /sys/devices/system/memory/memory41/online
:/# cat /proc/zoneinfo
Node 1, zone Movable
spanned 0
present 0
managed 0
This patch (of 6):
The third argument is actually number of pages. Change the variable name
from size to nr_pages to indicate this better.
No functional change in this patch.
Link: http://lkml.kernel.org/r/20191006085646.5768-3-david@redhat.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Pankaj Gupta <pagupta@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-02-04 01:34:06 +00:00
|
|
|
nr_pages, jiffies_to_msecs(jiffies - start));
|
2018-10-26 22:07:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
#endif
|
2008-02-05 06:29:26 +00:00
|
|
|
static void __meminit zone_init_free_lists(struct zone *zone)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2014-06-04 23:10:21 +00:00
|
|
|
unsigned int order, t;
|
2007-10-16 08:25:48 +00:00
|
|
|
for_each_migratetype_order(order, t) {
|
|
|
|
INIT_LIST_HEAD(&zone->free_area[order].free_list[t]);
|
2005-04-16 22:20:36 +00:00
|
|
|
zone->free_area[order].nr_free = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-02-24 20:06:17 +00:00
|
|
|
void __meminit __weak memmap_init_zone(struct zone *zone)
|
2018-10-26 22:09:32 +00:00
|
|
|
{
|
2021-02-24 20:06:17 +00:00
|
|
|
unsigned long zone_start_pfn = zone->zone_start_pfn;
|
|
|
|
unsigned long zone_end_pfn = zone_start_pfn + zone->spanned_pages;
|
|
|
|
int i, nid = zone_to_nid(zone), zone_id = zone_idx(zone);
|
2020-06-03 22:57:55 +00:00
|
|
|
unsigned long start_pfn, end_pfn;
|
|
|
|
|
|
|
|
for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
|
2021-02-24 20:06:17 +00:00
|
|
|
start_pfn = clamp(start_pfn, zone_start_pfn, zone_end_pfn);
|
|
|
|
end_pfn = clamp(end_pfn, zone_start_pfn, zone_end_pfn);
|
2020-06-03 22:57:55 +00:00
|
|
|
|
2021-02-24 20:06:17 +00:00
|
|
|
if (end_pfn > start_pfn)
|
|
|
|
memmap_init_range(end_pfn - start_pfn, nid,
|
|
|
|
zone_id, start_pfn, zone_end_pfn,
|
|
|
|
MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
|
2020-06-03 22:57:55 +00:00
|
|
|
}
|
2018-10-26 22:09:32 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2014-06-23 20:22:04 +00:00
|
|
|
static int zone_batchsize(struct zone *zone)
|
2005-06-22 00:14:47 +00:00
|
|
|
{
|
nommu: clamp zone_batchsize() to 0 under NOMMU conditions
Clamp zone_batchsize() to 0 under NOMMU conditions to stop
free_hot_cold_page() from queueing and batching frees.
The problem is that under NOMMU conditions it is really important to be
able to allocate large contiguous chunks of memory, but when munmap() or
exit_mmap() releases big stretches of memory, return of these to the buddy
allocator can be deferred, and when it does finally happen, it can be in
small chunks.
Whilst the fragmentation this incurs isn't so much of a problem under MMU
conditions as userspace VM is glued together from individual pages with
the aid of the MMU, it is a real problem if there isn't an MMU.
By clamping the page freeing queue size to 0, pages are returned to the
allocator immediately, and the buddy detector is more likely to be able to
glue them together into large chunks immediately, and fragmentation is
less likely to occur.
By disabling batching of frees, and by turning off the trimming of excess
space during boot, Coldfire can manage to boot.
Reported-by: Lanttor Guo <lanttor.guo@freescale.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Lanttor Guo <lanttor.guo@freescale.com>
Cc: Greg Ungerer <gerg@snapgear.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-05-06 23:03:03 +00:00
|
|
|
#ifdef CONFIG_MMU
|
2005-06-22 00:14:47 +00:00
|
|
|
int batch;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The per-cpu-pages pools are set to around 1000th of the
|
mm, page_alloc: double zone's batchsize
To improve page allocator's performance for order-0 pages, each CPU has
a Per-CPU-Pageset(PCP) per zone. Whenever an order-0 page is needed,
PCP will be checked first before asking pages from Buddy. When PCP is
used up, a batch of pages will be fetched from Buddy to improve
performance and the size of batch can affect performance.
zone's batch size gets doubled last time by commit ba56e91c9401("mm:
page_alloc: increase size of per-cpu-pages") over ten years ago. Since
then, CPU has envolved a lot and CPU's cache sizes also increased.
Dave Hansen is concerned the current batch size doesn't fit well with
modern hardware and suggested me to do two things: first, use a page
allocator intensive benchmark, e.g. will-it-scale/page_fault1 to find
out how performance changes with different batch sizes on various
machines and then choose a new default batch size; second, see how this
new batch size work with other workloads.
In the first test, we saw performance gains on high-core-count systems
and little to no effect on older systems with more modest core counts.
In this phase's test data, two candidates: 63 and 127 are chosen.
In the second step, ebizzy, oltp, kbuild, pigz, netperf, vm-scalability
and more will-it-scale sub-tests are tested to see how these two
candidates work with these workloads and decides a new default according
to their results.
Most test results are flat. will-it-scale/page_fault2 process mode has
10%-18% performance increase on 4-sockets Skylake and Broadwell.
vm-scalability/lru-file-mmap-read has 17%-47% performance increase for
4-sockets servers while for 2-sockets servers, it caused 3%-8% performance
drop. Further analysis showed that, with a larger pcp->batch and thus
larger pcp->high(the relationship of pcp->high=6 * pcp->batch is
maintained in this patch), zone lock contention shifted to LRU add side
lock contention and that caused performance drop. This performance drop
might be mitigated by others' work on optimizing LRU lock.
Another downside of increasing pcp->batch is, when PCP is used up and need
to fetch a batch of pages from Buddy, since batch is increased, that time
can be longer than before. My understanding is, this doesn't affect
slowpath where direct reclaim and compaction dominates. For fastpath,
throughput is a win(according to will-it-scale/page_fault1) but worst
latency can be larger now.
Overall, I think double the batch size from 31 to 63 is relatively safe
and provide good performance boost for high-core-count systems.
The two phase's test results are listed below(all tests are done with THP
disabled).
Phase one(will-it-scale/page_fault1) test results:
Skylake-EX: increased batch size has a good effect on zone->lock
contention, though LRU contention will rise at the same time and
limited the final performance increase.
batch score change zone_contention lru_contention total_contention
31 15345900 +0.00% 64% 8% 72%
53 17903847 +16.67% 32% 38% 70%
63 17992886 +17.25% 24% 45% 69%
73 18022825 +17.44% 10% 61% 71%
119 18023401 +17.45% 4% 66% 70%
127 18029012 +17.48% 3% 66% 69%
137 18036075 +17.53% 4% 66% 70%
165 18035964 +17.53% 2% 67% 69%
188 18101105 +17.95% 2% 67% 69%
223 18130951 +18.15% 2% 67% 69%
255 18118898 +18.07% 2% 67% 69%
267 18101559 +17.96% 2% 67% 69%
299 18160468 +18.34% 2% 68% 70%
320 18139845 +18.21% 2% 67% 69%
393 18160869 +18.34% 2% 68% 70%
424 18170999 +18.41% 2% 68% 70%
458 18144868 +18.24% 2% 68% 70%
467 18142366 +18.22% 2% 68% 70%
498 18154549 +18.30% 1% 68% 69%
511 18134525 +18.17% 1% 69% 70%
Broadwell-EX: similar pattern as Skylake-EX.
batch score change zone_contention lru_contention total_contention
31 16703983 +0.00% 67% 7% 74%
53 18195393 +8.93% 43% 28% 71%
63 18288885 +9.49% 38% 33% 71%
73 18344329 +9.82% 35% 37% 72%
119 18535529 +10.96% 24% 46% 70%
127 18513596 +10.83% 23% 48% 71%
137 18514327 +10.84% 23% 48% 71%
165 18511840 +10.82% 22% 49% 71%
188 18593478 +11.31% 17% 53% 70%
223 18601667 +11.36% 17% 52% 69%
255 18774825 +12.40% 12% 58% 70%
267 18754781 +12.28% 9% 60% 69%
299 18892265 +13.10% 7% 63% 70%
320 18873812 +12.99% 8% 62% 70%
393 18891174 +13.09% 6% 64% 70%
424 18975108 +13.60% 6% 64% 70%
458 18932364 +13.34% 8% 62% 70%
467 18960891 +13.51% 5% 65% 70%
498 18944526 +13.41% 5% 64% 69%
511 18960839 +13.51% 5% 64% 69%
Skylake-EP: although increased batch reduced zone->lock contention, but
the effect is not as good as EX: zone->lock contention is still as high as
20% with a very high batch value instead of 1% on Skylake-EX or 5% on
Broadwell-EX. Also, total_contention actually decreased with a higher
batch but that doesn't translate to performance increase.
batch score change zone_contention lru_contention total_contention
31 9554867 +0.00% 66% 3% 69%
53 9855486 +3.15% 63% 3% 66%
63 9980145 +4.45% 62% 4% 66%
73 10092774 +5.63% 62% 5% 67%
119 10310061 +7.90% 45% 19% 64%
127 10342019 +8.24% 42% 19% 61%
137 10358182 +8.41% 42% 21% 63%
165 10397060 +8.81% 37% 24% 61%
188 10341808 +8.24% 34% 26% 60%
223 10349135 +8.31% 31% 27% 58%
255 10327189 +8.08% 28% 29% 57%
267 10344204 +8.26% 27% 29% 56%
299 10325043 +8.06% 25% 30% 55%
320 10310325 +7.91% 25% 31% 56%
393 10293274 +7.73% 21% 31% 52%
424 10311099 +7.91% 21% 32% 53%
458 10321375 +8.02% 21% 32% 53%
467 10303881 +7.84% 21% 32% 53%
498 10332462 +8.14% 20% 33% 53%
511 10325016 +8.06% 20% 32% 52%
Broadwell-EP: zone->lock and lru lock had an agreement to make sure
performance doesn't increase and they successfully managed to keep total
contention at 70%.
batch score change zone_contention lru_contention total_contention
31 10121178 +0.00% 19% 50% 69%
53 10142366 +0.21% 6% 63% 69%
63 10117984 -0.03% 11% 58% 69%
73 10123330 +0.02% 7% 63% 70%
119 10108791 -0.12% 2% 67% 69%
127 10166074 +0.44% 3% 66% 69%
137 10141574 +0.20% 3% 66% 69%
165 10154499 +0.33% 2% 68% 70%
188 10124921 +0.04% 2% 67% 69%
223 10137399 +0.16% 2% 67% 69%
255 10143289 +0.22% 0% 68% 68%
267 10123535 +0.02% 1% 68% 69%
299 10140952 +0.20% 0% 68% 68%
320 10163170 +0.41% 0% 68% 68%
393 10000633 -1.19% 0% 69% 69%
424 10087998 -0.33% 0% 69% 69%
458 10187116 +0.65% 0% 69% 69%
467 10146790 +0.25% 0% 69% 69%
498 10197958 +0.76% 0% 69% 69%
511 10152326 +0.31% 0% 69% 69%
Haswell-EP: similar to Broadwell-EP.
batch score change zone_contention lru_contention total_contention
31 10442205 +0.00% 14% 48% 62%
53 10442255 +0.00% 5% 57% 62%
63 10452059 +0.09% 6% 57% 63%
73 10482349 +0.38% 5% 59% 64%
119 10454644 +0.12% 3% 60% 63%
127 10431514 -0.10% 3% 59% 62%
137 10423785 -0.18% 3% 60% 63%
165 10481216 +0.37% 2% 61% 63%
188 10448755 +0.06% 2% 61% 63%
223 10467144 +0.24% 2% 61% 63%
255 10480215 +0.36% 2% 61% 63%
267 10484279 +0.40% 2% 61% 63%
299 10466450 +0.23% 2% 61% 63%
320 10452578 +0.10% 2% 61% 63%
393 10499678 +0.55% 1% 62% 63%
424 10481454 +0.38% 1% 62% 63%
458 10473562 +0.30% 1% 62% 63%
467 10484269 +0.40% 0% 62% 62%
498 10505599 +0.61% 0% 62% 62%
511 10483395 +0.39% 0% 62% 62%
Westmere-EP: contention is pretty small so not interesting. Note too high
a batch value could hurt performance.
batch score change zone_contention lru_contention total_contention
31 4831523 +0.00% 2% 3% 5%
53 4834086 +0.05% 2% 4% 6%
63 4834262 +0.06% 2% 3% 5%
73 4832851 +0.03% 2% 4% 6%
119 4830534 -0.02% 1% 3% 4%
127 4827461 -0.08% 1% 4% 5%
137 4827459 -0.08% 1% 3% 4%
165 4820534 -0.23% 0% 4% 4%
188 4817947 -0.28% 0% 3% 3%
223 4809671 -0.45% 0% 3% 3%
255 4802463 -0.60% 0% 4% 4%
267 4801634 -0.62% 0% 3% 3%
299 4798047 -0.69% 0% 3% 3%
320 4793084 -0.80% 0% 3% 3%
393 4785877 -0.94% 0% 3% 3%
424 4782911 -1.01% 0% 3% 3%
458 4779346 -1.08% 0% 3% 3%
467 4780306 -1.06% 0% 3% 3%
498 4780589 -1.05% 0% 3% 3%
511 4773724 -1.20% 0% 3% 3%
Skylake-Desktop: similar to Westmere-EP, nothing interesting.
batch score change zone_contention lru_contention total_contention
31 3906608 +0.00% 2% 3% 5%
53 3940164 +0.86% 2% 3% 5%
63 3937289 +0.79% 2% 3% 5%
73 3940201 +0.86% 2% 3% 5%
119 3933240 +0.68% 2% 3% 5%
127 3930514 +0.61% 2% 4% 6%
137 3938639 +0.82% 0% 3% 3%
165 3908755 +0.05% 0% 3% 3%
188 3905621 -0.03% 0% 3% 3%
223 3903015 -0.09% 0% 4% 4%
255 3889480 -0.44% 0% 3% 3%
267 3891669 -0.38% 0% 4% 4%
299 3898728 -0.20% 0% 4% 4%
320 3894547 -0.31% 0% 4% 4%
393 3875137 -0.81% 0% 4% 4%
424 3874521 -0.82% 0% 3% 3%
458 3880432 -0.67% 0% 4% 4%
467 3888715 -0.46% 0% 3% 3%
498 3888633 -0.46% 0% 4% 4%
511 3875305 -0.80% 0% 5% 5%
Haswell-Desktop: zone->lock is pretty low as other desktops, though lru
contention is higher than other desktops.
batch score change zone_contention lru_contention total_contention
31 3511158 +0.00% 2% 5% 7%
53 3555445 +1.26% 2% 6% 8%
63 3561082 +1.42% 2% 6% 8%
73 3547218 +1.03% 2% 6% 8%
119 3571319 +1.71% 1% 7% 8%
127 3549375 +1.09% 0% 6% 6%
137 3560233 +1.40% 0% 6% 6%
165 3555176 +1.25% 2% 6% 8%
188 3551501 +1.15% 0% 8% 8%
223 3531462 +0.58% 0% 7% 7%
255 3570400 +1.69% 0% 7% 7%
267 3532235 +0.60% 1% 8% 9%
299 3562326 +1.46% 0% 6% 6%
320 3553569 +1.21% 0% 8% 8%
393 3539519 +0.81% 0% 7% 7%
424 3549271 +1.09% 0% 8% 8%
458 3528885 +0.50% 0% 8% 8%
467 3526554 +0.44% 0% 7% 7%
498 3525302 +0.40% 0% 9% 9%
511 3527556 +0.47% 0% 8% 8%
Sandybridge-Desktop: the 0% contention isn't accurate but caused by
dropped fractional part. Since multiple contention path's contentions
are all under 1% here, with some arithmetic operations like add, the
final deviation could be as large as 3%.
batch score change zone_contention lru_contention total_contention
31 1744495 +0.00% 0% 0% 0%
53 1755341 +0.62% 0% 0% 0%
63 1758469 +0.80% 0% 0% 0%
73 1759626 +0.87% 0% 0% 0%
119 1770417 +1.49% 0% 0% 0%
127 1768252 +1.36% 0% 0% 0%
137 1767848 +1.34% 0% 0% 0%
165 1765088 +1.18% 0% 0% 0%
188 1766918 +1.29% 0% 0% 0%
223 1767866 +1.34% 0% 0% 0%
255 1768074 +1.35% 0% 0% 0%
267 1763187 +1.07% 0% 0% 0%
299 1765620 +1.21% 0% 0% 0%
320 1767603 +1.32% 0% 0% 0%
393 1764612 +1.15% 0% 0% 0%
424 1758476 +0.80% 0% 0% 0%
458 1758593 +0.81% 0% 0% 0%
467 1757915 +0.77% 0% 0% 0%
498 1753363 +0.51% 0% 0% 0%
511 1755548 +0.63% 0% 0% 0%
Phase two test results:
Note: all percent change is against base(batch=31).
ebizzy.throughput (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 2410037±7% 2600451±2% +7.9% 2602878 +8.0%
lkp-bdw-ex1 1493328 1489243 -0.3% 1492145 -0.1%
lkp-skl-2sp2 1329674 1345891 +1.2% 1351056 +1.6%
lkp-bdw-ep2 711511 711511 0.0% 710708 -0.1%
lkp-wsm-ep2 75750 75528 -0.3% 75441 -0.4%
lkp-skl-d01 264126 262791 -0.5% 264113 +0.0%
lkp-hsw-d01 176601 176328 -0.2% 176368 -0.1%
lkp-sb02 98937 98937 +0.0% 99030 +0.1%
kbuild.buildtime (less is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 107.00 107.67 +0.6% 107.11 +0.1%
lkp-bdw-ex1 97.33 97.33 +0.0% 97.42 +0.1%
lkp-skl-2sp2 180.00 179.83 -0.1% 179.83 -0.1%
lkp-bdw-ep2 178.17 179.17 +0.6% 177.50 -0.4%
lkp-wsm-ep2 737.00 738.00 +0.1% 738.00 +0.1%
lkp-skl-d01 642.00 653.00 +1.7% 653.00 +1.7%
lkp-hsw-d01 1310.00 1316.00 +0.5% 1311.00 +0.1%
netperf/TCP_STREAM.Throughput_total_Mbps (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 948790 947144 -0.2% 948333 -0.0%
lkp-bdw-ex1 904224 904366 +0.0% 904926 +0.1%
lkp-skl-2sp2 239731 239607 -0.1% 239565 -0.1%
lk-bdw-ep2 365764 365933 +0.0% 365951 +0.1%
lkp-wsm-ep2 93736 93803 +0.1% 93808 +0.1%
lkp-skl-d01 77314 77303 -0.0% 77375 +0.1%
lkp-hsw-d01 58617 60387 +3.0% 60208 +2.7%
lkp-sb02 29990 30137 +0.5% 30103 +0.4%
oltp.transactions (higer is better)
machine batch=31 batch=63 batch=127
lkp-bdw-ex1 9073276 9100377 +0.3% 9036344 -0.4%
lkp-skl-2sp2 8898717 8852054 -0.5% 8894459 -0.0%
lkp-bdw-ep2 13426155 13384654 -0.3% 13333637 -0.7%
lkp-hsw-ep2 13146314 13232784 +0.7% 13193163 +0.4%
lkp-wsm-ep2 5035355 5019348 -0.3% 5033418 -0.0%
lkp-skl-d01 418485 4413339 -0.1% 4419039 +0.0%
lkp-hsw-d01 3517817±5% 3396120±3% -3.5% 3455138±3% -1.8%
pigz.throughput (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.513e+08 1.507e+08 -0.4% 1.511e+08 -0.2%
lkp-bdw-ex1 2.060e+08 2.052e+08 -0.4% 2.044e+08 -0.8%
lkp-skl-2sp2 8.836e+08 8.845e+08 +0.1% 8.836e+08 -0.0%
lkp-bdw-ep2 8.275e+08 8.464e+08 +2.3% 8.330e+08 +0.7%
lkp-wsm-ep2 2.224e+08 2.221e+08 -0.2% 2.218e+08 -0.3%
lkp-skl-d01 1.177e+08 1.177e+08 -0.0% 1.176e+08 -0.1%
lkp-hsw-d01 1.154e+08 1.154e+08 +0.1% 1.154e+08 -0.0%
lkp-sb02 0.633e+08 0.633e+08 +0.1% 0.633e+08 +0.0%
will-it-scale.malloc1.processes (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 620181 620484 +0.0% 620240 +0.0%
lkp-bdw-ex1 1403610 1401201 -0.2% 1417900 +1.0%
lkp-skl-2sp2 1288097 1284145 -0.3% 1283907 -0.3%
lkp-bdw-ep2 1427879 1427675 -0.0% 1428266 +0.0%
lkp-hsw-ep2 1362546 1353965 -0.6% 1354759 -0.6%
lkp-wsm-ep2 2099657 2107576 +0.4% 2100226 +0.0%
lkp-skl-d01 1476835 1476358 -0.0% 1474487 -0.2%
lkp-hsw-d01 1308810 1303429 -0.4% 1301299 -0.6%
lkp-sb02 589286 589284 -0.0% 588101 -0.2%
will-it-scale.malloc1.threads (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 21289 21125 -0.8% 21241 -0.2%
lkp-bdw-ex1 28114 28089 -0.1% 28007 -0.4%
lkp-skl-2sp2 91866 91946 +0.1% 92723 +0.9%
lkp-bdw-ep2 37637 37501 -0.4% 37317 -0.9%
lkp-hsw-ep2 43673 43590 -0.2% 43754 +0.2%
lkp-wsm-ep2 28577 28298 -1.0% 28545 -0.1%
lkp-skl-d01 175277 173343 -1.1% 173082 -1.3%
lkp-hsw-d01 130303 129566 -0.6% 129250 -0.8%
lkp-sb02 113742±3% 116911 +2.8% 116417±3% +2.4%
will-it-scale.malloc2.processes (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.206e+09 1.206e+09 -0.0% 1.206e+09 +0.0%
lkp-bdw-ex1 1.319e+09 1.319e+09 -0.0% 1.319e+09 +0.0%
lkp-skl-2sp2 8.000e+08 8.021e+08 +0.3% 7.995e+08 -0.1%
lkp-bdw-ep2 6.582e+08 6.634e+08 +0.8% 6.513e+08 -1.1%
lkp-hsw-ep2 6.671e+08 6.669e+08 -0.0% 6.665e+08 -0.1%
lkp-wsm-ep2 1.805e+08 1.806e+08 +0.0% 1.804e+08 -0.1%
lkp-skl-d01 1.611e+08 1.611e+08 -0.0% 1.610e+08 -0.0%
lkp-hsw-d01 1.333e+08 1.332e+08 -0.0% 1.332e+08 -0.0%
lkp-sb02 82485104 82478206 -0.0% 82473546 -0.0%
will-it-scale.malloc2.threads (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.574e+09 1.574e+09 -0.0% 1.574e+09 -0.0%
lkp-bdw-ex1 1.737e+09 1.737e+09 +0.0% 1.737e+09 -0.0%
lkp-skl-2sp2 9.161e+08 9.162e+08 +0.0% 9.181e+08 +0.2%
lkp-bdw-ep2 7.856e+08 8.015e+08 +2.0% 8.113e+08 +3.3%
lkp-hsw-ep2 6.908e+08 6.904e+08 -0.1% 6.907e+08 -0.0%
lkp-wsm-ep2 2.409e+08 2.409e+08 +0.0% 2.409e+08 -0.0%
lkp-skl-d01 1.199e+08 1.199e+08 -0.0% 1.199e+08 -0.0%
lkp-hsw-d01 1.029e+08 1.029e+08 -0.0% 1.029e+08 +0.0%
lkp-sb02 68081213 68061423 -0.0% 68076037 -0.0%
will-it-scale.page_fault2.processes (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 14509125±4% 16472364 +13.5% 17123117 +18.0%
lkp-bdw-ex1 14736381 16196588 +9.9% 16364011 +11.0%
lkp-skl-2sp2 6354925 6435444 +1.3% 6436644 +1.3%
lkp-bdw-ep2 8749584 8834422 +1.0% 8827179 +0.9%
lkp-hsw-ep2 8762591 8845920 +1.0% 8825697 +0.7%
lkp-wsm-ep2 3036083 3030428 -0.2% 3021741 -0.5%
lkp-skl-d01 2307834 2304731 -0.1% 2286142 -0.9%
lkp-hsw-d01 1806237 1800786 -0.3% 1795943 -0.6%
lkp-sb02 842616 837844 -0.6% 833921 -1.0%
will-it-scale.page_fault2.threads
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1623294 1615132±2% -0.5% 1656777 +2.1%
lkp-bdw-ex1 1995714 2025948 +1.5% 2113753±3% +5.9%
lkp-skl-2sp2 2346708 2415591 +2.9% 2416919 +3.0%
lkp-bdw-ep2 2342564 2344882 +0.1% 2300206 -1.8%
lkp-hsw-ep2 1820658 1831681 +0.6% 1844057 +1.3%
lkp-wsm-ep2 1725482 1733774 +0.5% 1740517 +0.9%
lkp-skl-d01 1832833 1823628 -0.5% 1806489 -1.4%
lkp-hsw-d01 1427913 1427287 -0.0% 1420226 -0.5%
lkp-sb02 750626 748615 -0.3% 746621 -0.5%
will-it-scale.page_fault3.processes (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 24382726 24400317 +0.1% 24668774 +1.2%
lkp-bdw-ex1 35399750 35683124 +0.8% 35829492 +1.2%
lkp-skl-2sp2 28136820 28068248 -0.2% 28147989 +0.0%
lkp-bdw-ep2 37269077 37459490 +0.5% 37373073 +0.3%
lkp-hsw-ep2 36224967 36114085 -0.3% 36104908 -0.3%
lkp-wsm-ep2 16820457 16911005 +0.5% 16968596 +0.9%
lkp-skl-d01 7721138 7725904 +0.1% 7756740 +0.5%
lkp-hsw-d01 7611979 7650928 +0.5% 7651323 +0.5%
lkp-sb02 3781546 3796502 +0.4% 3796827 +0.4%
will-it-scale.page_fault3.threads (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1865820±3% 1900917±2% +1.9% 1826245±4% -2.1%
lkp-bdw-ex1 3094060 3148326 +1.8% 3150036 +1.8%
lkp-skl-2sp2 3952940 3953898 +0.0% 3989360 +0.9%
lkp-bdw-ep2 3420373±3% 3643964 +6.5% 3644910±5% +6.6%
lkp-hsw-ep2 2609635±2% 2582310±3% -1.0% 2780459 +6.5%
lkp-wsm-ep2 4395001 4417196 +0.5% 4432499 +0.9%
lkp-skl-d01 5363977 5400003 +0.7% 5411370 +0.9%
lkp-hsw-d01 5274131 5311294 +0.7% 5319359 +0.9%
lkp-sb02 2917314 2913004 -0.1% 2935286 +0.6%
will-it-scale.read1.processes (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 73762279±14% 69322519±10% -6.0% 69349855±13% -6.0% (result unstable)
lkp-bdw-ex1 1.701e+08 1.704e+08 +0.1% 1.705e+08 +0.2%
lkp-skl-2sp2 63111570 63113953 +0.0% 63836573 +1.1%
lkp-bdw-ep2 79247409 79424610 +0.2% 78012656 -1.6%
lkp-hsw-ep2 67677026 68308800 +0.9% 67539106 -0.2%
lkp-wsm-ep2 13339630 13939817 +4.5% 13766865 +3.2%
lkp-skl-d01 10969487 10972650 +0.0% no data
lkp-hsw-d01 9857342±2% 10080592±2% +2.3% 10131560 +2.8%
lkp-sb02 5189076 5197473 +0.2% 5163253 -0.5%
will-it-scale.read1.threads (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 62468045±12% 73666726±7% +17.9% 79553123±12% +27.4% (result unstable)
lkp-bdw-ex1 1.62e+08 1.624e+08 +0.3% 1.614e+08 -0.3%
lkp-skl-2sp2 58319780 59181032 +1.5% 59821353 +2.6%
lkp-bdw-ep2 74057992 75698171 +2.2% 74990869 +1.3%
lkp-hsw-ep2 63672959 63639652 -0.1% 64387051 +1.1%
lkp-wsm-ep2 13489943 13526058 +0.3% 13259032 -1.7%
lkp-skl-d01 10297906 10338796 +0.4% 10407328 +1.1%
lkp-hsw-d01 9636721 9667376 +0.3% 9341147 -3.1%
lkp-sb02 4801938 4804496 +0.1% 4802290 +0.0%
will-it-scale.write1.processes (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.111e+08 1.104e+08±2% -0.7% 1.122e+08±2% +1.0%
lkp-bdw-ex1 1.392e+08 1.399e+08 +0.5% 1.397e+08 +0.4%
lkp-skl-2sp2 59369233 58994841 -0.6% 58715168 -1.1%
lkp-bdw-ep2 61820979 CPU throttle 63593123 +2.9%
lkp-hsw-ep2 57897587 57435605 -0.8% 56347450 -2.7%
lkp-wsm-ep2 7814203 7918017±2% +1.3% 7669068 -1.9%
lkp-skl-d01 8886557 8971422 +1.0% 8818366 -0.8%
lkp-hsw-d01 9171001±5% 9189915 +0.2% 9483909 +3.4%
lkp-sb02 4475406 4475294 -0.0% 4501756 +0.6%
will-it-scale.write1.threads (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.058e+08 1.055e+08±2% -0.2% 1.065e+08 +0.7%
lkp-bdw-ex1 1.316e+08 1.300e+08 -1.2% 1.308e+08 -0.6%
lkp-skl-2sp2 54492421 56086678 +2.9% 55975657 +2.7%
lkp-bdw-ep2 59360449 59003957 -0.6% 58101262 -2.1%
lkp-hsw-ep2 53346346±2% 52530876 -1.5% 52902487 -0.8%
lkp-wsm-ep2 7774006 7800092±2% +0.3% 7558833 -2.8%
lkp-skl-d01 8346174 8235695 -1.3% no data
lkp-hsw-d01 8636244 8655731 +0.2% 8658868 +0.3%
lkp-sb02 4181820 4204107 +0.5% 4182992 +0.0%
vm-scalability.anon-r-rand.throughput (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 11933873±3% 12356544±2% +3.5% 12188624 +2.1%
lkp-bdw-ex1 7114424±2% 7330949±2% +3.0% 7392419 +3.9%
lkp-skl-2sp2 6773277±5% 6492332±8% -4.1% 6543962 -3.4%
lkp-bdw-ep2 7133846±4% 7233508 +1.4% 7013518±3% -1.7%
lkp-hsw-ep2 4576626 4527098 -1.1% 4551679 -0.5%
lkp-wsm-ep2 2583599 2592492 +0.3% 2588039 +0.2%
lkp-hsw-d01 998199±2% 1028311 +3.0% 1006460±2% +0.8%
lkp-sb02 570572 567854 -0.5% 568449 -0.4%
vm-scalability.anon-r-rand-mt.throughput (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1789419 1787830 -0.1% 1788208 -0.1%
lkp-bdw-ex1 3492595±2% 3554966±2% +1.8% 3558835±3% +1.9%
lkp-skl-2sp2 3856238±2% 3975403±4% +3.1% 3994600 +3.6%
lkp-bdw-ep2 3726963±11% 3809292±6% +2.2% 3871924±4% +3.9%
lkp-hsw-ep2 2131760±3% 2033578±4% -4.6% 2130727±6% -0.0%
lkp-wsm-ep2 2369731 2368384 -0.1% 2370252 +0.0%
lkp-skl-d01 1207128 1206220 -0.1% 1205801 -0.1%
lkp-hsw-d01 964317 992329±2% +2.9% 992099±2% +2.9%
lkp-sb02 567137 567346 +0.0% 566144 -0.2%
vm-scalability.lru-file-mmap-read.throughput (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 19560469±6% 23018999 +17.7% 23418800 +19.7%
lkp-bdw-ex1 17769135±14% 26141676±3% +47.1% 26284723±5% +47.9%
lkp-skl-2sp2 14056512 13578884 -3.4% 13146214 -6.5%
lkp-bdw-ep2 15336542 14737654 -3.9% 14088159 -8.1%
lkp-hsw-ep2 16275498 15756296 -3.2% 15018090 -7.7%
lkp-wsm-ep2 11272160 11237231 -0.3% 11310047 +0.3%
lkp-skl-d01 7322119 7324569 +0.0% 7184148 -1.9%
lkp-hsw-d01 6449234 6404542 -0.7% 6356141 -1.4%
lkp-sb02 3517943 3520668 +0.1% 3527309 +0.3%
vm-scalability.lru-file-mmap-read-rand.throughput (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1689052 1697553 +0.5% 1698726 +0.6%
lkp-bdw-ex1 1675246 1699764 +1.5% 1712226 +2.2%
lkp-skl-2sp2 1800533 1799749 -0.0% 1800581 +0.0%
lkp-bdw-ep2 1807422 1807758 +0.0% 1804932 -0.1%
lkp-hsw-ep2 1809807 1808781 -0.1% 1807811 -0.1%
lkp-wsm-ep2 1800198 1802434 +0.1% 1801236 +0.1%
lkp-skl-d01 696689 695537 -0.2% 694106 -0.4%
lkp-hsw-d01 698364 698666 +0.0% 696686 -0.2%
lkp-sb02 258939 258787 -0.1% 258199 -0.3%
Link: http://lkml.kernel.org/r/20180711055855.29072-1-aaron.lu@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Suggested-by: Dave Hansen <dave.hansen@intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:49:14 +00:00
|
|
|
* size of the zone.
|
2005-06-22 00:14:47 +00:00
|
|
|
*/
|
2018-12-28 08:34:24 +00:00
|
|
|
batch = zone_managed_pages(zone) / 1024;
|
mm, page_alloc: double zone's batchsize
To improve page allocator's performance for order-0 pages, each CPU has
a Per-CPU-Pageset(PCP) per zone. Whenever an order-0 page is needed,
PCP will be checked first before asking pages from Buddy. When PCP is
used up, a batch of pages will be fetched from Buddy to improve
performance and the size of batch can affect performance.
zone's batch size gets doubled last time by commit ba56e91c9401("mm:
page_alloc: increase size of per-cpu-pages") over ten years ago. Since
then, CPU has envolved a lot and CPU's cache sizes also increased.
Dave Hansen is concerned the current batch size doesn't fit well with
modern hardware and suggested me to do two things: first, use a page
allocator intensive benchmark, e.g. will-it-scale/page_fault1 to find
out how performance changes with different batch sizes on various
machines and then choose a new default batch size; second, see how this
new batch size work with other workloads.
In the first test, we saw performance gains on high-core-count systems
and little to no effect on older systems with more modest core counts.
In this phase's test data, two candidates: 63 and 127 are chosen.
In the second step, ebizzy, oltp, kbuild, pigz, netperf, vm-scalability
and more will-it-scale sub-tests are tested to see how these two
candidates work with these workloads and decides a new default according
to their results.
Most test results are flat. will-it-scale/page_fault2 process mode has
10%-18% performance increase on 4-sockets Skylake and Broadwell.
vm-scalability/lru-file-mmap-read has 17%-47% performance increase for
4-sockets servers while for 2-sockets servers, it caused 3%-8% performance
drop. Further analysis showed that, with a larger pcp->batch and thus
larger pcp->high(the relationship of pcp->high=6 * pcp->batch is
maintained in this patch), zone lock contention shifted to LRU add side
lock contention and that caused performance drop. This performance drop
might be mitigated by others' work on optimizing LRU lock.
Another downside of increasing pcp->batch is, when PCP is used up and need
to fetch a batch of pages from Buddy, since batch is increased, that time
can be longer than before. My understanding is, this doesn't affect
slowpath where direct reclaim and compaction dominates. For fastpath,
throughput is a win(according to will-it-scale/page_fault1) but worst
latency can be larger now.
Overall, I think double the batch size from 31 to 63 is relatively safe
and provide good performance boost for high-core-count systems.
The two phase's test results are listed below(all tests are done with THP
disabled).
Phase one(will-it-scale/page_fault1) test results:
Skylake-EX: increased batch size has a good effect on zone->lock
contention, though LRU contention will rise at the same time and
limited the final performance increase.
batch score change zone_contention lru_contention total_contention
31 15345900 +0.00% 64% 8% 72%
53 17903847 +16.67% 32% 38% 70%
63 17992886 +17.25% 24% 45% 69%
73 18022825 +17.44% 10% 61% 71%
119 18023401 +17.45% 4% 66% 70%
127 18029012 +17.48% 3% 66% 69%
137 18036075 +17.53% 4% 66% 70%
165 18035964 +17.53% 2% 67% 69%
188 18101105 +17.95% 2% 67% 69%
223 18130951 +18.15% 2% 67% 69%
255 18118898 +18.07% 2% 67% 69%
267 18101559 +17.96% 2% 67% 69%
299 18160468 +18.34% 2% 68% 70%
320 18139845 +18.21% 2% 67% 69%
393 18160869 +18.34% 2% 68% 70%
424 18170999 +18.41% 2% 68% 70%
458 18144868 +18.24% 2% 68% 70%
467 18142366 +18.22% 2% 68% 70%
498 18154549 +18.30% 1% 68% 69%
511 18134525 +18.17% 1% 69% 70%
Broadwell-EX: similar pattern as Skylake-EX.
batch score change zone_contention lru_contention total_contention
31 16703983 +0.00% 67% 7% 74%
53 18195393 +8.93% 43% 28% 71%
63 18288885 +9.49% 38% 33% 71%
73 18344329 +9.82% 35% 37% 72%
119 18535529 +10.96% 24% 46% 70%
127 18513596 +10.83% 23% 48% 71%
137 18514327 +10.84% 23% 48% 71%
165 18511840 +10.82% 22% 49% 71%
188 18593478 +11.31% 17% 53% 70%
223 18601667 +11.36% 17% 52% 69%
255 18774825 +12.40% 12% 58% 70%
267 18754781 +12.28% 9% 60% 69%
299 18892265 +13.10% 7% 63% 70%
320 18873812 +12.99% 8% 62% 70%
393 18891174 +13.09% 6% 64% 70%
424 18975108 +13.60% 6% 64% 70%
458 18932364 +13.34% 8% 62% 70%
467 18960891 +13.51% 5% 65% 70%
498 18944526 +13.41% 5% 64% 69%
511 18960839 +13.51% 5% 64% 69%
Skylake-EP: although increased batch reduced zone->lock contention, but
the effect is not as good as EX: zone->lock contention is still as high as
20% with a very high batch value instead of 1% on Skylake-EX or 5% on
Broadwell-EX. Also, total_contention actually decreased with a higher
batch but that doesn't translate to performance increase.
batch score change zone_contention lru_contention total_contention
31 9554867 +0.00% 66% 3% 69%
53 9855486 +3.15% 63% 3% 66%
63 9980145 +4.45% 62% 4% 66%
73 10092774 +5.63% 62% 5% 67%
119 10310061 +7.90% 45% 19% 64%
127 10342019 +8.24% 42% 19% 61%
137 10358182 +8.41% 42% 21% 63%
165 10397060 +8.81% 37% 24% 61%
188 10341808 +8.24% 34% 26% 60%
223 10349135 +8.31% 31% 27% 58%
255 10327189 +8.08% 28% 29% 57%
267 10344204 +8.26% 27% 29% 56%
299 10325043 +8.06% 25% 30% 55%
320 10310325 +7.91% 25% 31% 56%
393 10293274 +7.73% 21% 31% 52%
424 10311099 +7.91% 21% 32% 53%
458 10321375 +8.02% 21% 32% 53%
467 10303881 +7.84% 21% 32% 53%
498 10332462 +8.14% 20% 33% 53%
511 10325016 +8.06% 20% 32% 52%
Broadwell-EP: zone->lock and lru lock had an agreement to make sure
performance doesn't increase and they successfully managed to keep total
contention at 70%.
batch score change zone_contention lru_contention total_contention
31 10121178 +0.00% 19% 50% 69%
53 10142366 +0.21% 6% 63% 69%
63 10117984 -0.03% 11% 58% 69%
73 10123330 +0.02% 7% 63% 70%
119 10108791 -0.12% 2% 67% 69%
127 10166074 +0.44% 3% 66% 69%
137 10141574 +0.20% 3% 66% 69%
165 10154499 +0.33% 2% 68% 70%
188 10124921 +0.04% 2% 67% 69%
223 10137399 +0.16% 2% 67% 69%
255 10143289 +0.22% 0% 68% 68%
267 10123535 +0.02% 1% 68% 69%
299 10140952 +0.20% 0% 68% 68%
320 10163170 +0.41% 0% 68% 68%
393 10000633 -1.19% 0% 69% 69%
424 10087998 -0.33% 0% 69% 69%
458 10187116 +0.65% 0% 69% 69%
467 10146790 +0.25% 0% 69% 69%
498 10197958 +0.76% 0% 69% 69%
511 10152326 +0.31% 0% 69% 69%
Haswell-EP: similar to Broadwell-EP.
batch score change zone_contention lru_contention total_contention
31 10442205 +0.00% 14% 48% 62%
53 10442255 +0.00% 5% 57% 62%
63 10452059 +0.09% 6% 57% 63%
73 10482349 +0.38% 5% 59% 64%
119 10454644 +0.12% 3% 60% 63%
127 10431514 -0.10% 3% 59% 62%
137 10423785 -0.18% 3% 60% 63%
165 10481216 +0.37% 2% 61% 63%
188 10448755 +0.06% 2% 61% 63%
223 10467144 +0.24% 2% 61% 63%
255 10480215 +0.36% 2% 61% 63%
267 10484279 +0.40% 2% 61% 63%
299 10466450 +0.23% 2% 61% 63%
320 10452578 +0.10% 2% 61% 63%
393 10499678 +0.55% 1% 62% 63%
424 10481454 +0.38% 1% 62% 63%
458 10473562 +0.30% 1% 62% 63%
467 10484269 +0.40% 0% 62% 62%
498 10505599 +0.61% 0% 62% 62%
511 10483395 +0.39% 0% 62% 62%
Westmere-EP: contention is pretty small so not interesting. Note too high
a batch value could hurt performance.
batch score change zone_contention lru_contention total_contention
31 4831523 +0.00% 2% 3% 5%
53 4834086 +0.05% 2% 4% 6%
63 4834262 +0.06% 2% 3% 5%
73 4832851 +0.03% 2% 4% 6%
119 4830534 -0.02% 1% 3% 4%
127 4827461 -0.08% 1% 4% 5%
137 4827459 -0.08% 1% 3% 4%
165 4820534 -0.23% 0% 4% 4%
188 4817947 -0.28% 0% 3% 3%
223 4809671 -0.45% 0% 3% 3%
255 4802463 -0.60% 0% 4% 4%
267 4801634 -0.62% 0% 3% 3%
299 4798047 -0.69% 0% 3% 3%
320 4793084 -0.80% 0% 3% 3%
393 4785877 -0.94% 0% 3% 3%
424 4782911 -1.01% 0% 3% 3%
458 4779346 -1.08% 0% 3% 3%
467 4780306 -1.06% 0% 3% 3%
498 4780589 -1.05% 0% 3% 3%
511 4773724 -1.20% 0% 3% 3%
Skylake-Desktop: similar to Westmere-EP, nothing interesting.
batch score change zone_contention lru_contention total_contention
31 3906608 +0.00% 2% 3% 5%
53 3940164 +0.86% 2% 3% 5%
63 3937289 +0.79% 2% 3% 5%
73 3940201 +0.86% 2% 3% 5%
119 3933240 +0.68% 2% 3% 5%
127 3930514 +0.61% 2% 4% 6%
137 3938639 +0.82% 0% 3% 3%
165 3908755 +0.05% 0% 3% 3%
188 3905621 -0.03% 0% 3% 3%
223 3903015 -0.09% 0% 4% 4%
255 3889480 -0.44% 0% 3% 3%
267 3891669 -0.38% 0% 4% 4%
299 3898728 -0.20% 0% 4% 4%
320 3894547 -0.31% 0% 4% 4%
393 3875137 -0.81% 0% 4% 4%
424 3874521 -0.82% 0% 3% 3%
458 3880432 -0.67% 0% 4% 4%
467 3888715 -0.46% 0% 3% 3%
498 3888633 -0.46% 0% 4% 4%
511 3875305 -0.80% 0% 5% 5%
Haswell-Desktop: zone->lock is pretty low as other desktops, though lru
contention is higher than other desktops.
batch score change zone_contention lru_contention total_contention
31 3511158 +0.00% 2% 5% 7%
53 3555445 +1.26% 2% 6% 8%
63 3561082 +1.42% 2% 6% 8%
73 3547218 +1.03% 2% 6% 8%
119 3571319 +1.71% 1% 7% 8%
127 3549375 +1.09% 0% 6% 6%
137 3560233 +1.40% 0% 6% 6%
165 3555176 +1.25% 2% 6% 8%
188 3551501 +1.15% 0% 8% 8%
223 3531462 +0.58% 0% 7% 7%
255 3570400 +1.69% 0% 7% 7%
267 3532235 +0.60% 1% 8% 9%
299 3562326 +1.46% 0% 6% 6%
320 3553569 +1.21% 0% 8% 8%
393 3539519 +0.81% 0% 7% 7%
424 3549271 +1.09% 0% 8% 8%
458 3528885 +0.50% 0% 8% 8%
467 3526554 +0.44% 0% 7% 7%
498 3525302 +0.40% 0% 9% 9%
511 3527556 +0.47% 0% 8% 8%
Sandybridge-Desktop: the 0% contention isn't accurate but caused by
dropped fractional part. Since multiple contention path's contentions
are all under 1% here, with some arithmetic operations like add, the
final deviation could be as large as 3%.
batch score change zone_contention lru_contention total_contention
31 1744495 +0.00% 0% 0% 0%
53 1755341 +0.62% 0% 0% 0%
63 1758469 +0.80% 0% 0% 0%
73 1759626 +0.87% 0% 0% 0%
119 1770417 +1.49% 0% 0% 0%
127 1768252 +1.36% 0% 0% 0%
137 1767848 +1.34% 0% 0% 0%
165 1765088 +1.18% 0% 0% 0%
188 1766918 +1.29% 0% 0% 0%
223 1767866 +1.34% 0% 0% 0%
255 1768074 +1.35% 0% 0% 0%
267 1763187 +1.07% 0% 0% 0%
299 1765620 +1.21% 0% 0% 0%
320 1767603 +1.32% 0% 0% 0%
393 1764612 +1.15% 0% 0% 0%
424 1758476 +0.80% 0% 0% 0%
458 1758593 +0.81% 0% 0% 0%
467 1757915 +0.77% 0% 0% 0%
498 1753363 +0.51% 0% 0% 0%
511 1755548 +0.63% 0% 0% 0%
Phase two test results:
Note: all percent change is against base(batch=31).
ebizzy.throughput (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 2410037±7% 2600451±2% +7.9% 2602878 +8.0%
lkp-bdw-ex1 1493328 1489243 -0.3% 1492145 -0.1%
lkp-skl-2sp2 1329674 1345891 +1.2% 1351056 +1.6%
lkp-bdw-ep2 711511 711511 0.0% 710708 -0.1%
lkp-wsm-ep2 75750 75528 -0.3% 75441 -0.4%
lkp-skl-d01 264126 262791 -0.5% 264113 +0.0%
lkp-hsw-d01 176601 176328 -0.2% 176368 -0.1%
lkp-sb02 98937 98937 +0.0% 99030 +0.1%
kbuild.buildtime (less is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 107.00 107.67 +0.6% 107.11 +0.1%
lkp-bdw-ex1 97.33 97.33 +0.0% 97.42 +0.1%
lkp-skl-2sp2 180.00 179.83 -0.1% 179.83 -0.1%
lkp-bdw-ep2 178.17 179.17 +0.6% 177.50 -0.4%
lkp-wsm-ep2 737.00 738.00 +0.1% 738.00 +0.1%
lkp-skl-d01 642.00 653.00 +1.7% 653.00 +1.7%
lkp-hsw-d01 1310.00 1316.00 +0.5% 1311.00 +0.1%
netperf/TCP_STREAM.Throughput_total_Mbps (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 948790 947144 -0.2% 948333 -0.0%
lkp-bdw-ex1 904224 904366 +0.0% 904926 +0.1%
lkp-skl-2sp2 239731 239607 -0.1% 239565 -0.1%
lk-bdw-ep2 365764 365933 +0.0% 365951 +0.1%
lkp-wsm-ep2 93736 93803 +0.1% 93808 +0.1%
lkp-skl-d01 77314 77303 -0.0% 77375 +0.1%
lkp-hsw-d01 58617 60387 +3.0% 60208 +2.7%
lkp-sb02 29990 30137 +0.5% 30103 +0.4%
oltp.transactions (higer is better)
machine batch=31 batch=63 batch=127
lkp-bdw-ex1 9073276 9100377 +0.3% 9036344 -0.4%
lkp-skl-2sp2 8898717 8852054 -0.5% 8894459 -0.0%
lkp-bdw-ep2 13426155 13384654 -0.3% 13333637 -0.7%
lkp-hsw-ep2 13146314 13232784 +0.7% 13193163 +0.4%
lkp-wsm-ep2 5035355 5019348 -0.3% 5033418 -0.0%
lkp-skl-d01 418485 4413339 -0.1% 4419039 +0.0%
lkp-hsw-d01 3517817±5% 3396120±3% -3.5% 3455138±3% -1.8%
pigz.throughput (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.513e+08 1.507e+08 -0.4% 1.511e+08 -0.2%
lkp-bdw-ex1 2.060e+08 2.052e+08 -0.4% 2.044e+08 -0.8%
lkp-skl-2sp2 8.836e+08 8.845e+08 +0.1% 8.836e+08 -0.0%
lkp-bdw-ep2 8.275e+08 8.464e+08 +2.3% 8.330e+08 +0.7%
lkp-wsm-ep2 2.224e+08 2.221e+08 -0.2% 2.218e+08 -0.3%
lkp-skl-d01 1.177e+08 1.177e+08 -0.0% 1.176e+08 -0.1%
lkp-hsw-d01 1.154e+08 1.154e+08 +0.1% 1.154e+08 -0.0%
lkp-sb02 0.633e+08 0.633e+08 +0.1% 0.633e+08 +0.0%
will-it-scale.malloc1.processes (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 620181 620484 +0.0% 620240 +0.0%
lkp-bdw-ex1 1403610 1401201 -0.2% 1417900 +1.0%
lkp-skl-2sp2 1288097 1284145 -0.3% 1283907 -0.3%
lkp-bdw-ep2 1427879 1427675 -0.0% 1428266 +0.0%
lkp-hsw-ep2 1362546 1353965 -0.6% 1354759 -0.6%
lkp-wsm-ep2 2099657 2107576 +0.4% 2100226 +0.0%
lkp-skl-d01 1476835 1476358 -0.0% 1474487 -0.2%
lkp-hsw-d01 1308810 1303429 -0.4% 1301299 -0.6%
lkp-sb02 589286 589284 -0.0% 588101 -0.2%
will-it-scale.malloc1.threads (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 21289 21125 -0.8% 21241 -0.2%
lkp-bdw-ex1 28114 28089 -0.1% 28007 -0.4%
lkp-skl-2sp2 91866 91946 +0.1% 92723 +0.9%
lkp-bdw-ep2 37637 37501 -0.4% 37317 -0.9%
lkp-hsw-ep2 43673 43590 -0.2% 43754 +0.2%
lkp-wsm-ep2 28577 28298 -1.0% 28545 -0.1%
lkp-skl-d01 175277 173343 -1.1% 173082 -1.3%
lkp-hsw-d01 130303 129566 -0.6% 129250 -0.8%
lkp-sb02 113742±3% 116911 +2.8% 116417±3% +2.4%
will-it-scale.malloc2.processes (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.206e+09 1.206e+09 -0.0% 1.206e+09 +0.0%
lkp-bdw-ex1 1.319e+09 1.319e+09 -0.0% 1.319e+09 +0.0%
lkp-skl-2sp2 8.000e+08 8.021e+08 +0.3% 7.995e+08 -0.1%
lkp-bdw-ep2 6.582e+08 6.634e+08 +0.8% 6.513e+08 -1.1%
lkp-hsw-ep2 6.671e+08 6.669e+08 -0.0% 6.665e+08 -0.1%
lkp-wsm-ep2 1.805e+08 1.806e+08 +0.0% 1.804e+08 -0.1%
lkp-skl-d01 1.611e+08 1.611e+08 -0.0% 1.610e+08 -0.0%
lkp-hsw-d01 1.333e+08 1.332e+08 -0.0% 1.332e+08 -0.0%
lkp-sb02 82485104 82478206 -0.0% 82473546 -0.0%
will-it-scale.malloc2.threads (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.574e+09 1.574e+09 -0.0% 1.574e+09 -0.0%
lkp-bdw-ex1 1.737e+09 1.737e+09 +0.0% 1.737e+09 -0.0%
lkp-skl-2sp2 9.161e+08 9.162e+08 +0.0% 9.181e+08 +0.2%
lkp-bdw-ep2 7.856e+08 8.015e+08 +2.0% 8.113e+08 +3.3%
lkp-hsw-ep2 6.908e+08 6.904e+08 -0.1% 6.907e+08 -0.0%
lkp-wsm-ep2 2.409e+08 2.409e+08 +0.0% 2.409e+08 -0.0%
lkp-skl-d01 1.199e+08 1.199e+08 -0.0% 1.199e+08 -0.0%
lkp-hsw-d01 1.029e+08 1.029e+08 -0.0% 1.029e+08 +0.0%
lkp-sb02 68081213 68061423 -0.0% 68076037 -0.0%
will-it-scale.page_fault2.processes (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 14509125±4% 16472364 +13.5% 17123117 +18.0%
lkp-bdw-ex1 14736381 16196588 +9.9% 16364011 +11.0%
lkp-skl-2sp2 6354925 6435444 +1.3% 6436644 +1.3%
lkp-bdw-ep2 8749584 8834422 +1.0% 8827179 +0.9%
lkp-hsw-ep2 8762591 8845920 +1.0% 8825697 +0.7%
lkp-wsm-ep2 3036083 3030428 -0.2% 3021741 -0.5%
lkp-skl-d01 2307834 2304731 -0.1% 2286142 -0.9%
lkp-hsw-d01 1806237 1800786 -0.3% 1795943 -0.6%
lkp-sb02 842616 837844 -0.6% 833921 -1.0%
will-it-scale.page_fault2.threads
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1623294 1615132±2% -0.5% 1656777 +2.1%
lkp-bdw-ex1 1995714 2025948 +1.5% 2113753±3% +5.9%
lkp-skl-2sp2 2346708 2415591 +2.9% 2416919 +3.0%
lkp-bdw-ep2 2342564 2344882 +0.1% 2300206 -1.8%
lkp-hsw-ep2 1820658 1831681 +0.6% 1844057 +1.3%
lkp-wsm-ep2 1725482 1733774 +0.5% 1740517 +0.9%
lkp-skl-d01 1832833 1823628 -0.5% 1806489 -1.4%
lkp-hsw-d01 1427913 1427287 -0.0% 1420226 -0.5%
lkp-sb02 750626 748615 -0.3% 746621 -0.5%
will-it-scale.page_fault3.processes (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 24382726 24400317 +0.1% 24668774 +1.2%
lkp-bdw-ex1 35399750 35683124 +0.8% 35829492 +1.2%
lkp-skl-2sp2 28136820 28068248 -0.2% 28147989 +0.0%
lkp-bdw-ep2 37269077 37459490 +0.5% 37373073 +0.3%
lkp-hsw-ep2 36224967 36114085 -0.3% 36104908 -0.3%
lkp-wsm-ep2 16820457 16911005 +0.5% 16968596 +0.9%
lkp-skl-d01 7721138 7725904 +0.1% 7756740 +0.5%
lkp-hsw-d01 7611979 7650928 +0.5% 7651323 +0.5%
lkp-sb02 3781546 3796502 +0.4% 3796827 +0.4%
will-it-scale.page_fault3.threads (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1865820±3% 1900917±2% +1.9% 1826245±4% -2.1%
lkp-bdw-ex1 3094060 3148326 +1.8% 3150036 +1.8%
lkp-skl-2sp2 3952940 3953898 +0.0% 3989360 +0.9%
lkp-bdw-ep2 3420373±3% 3643964 +6.5% 3644910±5% +6.6%
lkp-hsw-ep2 2609635±2% 2582310±3% -1.0% 2780459 +6.5%
lkp-wsm-ep2 4395001 4417196 +0.5% 4432499 +0.9%
lkp-skl-d01 5363977 5400003 +0.7% 5411370 +0.9%
lkp-hsw-d01 5274131 5311294 +0.7% 5319359 +0.9%
lkp-sb02 2917314 2913004 -0.1% 2935286 +0.6%
will-it-scale.read1.processes (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 73762279±14% 69322519±10% -6.0% 69349855±13% -6.0% (result unstable)
lkp-bdw-ex1 1.701e+08 1.704e+08 +0.1% 1.705e+08 +0.2%
lkp-skl-2sp2 63111570 63113953 +0.0% 63836573 +1.1%
lkp-bdw-ep2 79247409 79424610 +0.2% 78012656 -1.6%
lkp-hsw-ep2 67677026 68308800 +0.9% 67539106 -0.2%
lkp-wsm-ep2 13339630 13939817 +4.5% 13766865 +3.2%
lkp-skl-d01 10969487 10972650 +0.0% no data
lkp-hsw-d01 9857342±2% 10080592±2% +2.3% 10131560 +2.8%
lkp-sb02 5189076 5197473 +0.2% 5163253 -0.5%
will-it-scale.read1.threads (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 62468045±12% 73666726±7% +17.9% 79553123±12% +27.4% (result unstable)
lkp-bdw-ex1 1.62e+08 1.624e+08 +0.3% 1.614e+08 -0.3%
lkp-skl-2sp2 58319780 59181032 +1.5% 59821353 +2.6%
lkp-bdw-ep2 74057992 75698171 +2.2% 74990869 +1.3%
lkp-hsw-ep2 63672959 63639652 -0.1% 64387051 +1.1%
lkp-wsm-ep2 13489943 13526058 +0.3% 13259032 -1.7%
lkp-skl-d01 10297906 10338796 +0.4% 10407328 +1.1%
lkp-hsw-d01 9636721 9667376 +0.3% 9341147 -3.1%
lkp-sb02 4801938 4804496 +0.1% 4802290 +0.0%
will-it-scale.write1.processes (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.111e+08 1.104e+08±2% -0.7% 1.122e+08±2% +1.0%
lkp-bdw-ex1 1.392e+08 1.399e+08 +0.5% 1.397e+08 +0.4%
lkp-skl-2sp2 59369233 58994841 -0.6% 58715168 -1.1%
lkp-bdw-ep2 61820979 CPU throttle 63593123 +2.9%
lkp-hsw-ep2 57897587 57435605 -0.8% 56347450 -2.7%
lkp-wsm-ep2 7814203 7918017±2% +1.3% 7669068 -1.9%
lkp-skl-d01 8886557 8971422 +1.0% 8818366 -0.8%
lkp-hsw-d01 9171001±5% 9189915 +0.2% 9483909 +3.4%
lkp-sb02 4475406 4475294 -0.0% 4501756 +0.6%
will-it-scale.write1.threads (higer is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1.058e+08 1.055e+08±2% -0.2% 1.065e+08 +0.7%
lkp-bdw-ex1 1.316e+08 1.300e+08 -1.2% 1.308e+08 -0.6%
lkp-skl-2sp2 54492421 56086678 +2.9% 55975657 +2.7%
lkp-bdw-ep2 59360449 59003957 -0.6% 58101262 -2.1%
lkp-hsw-ep2 53346346±2% 52530876 -1.5% 52902487 -0.8%
lkp-wsm-ep2 7774006 7800092±2% +0.3% 7558833 -2.8%
lkp-skl-d01 8346174 8235695 -1.3% no data
lkp-hsw-d01 8636244 8655731 +0.2% 8658868 +0.3%
lkp-sb02 4181820 4204107 +0.5% 4182992 +0.0%
vm-scalability.anon-r-rand.throughput (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 11933873±3% 12356544±2% +3.5% 12188624 +2.1%
lkp-bdw-ex1 7114424±2% 7330949±2% +3.0% 7392419 +3.9%
lkp-skl-2sp2 6773277±5% 6492332±8% -4.1% 6543962 -3.4%
lkp-bdw-ep2 7133846±4% 7233508 +1.4% 7013518±3% -1.7%
lkp-hsw-ep2 4576626 4527098 -1.1% 4551679 -0.5%
lkp-wsm-ep2 2583599 2592492 +0.3% 2588039 +0.2%
lkp-hsw-d01 998199±2% 1028311 +3.0% 1006460±2% +0.8%
lkp-sb02 570572 567854 -0.5% 568449 -0.4%
vm-scalability.anon-r-rand-mt.throughput (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1789419 1787830 -0.1% 1788208 -0.1%
lkp-bdw-ex1 3492595±2% 3554966±2% +1.8% 3558835±3% +1.9%
lkp-skl-2sp2 3856238±2% 3975403±4% +3.1% 3994600 +3.6%
lkp-bdw-ep2 3726963±11% 3809292±6% +2.2% 3871924±4% +3.9%
lkp-hsw-ep2 2131760±3% 2033578±4% -4.6% 2130727±6% -0.0%
lkp-wsm-ep2 2369731 2368384 -0.1% 2370252 +0.0%
lkp-skl-d01 1207128 1206220 -0.1% 1205801 -0.1%
lkp-hsw-d01 964317 992329±2% +2.9% 992099±2% +2.9%
lkp-sb02 567137 567346 +0.0% 566144 -0.2%
vm-scalability.lru-file-mmap-read.throughput (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 19560469±6% 23018999 +17.7% 23418800 +19.7%
lkp-bdw-ex1 17769135±14% 26141676±3% +47.1% 26284723±5% +47.9%
lkp-skl-2sp2 14056512 13578884 -3.4% 13146214 -6.5%
lkp-bdw-ep2 15336542 14737654 -3.9% 14088159 -8.1%
lkp-hsw-ep2 16275498 15756296 -3.2% 15018090 -7.7%
lkp-wsm-ep2 11272160 11237231 -0.3% 11310047 +0.3%
lkp-skl-d01 7322119 7324569 +0.0% 7184148 -1.9%
lkp-hsw-d01 6449234 6404542 -0.7% 6356141 -1.4%
lkp-sb02 3517943 3520668 +0.1% 3527309 +0.3%
vm-scalability.lru-file-mmap-read-rand.throughput (higher is better)
machine batch=31 batch=63 batch=127
lkp-skl-4sp1 1689052 1697553 +0.5% 1698726 +0.6%
lkp-bdw-ex1 1675246 1699764 +1.5% 1712226 +2.2%
lkp-skl-2sp2 1800533 1799749 -0.0% 1800581 +0.0%
lkp-bdw-ep2 1807422 1807758 +0.0% 1804932 -0.1%
lkp-hsw-ep2 1809807 1808781 -0.1% 1807811 -0.1%
lkp-wsm-ep2 1800198 1802434 +0.1% 1801236 +0.1%
lkp-skl-d01 696689 695537 -0.2% 694106 -0.4%
lkp-hsw-d01 698364 698666 +0.0% 696686 -0.2%
lkp-sb02 258939 258787 -0.1% 258199 -0.3%
Link: http://lkml.kernel.org/r/20180711055855.29072-1-aaron.lu@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Suggested-by: Dave Hansen <dave.hansen@intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:49:14 +00:00
|
|
|
/* But no more than a meg. */
|
|
|
|
if (batch * PAGE_SIZE > 1024 * 1024)
|
|
|
|
batch = (1024 * 1024) / PAGE_SIZE;
|
2005-06-22 00:14:47 +00:00
|
|
|
batch /= 4; /* We effectively *= 4 below */
|
|
|
|
if (batch < 1)
|
|
|
|
batch = 1;
|
|
|
|
|
|
|
|
/*
|
2005-12-04 02:55:25 +00:00
|
|
|
* Clamp the batch to a 2^n - 1 value. Having a power
|
|
|
|
* of 2 value was found to be more likely to have
|
|
|
|
* suboptimal cache aliasing properties in some cases.
|
2005-06-22 00:14:47 +00:00
|
|
|
*
|
2005-12-04 02:55:25 +00:00
|
|
|
* For example if 2 tasks are alternately allocating
|
|
|
|
* batches of pages, one task can end up with a lot
|
|
|
|
* of pages of one half of the possible page colors
|
|
|
|
* and the other with pages of the other colors.
|
2005-06-22 00:14:47 +00:00
|
|
|
*/
|
2009-05-06 23:03:02 +00:00
|
|
|
batch = rounddown_pow_of_two(batch + batch/2) - 1;
|
2005-10-30 01:15:47 +00:00
|
|
|
|
2005-06-22 00:14:47 +00:00
|
|
|
return batch;
|
nommu: clamp zone_batchsize() to 0 under NOMMU conditions
Clamp zone_batchsize() to 0 under NOMMU conditions to stop
free_hot_cold_page() from queueing and batching frees.
The problem is that under NOMMU conditions it is really important to be
able to allocate large contiguous chunks of memory, but when munmap() or
exit_mmap() releases big stretches of memory, return of these to the buddy
allocator can be deferred, and when it does finally happen, it can be in
small chunks.
Whilst the fragmentation this incurs isn't so much of a problem under MMU
conditions as userspace VM is glued together from individual pages with
the aid of the MMU, it is a real problem if there isn't an MMU.
By clamping the page freeing queue size to 0, pages are returned to the
allocator immediately, and the buddy detector is more likely to be able to
glue them together into large chunks immediately, and fragmentation is
less likely to occur.
By disabling batching of frees, and by turning off the trimming of excess
space during boot, Coldfire can manage to boot.
Reported-by: Lanttor Guo <lanttor.guo@freescale.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Lanttor Guo <lanttor.guo@freescale.com>
Cc: Greg Ungerer <gerg@snapgear.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-05-06 23:03:03 +00:00
|
|
|
|
|
|
|
#else
|
|
|
|
/* The deferral and batching of frees should be suppressed under NOMMU
|
|
|
|
* conditions.
|
|
|
|
*
|
|
|
|
* The problem is that NOMMU needs to be able to allocate large chunks
|
|
|
|
* of contiguous memory as there's no hardware page translation to
|
|
|
|
* assemble apparent contiguous memory from discontiguous pages.
|
|
|
|
*
|
|
|
|
* Queueing large contiguous runs of pages for batching, however,
|
|
|
|
* causes the pages to actually be freed in smaller chunks. As there
|
|
|
|
* can be a significant delay between the individual batches being
|
|
|
|
* recycled, this leads to the once large chunks of space being
|
|
|
|
* fragmented and becoming unavailable for high-order allocations.
|
|
|
|
*/
|
|
|
|
return 0;
|
|
|
|
#endif
|
2005-06-22 00:14:47 +00:00
|
|
|
}
|
|
|
|
|
2013-07-03 22:01:31 +00:00
|
|
|
/*
|
2020-12-15 03:10:50 +00:00
|
|
|
* pcp->high and pcp->batch values are related and generally batch is lower
|
|
|
|
* than high. They are also related to pcp->count such that count is lower
|
|
|
|
* than high, and as soon as it reaches high, the pcplist is flushed.
|
2013-07-03 22:01:31 +00:00
|
|
|
*
|
2020-12-15 03:10:50 +00:00
|
|
|
* However, guaranteeing these relations at all times would require e.g. write
|
|
|
|
* barriers here but also careful usage of read barriers at the read side, and
|
|
|
|
* thus be prone to error and bad for performance. Thus the update only prevents
|
|
|
|
* store tearing. Any new users of pcp->batch and pcp->high should ensure they
|
|
|
|
* can cope with those fields changing asynchronously, and fully trust only the
|
|
|
|
* pcp->count field on the local CPU with interrupts disabled.
|
2013-07-03 22:01:31 +00:00
|
|
|
*
|
|
|
|
* mutex_is_locked(&pcp_batch_high_lock) required when calling this function
|
|
|
|
* outside of boot time (or some other assurance that no concurrent updaters
|
|
|
|
* exist).
|
|
|
|
*/
|
|
|
|
static void pageset_update(struct per_cpu_pages *pcp, unsigned long high,
|
|
|
|
unsigned long batch)
|
|
|
|
{
|
2020-12-15 03:10:50 +00:00
|
|
|
WRITE_ONCE(pcp->batch, batch);
|
|
|
|
WRITE_ONCE(pcp->high, high);
|
2013-07-03 22:01:31 +00:00
|
|
|
}
|
|
|
|
|
2013-07-03 22:01:35 +00:00
|
|
|
static void pageset_init(struct per_cpu_pageset *p)
|
2005-06-22 00:15:00 +00:00
|
|
|
{
|
|
|
|
struct per_cpu_pages *pcp;
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
int migratetype;
|
2005-06-22 00:15:00 +00:00
|
|
|
|
2005-10-26 08:58:59 +00:00
|
|
|
memset(p, 0, sizeof(*p));
|
|
|
|
|
2008-02-05 06:29:19 +00:00
|
|
|
pcp = &p->pcp;
|
page-allocator: split per-cpu list into one-list-per-migrate-type
The following two patches remove searching in the page allocator fast-path
by maintaining multiple free-lists in the per-cpu structure. At the time
the search was introduced, increasing the per-cpu structures would waste a
lot of memory as per-cpu structures were statically allocated at
compile-time. This is no longer the case.
The patches are as follows. They are based on mmotm-2009-08-27.
Patch 1 adds multiple lists to struct per_cpu_pages, one per
migratetype that can be stored on the PCP lists.
Patch 2 notes that the pcpu drain path check empty lists multiple times. The
patch reduces the number of checks by maintaining a count of free
lists encountered. Lists containing pages will then free multiple
pages in batch
The patches were tested with kernbench, netperf udp/tcp, hackbench and
sysbench. The netperf tests were not bound to any CPU in particular and
were run such that the results should be 99% confidence that the reported
results are within 1% of the estimated mean. sysbench was run with a
postgres background and read-only tests. Similar to netperf, it was run
multiple times so that it's 99% confidence results are within 1%. The
patches were tested on x86, x86-64 and ppc64 as
x86: Intel Pentium D 3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.34% to 2.28% gain
netperf-tcp - 0.45% to 1.22% gain
hackbench - Small variances, very close to noise
sysbench - Very small gains
x86-64: AMD Phenom 9950 1.3GHz with 8G RAM (no-brand machine)
kernbench - No significant difference, variance well within noise
netperf-udp - 1.83% to 10.42% gains
netperf-tcp - No conclusive until buffer >= PAGE_SIZE
4096 +15.83%
8192 + 0.34% (not significant)
16384 + 1%
hackbench - Small gains, very close to noise
sysbench - 0.79% to 1.6% gain
ppc64: PPC970MP 2.5GHz with 10GB RAM (it's a terrasoft powerstation)
kernbench - No significant difference, variance well within noise
netperf-udp - 2-3% gain for almost all buffer sizes tested
netperf-tcp - losses on small buffers, gains on larger buffers
possibly indicates some bad caching effect.
hackbench - No significant difference
sysbench - 2-4% gain
This patch:
Currently the per-cpu page allocator searches the PCP list for pages of
the correct migrate-type to reduce the possibility of pages being
inappropriate placed from a fragmentation perspective. This search is
potentially expensive in a fast-path and undesirable. Splitting the
per-cpu list into multiple lists increases the size of a per-cpu structure
and this was potentially a major problem at the time the search was
introduced. These problem has been mitigated as now only the necessary
number of structures is allocated for the running system.
This patch replaces a list search in the per-cpu allocator with one list
per migrate type. The potential snag with this approach is when bulk
freeing pages. We round-robin free pages based on migrate type which has
little bearing on the cache hotness of the page and potentially checks
empty lists repeatedly in the event the majority of PCP pages are of one
type.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:03:19 +00:00
|
|
|
for (migratetype = 0; migratetype < MIGRATE_PCPTYPES; migratetype++)
|
|
|
|
INIT_LIST_HEAD(&pcp->lists[migratetype]);
|
2005-06-22 00:15:00 +00:00
|
|
|
|
2020-12-15 03:10:47 +00:00
|
|
|
/*
|
|
|
|
* Set batch and high values safe for a boot pageset. A true percpu
|
|
|
|
* pageset's initialization will update them subsequently. Here we don't
|
|
|
|
* need to be as careful as pageset_update() as nobody can access the
|
|
|
|
* pageset yet.
|
|
|
|
*/
|
2020-12-15 03:10:53 +00:00
|
|
|
pcp->high = BOOT_PAGESET_HIGH;
|
|
|
|
pcp->batch = BOOT_PAGESET_BATCH;
|
2013-07-03 22:01:35 +00:00
|
|
|
}
|
|
|
|
|
2020-12-15 03:11:12 +00:00
|
|
|
static void __zone_set_pageset_high_and_batch(struct zone *zone, unsigned long high,
|
mm, page_alloc: disable pcplists during memory offline
Memory offlining relies on page isolation to guarantee a forward progress
because pages cannot be reused while they are isolated. But the page
isolation itself doesn't prevent from races while freed pages are stored
on pcp lists and thus can be reused. This can be worked around by
repeated draining of pcplists, as done by commit 968318261221
("mm/memory_hotplug: drain per-cpu pages again during memory offline").
David and Michal would prefer that this race was closed in a way that
callers of page isolation who need stronger guarantees don't need to
repeatedly drain. David suggested disabling pcplists usage completely
during page isolation, instead of repeatedly draining them.
To achieve this without adding special cases in alloc/free fastpath, we
can use the same approach as boot pagesets - when pcp->high is 0, any
pcplist addition will be immediately flushed.
The race can thus be closed by setting pcp->high to 0 and draining
pcplists once, before calling start_isolate_page_range(). The draining
will serialize after processes that already disabled interrupts and read
the old value of pcp->high in free_unref_page_commit(), and processes that
have not yet disabled interrupts, will observe pcp->high == 0 when they
are rescheduled, and skip pcplists. This guarantees no stray pages on
pcplists in zones where isolation happens.
This patch thus adds zone_pcp_disable() and zone_pcp_enable() functions
that page isolation users can call before start_isolate_page_range() and
after unisolating (or offlining) the isolated pages.
Also, drain_all_pages() is optimized to only execute on cpus where
pcplists are not empty. The check can however race with a free to pcplist
that has not yet increased the pcp->count from 0 to 1. Thus make the
drain optionally skip the racy check and drain on all cpus, and use this
option in zone_pcp_disable().
As we have to avoid external updates to high and batch while pcplists are
disabled, we take pcp_batch_high_lock in zone_pcp_disable() and release it
in zone_pcp_enable(). This also synchronizes multiple users of
zone_pcp_disable()/enable().
Currently the only user of this functionality is offline_pages().
[vbabka@suse.cz: add comment, per David]
Link: https://lkml.kernel.org/r/527480ef-ed72-e1c1-52a0-1c5b0113df45@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Suggested-by: David Hildenbrand <david@redhat.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:59 +00:00
|
|
|
unsigned long batch)
|
|
|
|
{
|
|
|
|
struct per_cpu_pageset *p;
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
p = per_cpu_ptr(zone->pageset, cpu);
|
|
|
|
pageset_update(&p->pcp, high, batch);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-01-08 09:00:40 +00:00
|
|
|
/*
|
2020-12-15 03:10:43 +00:00
|
|
|
* Calculate and set new high and batch values for all per-cpu pagesets of a
|
mm, page_alloc: clean up pageset high and batch update
Patch series "disable pcplists during memory offline", v3.
As per the discussions [1] [2] this is an attempt to implement David's
suggestion that page isolation should disable pcplists to avoid races with
page freeing in progress. This is done without extra checks in fast
paths, as explained in Patch 9. The repeated draining done by [2] is then
no longer needed. Previous version (RFC) is at [3].
The RFC tried to hide pcplists disabling/enabling into page isolation, but
it wasn't completely possible, as memory offline does not unisolation.
Michal suggested an explicit API in [4] so that's the current
implementation and it seems indeed nicer.
Once we accept that page isolation users need to do explicit actions
around it depending on the needed guarantees, we can also IMHO accept that
the current pcplist draining can be also done by the callers, which is
more effective. After all, there are only two users of page isolation.
So patch 6 does effectively the same thing as Pavel proposed in [5], and
patch 7 implement stronger guarantees only for memory offline. If CMA
decides to opt-in to the stronger guarantee, it can be added later.
Patches 1-5 are preparatory cleanups for pcplist disabling.
Patchset was briefly tested in QEMU so that memory online/offline works,
but I haven't done a stress test that would prove the race fixed by [2] is
eliminated.
Note that patch 7 could be avoided if we instead adjusted page freeing in
shown in [6], but I believe the current implementation of disabling
pcplists is not too much complex, so I would prefer this instead of adding
new checks and longer irq-disabled section into page freeing hotpaths.
[1] https://lore.kernel.org/linux-mm/20200901124615.137200-1-pasha.tatashin@soleen.com/
[2] https://lore.kernel.org/linux-mm/20200903140032.380431-1-pasha.tatashin@soleen.com/
[3] https://lore.kernel.org/linux-mm/20200907163628.26495-1-vbabka@suse.cz/
[4] https://lore.kernel.org/linux-mm/20200909113647.GG7348@dhcp22.suse.cz/
[5] https://lore.kernel.org/linux-mm/20200904151448.100489-3-pasha.tatashin@soleen.com/
[6] https://lore.kernel.org/linux-mm/3d3b53db-aeaa-ff24-260b-36427fac9b1c@suse.cz/
[7] https://lore.kernel.org/linux-mm/20200922143712.12048-1-vbabka@suse.cz/
[8] https://lore.kernel.org/linux-mm/20201008114201.18824-1-vbabka@suse.cz/
This patch (of 7):
The updates to pcplists' high and batch values are handled by multiple
functions that make the calculations hard to follow. Consolidate
everything to pageset_set_high_and_batch() and remove pageset_set_batch()
and pageset_set_high() wrappers.
The only special case using one of the removed wrappers was:
build_all_zonelists_init()
setup_pageset()
pageset_set_batch()
which was hardcoding batch as 0, so we can just open-code a call to
pageset_update() with constant parameters instead.
No functional change.
Link: https://lkml.kernel.org/r/20201111092812.11329-1-vbabka@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Pankaj Gupta <pankaj.gupta@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:40 +00:00
|
|
|
* zone, based on the zone's size and the percpu_pagelist_fraction sysctl.
|
2006-01-08 09:00:40 +00:00
|
|
|
*/
|
2020-12-15 03:10:43 +00:00
|
|
|
static void zone_set_pageset_high_and_batch(struct zone *zone)
|
2013-07-03 22:01:38 +00:00
|
|
|
{
|
mm, page_alloc: clean up pageset high and batch update
Patch series "disable pcplists during memory offline", v3.
As per the discussions [1] [2] this is an attempt to implement David's
suggestion that page isolation should disable pcplists to avoid races with
page freeing in progress. This is done without extra checks in fast
paths, as explained in Patch 9. The repeated draining done by [2] is then
no longer needed. Previous version (RFC) is at [3].
The RFC tried to hide pcplists disabling/enabling into page isolation, but
it wasn't completely possible, as memory offline does not unisolation.
Michal suggested an explicit API in [4] so that's the current
implementation and it seems indeed nicer.
Once we accept that page isolation users need to do explicit actions
around it depending on the needed guarantees, we can also IMHO accept that
the current pcplist draining can be also done by the callers, which is
more effective. After all, there are only two users of page isolation.
So patch 6 does effectively the same thing as Pavel proposed in [5], and
patch 7 implement stronger guarantees only for memory offline. If CMA
decides to opt-in to the stronger guarantee, it can be added later.
Patches 1-5 are preparatory cleanups for pcplist disabling.
Patchset was briefly tested in QEMU so that memory online/offline works,
but I haven't done a stress test that would prove the race fixed by [2] is
eliminated.
Note that patch 7 could be avoided if we instead adjusted page freeing in
shown in [6], but I believe the current implementation of disabling
pcplists is not too much complex, so I would prefer this instead of adding
new checks and longer irq-disabled section into page freeing hotpaths.
[1] https://lore.kernel.org/linux-mm/20200901124615.137200-1-pasha.tatashin@soleen.com/
[2] https://lore.kernel.org/linux-mm/20200903140032.380431-1-pasha.tatashin@soleen.com/
[3] https://lore.kernel.org/linux-mm/20200907163628.26495-1-vbabka@suse.cz/
[4] https://lore.kernel.org/linux-mm/20200909113647.GG7348@dhcp22.suse.cz/
[5] https://lore.kernel.org/linux-mm/20200904151448.100489-3-pasha.tatashin@soleen.com/
[6] https://lore.kernel.org/linux-mm/3d3b53db-aeaa-ff24-260b-36427fac9b1c@suse.cz/
[7] https://lore.kernel.org/linux-mm/20200922143712.12048-1-vbabka@suse.cz/
[8] https://lore.kernel.org/linux-mm/20201008114201.18824-1-vbabka@suse.cz/
This patch (of 7):
The updates to pcplists' high and batch values are handled by multiple
functions that make the calculations hard to follow. Consolidate
everything to pageset_set_high_and_batch() and remove pageset_set_batch()
and pageset_set_high() wrappers.
The only special case using one of the removed wrappers was:
build_all_zonelists_init()
setup_pageset()
pageset_set_batch()
which was hardcoding batch as 0, so we can just open-code a call to
pageset_update() with constant parameters instead.
No functional change.
Link: https://lkml.kernel.org/r/20201111092812.11329-1-vbabka@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-2-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Pankaj Gupta <pankaj.gupta@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:40 +00:00
|
|
|
unsigned long new_high, new_batch;
|
|
|
|
|
|
|
|
if (percpu_pagelist_fraction) {
|
|
|
|
new_high = zone_managed_pages(zone) / percpu_pagelist_fraction;
|
|
|
|
new_batch = max(1UL, new_high / 4);
|
|
|
|
if ((new_high / 4) > (PAGE_SHIFT * 8))
|
|
|
|
new_batch = PAGE_SHIFT * 8;
|
|
|
|
} else {
|
|
|
|
new_batch = zone_batchsize(zone);
|
|
|
|
new_high = 6 * new_batch;
|
|
|
|
new_batch = max(1UL, 1 * new_batch);
|
|
|
|
}
|
2013-07-03 22:01:41 +00:00
|
|
|
|
2020-12-15 03:10:53 +00:00
|
|
|
if (zone->pageset_high == new_high &&
|
|
|
|
zone->pageset_batch == new_batch)
|
|
|
|
return;
|
|
|
|
|
|
|
|
zone->pageset_high = new_high;
|
|
|
|
zone->pageset_batch = new_batch;
|
|
|
|
|
mm, page_alloc: disable pcplists during memory offline
Memory offlining relies on page isolation to guarantee a forward progress
because pages cannot be reused while they are isolated. But the page
isolation itself doesn't prevent from races while freed pages are stored
on pcp lists and thus can be reused. This can be worked around by
repeated draining of pcplists, as done by commit 968318261221
("mm/memory_hotplug: drain per-cpu pages again during memory offline").
David and Michal would prefer that this race was closed in a way that
callers of page isolation who need stronger guarantees don't need to
repeatedly drain. David suggested disabling pcplists usage completely
during page isolation, instead of repeatedly draining them.
To achieve this without adding special cases in alloc/free fastpath, we
can use the same approach as boot pagesets - when pcp->high is 0, any
pcplist addition will be immediately flushed.
The race can thus be closed by setting pcp->high to 0 and draining
pcplists once, before calling start_isolate_page_range(). The draining
will serialize after processes that already disabled interrupts and read
the old value of pcp->high in free_unref_page_commit(), and processes that
have not yet disabled interrupts, will observe pcp->high == 0 when they
are rescheduled, and skip pcplists. This guarantees no stray pages on
pcplists in zones where isolation happens.
This patch thus adds zone_pcp_disable() and zone_pcp_enable() functions
that page isolation users can call before start_isolate_page_range() and
after unisolating (or offlining) the isolated pages.
Also, drain_all_pages() is optimized to only execute on cpus where
pcplists are not empty. The check can however race with a free to pcplist
that has not yet increased the pcp->count from 0 to 1. Thus make the
drain optionally skip the racy check and drain on all cpus, and use this
option in zone_pcp_disable().
As we have to avoid external updates to high and batch while pcplists are
disabled, we take pcp_batch_high_lock in zone_pcp_disable() and release it
in zone_pcp_enable(). This also synchronizes multiple users of
zone_pcp_disable()/enable().
Currently the only user of this functionality is offline_pages().
[vbabka@suse.cz: add comment, per David]
Link: https://lkml.kernel.org/r/527480ef-ed72-e1c1-52a0-1c5b0113df45@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Suggested-by: David Hildenbrand <david@redhat.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:59 +00:00
|
|
|
__zone_set_pageset_high_and_batch(zone, new_high, new_batch);
|
2013-07-03 22:01:41 +00:00
|
|
|
}
|
|
|
|
|
2017-09-06 23:20:24 +00:00
|
|
|
void __meminit setup_zone_pageset(struct zone *zone)
|
2010-05-24 21:32:49 +00:00
|
|
|
{
|
2020-12-15 03:10:43 +00:00
|
|
|
struct per_cpu_pageset *p;
|
2010-05-24 21:32:49 +00:00
|
|
|
int cpu;
|
2020-12-15 03:10:43 +00:00
|
|
|
|
2010-05-24 21:32:49 +00:00
|
|
|
zone->pageset = alloc_percpu(struct per_cpu_pageset);
|
2020-12-15 03:10:43 +00:00
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
p = per_cpu_ptr(zone->pageset, cpu);
|
|
|
|
pageset_init(p);
|
|
|
|
}
|
|
|
|
|
|
|
|
zone_set_pageset_high_and_batch(zone);
|
2010-05-24 21:32:49 +00:00
|
|
|
}
|
|
|
|
|
2005-06-22 00:15:00 +00:00
|
|
|
/*
|
2010-01-05 06:34:51 +00:00
|
|
|
* Allocate per cpu pagesets and initialize them.
|
|
|
|
* Before this call only boot pagesets were available.
|
2005-06-22 00:14:47 +00:00
|
|
|
*/
|
2010-01-05 06:34:51 +00:00
|
|
|
void __init setup_per_cpu_pageset(void)
|
2005-06-22 00:14:47 +00:00
|
|
|
{
|
mm: initialise per_cpu_nodestats for all online pgdats at boot
Paul Mackerras and Reza Arbab reported that machines with memoryless
nodes fail when vmstats are refreshed. Paul reported an oops as follows
Unable to handle kernel paging request for data at address 0xff7a10000
Faulting instruction address: 0xc000000000270cd0
Oops: Kernel access of bad area, sig: 11 [#1]
SMP NR_CPUS=2048 NUMA PowerNV
Modules linked in:
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.7.0-kvm+ #118
task: c000000ff0680010 task.stack: c000000ff0704000
NIP: c000000000270cd0 LR: c000000000270ce8 CTR: 0000000000000000
REGS: c000000ff0707900 TRAP: 0300 Not tainted (4.7.0-kvm+)
MSR: 9000000102009033 <SF,HV,VEC,EE,ME,IR,DR,RI,LE,TM[E]> CR: 846b6824 XER: 20000000
CFAR: c000000000008768 DAR: 0000000ff7a10000 DSISR: 42000000 SOFTE: 1
NIP refresh_zone_stat_thresholds+0x80/0x240
LR refresh_zone_stat_thresholds+0x98/0x240
Call Trace:
refresh_zone_stat_thresholds+0xb8/0x240 (unreliable)
Both supplied potential fixes but one potentially misses checks and
another had redundant initialisations. This version initialises
per_cpu_nodestats on a per-pgdat basis instead of on a per-zone basis.
Link: http://lkml.kernel.org/r/20160804092404.GI2799@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Paul Mackerras <paulus@ozlabs.org>
Reported-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Tested-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-04 22:31:49 +00:00
|
|
|
struct pglist_data *pgdat;
|
2010-01-05 06:34:51 +00:00
|
|
|
struct zone *zone;
|
2020-06-03 22:59:11 +00:00
|
|
|
int __maybe_unused cpu;
|
2005-06-22 00:14:47 +00:00
|
|
|
|
2010-05-24 21:32:49 +00:00
|
|
|
for_each_populated_zone(zone)
|
|
|
|
setup_zone_pageset(zone);
|
mm: initialise per_cpu_nodestats for all online pgdats at boot
Paul Mackerras and Reza Arbab reported that machines with memoryless
nodes fail when vmstats are refreshed. Paul reported an oops as follows
Unable to handle kernel paging request for data at address 0xff7a10000
Faulting instruction address: 0xc000000000270cd0
Oops: Kernel access of bad area, sig: 11 [#1]
SMP NR_CPUS=2048 NUMA PowerNV
Modules linked in:
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.7.0-kvm+ #118
task: c000000ff0680010 task.stack: c000000ff0704000
NIP: c000000000270cd0 LR: c000000000270ce8 CTR: 0000000000000000
REGS: c000000ff0707900 TRAP: 0300 Not tainted (4.7.0-kvm+)
MSR: 9000000102009033 <SF,HV,VEC,EE,ME,IR,DR,RI,LE,TM[E]> CR: 846b6824 XER: 20000000
CFAR: c000000000008768 DAR: 0000000ff7a10000 DSISR: 42000000 SOFTE: 1
NIP refresh_zone_stat_thresholds+0x80/0x240
LR refresh_zone_stat_thresholds+0x98/0x240
Call Trace:
refresh_zone_stat_thresholds+0xb8/0x240 (unreliable)
Both supplied potential fixes but one potentially misses checks and
another had redundant initialisations. This version initialises
per_cpu_nodestats on a per-pgdat basis instead of on a per-zone basis.
Link: http://lkml.kernel.org/r/20160804092404.GI2799@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Paul Mackerras <paulus@ozlabs.org>
Reported-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Tested-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-04 22:31:49 +00:00
|
|
|
|
2020-06-03 22:59:11 +00:00
|
|
|
#ifdef CONFIG_NUMA
|
|
|
|
/*
|
|
|
|
* Unpopulated zones continue using the boot pagesets.
|
|
|
|
* The numa stats for these pagesets need to be reset.
|
|
|
|
* Otherwise, they will end up skewing the stats of
|
|
|
|
* the nodes these zones are associated with.
|
|
|
|
*/
|
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
struct per_cpu_pageset *pcp = &per_cpu(boot_pageset, cpu);
|
|
|
|
memset(pcp->vm_numa_stat_diff, 0,
|
|
|
|
sizeof(pcp->vm_numa_stat_diff));
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
mm: initialise per_cpu_nodestats for all online pgdats at boot
Paul Mackerras and Reza Arbab reported that machines with memoryless
nodes fail when vmstats are refreshed. Paul reported an oops as follows
Unable to handle kernel paging request for data at address 0xff7a10000
Faulting instruction address: 0xc000000000270cd0
Oops: Kernel access of bad area, sig: 11 [#1]
SMP NR_CPUS=2048 NUMA PowerNV
Modules linked in:
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.7.0-kvm+ #118
task: c000000ff0680010 task.stack: c000000ff0704000
NIP: c000000000270cd0 LR: c000000000270ce8 CTR: 0000000000000000
REGS: c000000ff0707900 TRAP: 0300 Not tainted (4.7.0-kvm+)
MSR: 9000000102009033 <SF,HV,VEC,EE,ME,IR,DR,RI,LE,TM[E]> CR: 846b6824 XER: 20000000
CFAR: c000000000008768 DAR: 0000000ff7a10000 DSISR: 42000000 SOFTE: 1
NIP refresh_zone_stat_thresholds+0x80/0x240
LR refresh_zone_stat_thresholds+0x98/0x240
Call Trace:
refresh_zone_stat_thresholds+0xb8/0x240 (unreliable)
Both supplied potential fixes but one potentially misses checks and
another had redundant initialisations. This version initialises
per_cpu_nodestats on a per-pgdat basis instead of on a per-zone basis.
Link: http://lkml.kernel.org/r/20160804092404.GI2799@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Paul Mackerras <paulus@ozlabs.org>
Reported-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Tested-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-04 22:31:49 +00:00
|
|
|
for_each_online_pgdat(pgdat)
|
|
|
|
pgdat->per_cpu_nodestats =
|
|
|
|
alloc_percpu(struct per_cpu_nodestat);
|
2005-06-22 00:14:47 +00:00
|
|
|
}
|
|
|
|
|
2006-01-17 06:03:44 +00:00
|
|
|
static __meminit void zone_pcp_init(struct zone *zone)
|
2005-10-30 01:16:50 +00:00
|
|
|
{
|
2010-01-05 06:34:51 +00:00
|
|
|
/*
|
|
|
|
* per cpu subsystem is not up at this point. The following code
|
|
|
|
* relies on the ability of the linker to provide the
|
|
|
|
* offset of a (static) per cpu variable into the per cpu area.
|
|
|
|
*/
|
|
|
|
zone->pageset = &boot_pageset;
|
2020-12-15 03:10:53 +00:00
|
|
|
zone->pageset_high = BOOT_PAGESET_HIGH;
|
|
|
|
zone->pageset_batch = BOOT_PAGESET_BATCH;
|
2005-10-30 01:16:50 +00:00
|
|
|
|
2013-11-12 23:07:20 +00:00
|
|
|
if (populated_zone(zone))
|
2010-01-05 06:34:51 +00:00
|
|
|
printk(KERN_DEBUG " %s zone: %lu pages, LIFO batch:%u\n",
|
|
|
|
zone->name, zone->present_pages,
|
|
|
|
zone_batchsize(zone));
|
2005-10-30 01:16:50 +00:00
|
|
|
}
|
|
|
|
|
mm: remove return value from init_currently_empty_zone
Patch series "mm: make movable onlining suck less", v4.
Movable onlining is a real hack with many downsides - mainly
reintroduction of lowmem/highmem issues we used to have on 32b systems -
but it is the only way to make the memory hotremove more reliable which
is something that people are asking for.
The current semantic of memory movable onlinening is really cumbersome,
however. The main reason for this is that the udev driven approach is
basically unusable because udev races with the memory probing while only
the last memory block or the one adjacent to the existing zone_movable
are allowed to be onlined movable. In short the criterion for the
successful online_movable changes under udev's feet. A reliable udev
approach would require a 2 phase approach where the first successful
movable online would have to check all the previous blocks and online
them in descending order. This is hard to be considered sane.
This patchset aims at making the onlining semantic more usable. First
of all it allows to online memory movable as long as it doesn't clash
with the existing ZONE_NORMAL. That means that ZONE_NORMAL and
ZONE_MOVABLE cannot overlap. Currently I preserve the original ordering
semantic so the zone always precedes the movable zone but I have plans
to remove this restriction in future because it is not really necessary.
First 3 patches are cleanups which should be ready to be merged right
away (unless I have missed something subtle of course).
Patch 4 deals with ZONE_DEVICE dependencies down the __add_pages path.
Patch 5 deals with implicit assumptions of register_one_node on pgdat
initialization.
Patches 6-10 deal with offline holes in the zone for pfn walkers. I
hope I got all of them right but people familiar with compaction should
double check this.
Patch 11 is the core of the change. In order to make it easier to
review I have tried it to be as minimalistic as possible and the large
code removal is moved to patch 14.
Patch 12 is a trivial follow up cleanup. Patch 13 fixes sparse warnings
and finally patch 14 removes the unused code.
I have tested the patches in kvm:
# qemu-system-x86_64 -enable-kvm -monitor pty -m 2G,slots=4,maxmem=4G -numa node,mem=1G -numa node,mem=1G ...
and then probed the additional memory by
(qemu) object_add memory-backend-ram,id=mem1,size=1G
(qemu) device_add pc-dimm,id=dimm1,memdev=mem1
Then I have used this simple script to probe the memory block by hand
# cat probe_memblock.sh
#!/bin/sh
BLOCK_NR=$1
# echo $((0x100000000+$BLOCK_NR*(128<<20))) > /sys/devices/system/memory/probe
# for i in $(seq 10); do sh probe_memblock.sh $i; done
# grep . /sys/devices/system/memory/memory3?/valid_zones 2>/dev/null
/sys/devices/system/memory/memory33/valid_zones:Normal Movable
/sys/devices/system/memory/memory34/valid_zones:Normal Movable
/sys/devices/system/memory/memory35/valid_zones:Normal Movable
/sys/devices/system/memory/memory36/valid_zones:Normal Movable
/sys/devices/system/memory/memory37/valid_zones:Normal Movable
/sys/devices/system/memory/memory38/valid_zones:Normal Movable
/sys/devices/system/memory/memory39/valid_zones:Normal Movable
The main difference to the original implementation is that all new
memblocks can be both online_kernel and online_movable initially because
there is no clash obviously. For the comparison the original
implementation would have
/sys/devices/system/memory/memory33/valid_zones:Normal
/sys/devices/system/memory/memory34/valid_zones:Normal
/sys/devices/system/memory/memory35/valid_zones:Normal
/sys/devices/system/memory/memory36/valid_zones:Normal
/sys/devices/system/memory/memory37/valid_zones:Normal
/sys/devices/system/memory/memory38/valid_zones:Normal
/sys/devices/system/memory/memory39/valid_zones:Normal Movable
Now
# echo online_movable > /sys/devices/system/memory/memory34/state
# grep . /sys/devices/system/memory/memory3?/valid_zones 2>/dev/null
/sys/devices/system/memory/memory33/valid_zones:Normal Movable
/sys/devices/system/memory/memory34/valid_zones:Movable
/sys/devices/system/memory/memory35/valid_zones:Movable
/sys/devices/system/memory/memory36/valid_zones:Movable
/sys/devices/system/memory/memory37/valid_zones:Movable
/sys/devices/system/memory/memory38/valid_zones:Movable
/sys/devices/system/memory/memory39/valid_zones:Movable
Block 33 can still be online both kernel and movable while all
the remaining can be only movable.
/proc/zonelist says
Node 0, zone Normal
pages free 0
min 0
low 0
high 0
spanned 0
present 0
--
Node 0, zone Movable
pages free 32753
min 85
low 117
high 149
spanned 32768
present 32768
A new memblock at a lower address will result in a new memblock (32)
which will still allow both Normal and Movable.
# sh probe_memblock.sh 0
# grep . /sys/devices/system/memory/memory3[2-5]/valid_zones 2>/dev/null
/sys/devices/system/memory/memory32/valid_zones:Normal Movable
/sys/devices/system/memory/memory33/valid_zones:Normal Movable
/sys/devices/system/memory/memory34/valid_zones:Movable
/sys/devices/system/memory/memory35/valid_zones:Movable
and online_kernel will convert it to the zone normal properly
while 33 can be still onlined both ways.
# echo online_kernel > /sys/devices/system/memory/memory32/state
# grep . /sys/devices/system/memory/memory3[2-5]/valid_zones 2>/dev/null
/sys/devices/system/memory/memory32/valid_zones:Normal
/sys/devices/system/memory/memory33/valid_zones:Normal Movable
/sys/devices/system/memory/memory34/valid_zones:Movable
/sys/devices/system/memory/memory35/valid_zones:Movable
/proc/zoneinfo will now tell
Node 0, zone Normal
pages free 65441
min 165
low 230
high 295
spanned 65536
present 65536
--
Node 0, zone Movable
pages free 32740
min 82
low 114
high 146
spanned 32768
present 32768
so both zones have one memblock spanned and present.
Onlining 39 should associate this block to the movable zone
# echo online > /sys/devices/system/memory/memory39/state
/proc/zoneinfo will now tell
Node 0, zone Normal
pages free 32765
min 80
low 112
high 144
spanned 32768
present 32768
--
Node 0, zone Movable
pages free 65501
min 160
low 225
high 290
spanned 196608
present 65536
so we will have a movable zone which spans 6 memblocks, 2 present and 4
representing a hole.
Offlining both movable blocks will lead to the zone with no present
pages which is the expected behavior I believe.
# echo offline > /sys/devices/system/memory/memory39/state
# echo offline > /sys/devices/system/memory/memory34/state
# grep -A6 "Movable\|Normal" /proc/zoneinfo
Node 0, zone Normal
pages free 32735
min 90
low 122
high 154
spanned 32768
present 32768
--
Node 0, zone Movable
pages free 0
min 0
low 0
high 0
spanned 196608
present 0
As a bonus we will get a nice cleanup in the memory hotplug codebase.
This patch (of 16):
init_currently_empty_zone doesn't have any error to return yet it is
still an int and callers try to be defensive and try to handle potential
error. Remove this nonsense and simplify all callers.
This patch shouldn't have any visible effect
Link: http://lkml.kernel.org/r/20170515085827.16474-2-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Acked-by: Balbir Singh <bsingharora@gmail.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Daniel Kiper <daniel.kiper@oracle.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Joonsoo Kim <js1304@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Reza Arbab <arbab@linux.vnet.ibm.com>
Cc: Tobias Regnery <tobias.regnery@gmail.com>
Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-06 22:37:35 +00:00
|
|
|
void __meminit init_currently_empty_zone(struct zone *zone,
|
2006-06-23 09:03:10 +00:00
|
|
|
unsigned long zone_start_pfn,
|
2015-11-06 02:47:06 +00:00
|
|
|
unsigned long size)
|
2005-10-30 01:16:50 +00:00
|
|
|
{
|
|
|
|
struct pglist_data *pgdat = zone->zone_pgdat;
|
mm/page_alloc.c: fix calculation of pgdat->nr_zones
init_currently_empty_zone() will adjust pgdat->nr_zones and set it to
'zone_idx(zone) + 1' unconditionally. This is correct in the normal
case, while not exact in hot-plug situation.
This function is used in two places:
* free_area_init_core()
* move_pfn_range_to_zone()
In the first case, we are sure zone index increase monotonically. While
in the second one, this is under users control.
One way to reproduce this is:
----------------------------
1. create a virtual machine with empty node1
-m 4G,slots=32,maxmem=32G \
-smp 4,maxcpus=8 \
-numa node,nodeid=0,mem=4G,cpus=0-3 \
-numa node,nodeid=1,mem=0G,cpus=4-7
2. hot-add cpu 3-7
cpu-add [3-7]
2. hot-add memory to nod1
object_add memory-backend-ram,id=ram0,size=1G
device_add pc-dimm,id=dimm0,memdev=ram0,node=1
3. online memory with following order
echo online_movable > memory47/state
echo online > memory40/state
After this, node1 will have its nr_zones equals to (ZONE_NORMAL + 1)
instead of (ZONE_MOVABLE + 1).
Michal said:
"Having an incorrect nr_zones might result in all sorts of problems
which would be quite hard to debug (e.g. reclaim not considering the
movable zone). I do not expect many users would suffer from this it
but still this is trivial and obviously right thing to do so
backporting to the stable tree shouldn't be harmful (last famous
words)"
Link: http://lkml.kernel.org/r/20181117022022.9956-1-richard.weiyang@gmail.com
Fixes: f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online")
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-11-30 22:09:07 +00:00
|
|
|
int zone_idx = zone_idx(zone) + 1;
|
mm: remove per-zone hashtable of bitlock waitqueues
The per-zone waitqueues exist because of a scalability issue with the
page waitqueues on some NUMA machines, but it turns out that they hurt
normal loads, and now with the vmalloced stacks they also end up
breaking gfs2 that uses a bit_wait on a stack object:
wait_on_bit(&gh->gh_iflags, HIF_WAIT, TASK_UNINTERRUPTIBLE)
where 'gh' can be a reference to the local variable 'mount_gh' on the
stack of fill_super().
The reason the per-zone hash table breaks for this case is that there is
no "zone" for virtual allocations, and trying to look up the physical
page to get at it will fail (with a BUG_ON()).
It turns out that I actually complained to the mm people about the
per-zone hash table for another reason just a month ago: the zone lookup
also hurts the regular use of "unlock_page()" a lot, because the zone
lookup ends up forcing several unnecessary cache misses and generates
horrible code.
As part of that earlier discussion, we had a much better solution for
the NUMA scalability issue - by just making the page lock have a
separate contention bit, the waitqueue doesn't even have to be looked at
for the normal case.
Peter Zijlstra already has a patch for that, but let's see if anybody
even notices. In the meantime, let's fix the actual gfs2 breakage by
simplifying the bitlock waitqueues and removing the per-zone issue.
Reported-by: Andreas Gruenbacher <agruenba@redhat.com>
Tested-by: Bob Peterson <rpeterso@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-26 17:15:30 +00:00
|
|
|
|
mm/page_alloc.c: fix calculation of pgdat->nr_zones
init_currently_empty_zone() will adjust pgdat->nr_zones and set it to
'zone_idx(zone) + 1' unconditionally. This is correct in the normal
case, while not exact in hot-plug situation.
This function is used in two places:
* free_area_init_core()
* move_pfn_range_to_zone()
In the first case, we are sure zone index increase monotonically. While
in the second one, this is under users control.
One way to reproduce this is:
----------------------------
1. create a virtual machine with empty node1
-m 4G,slots=32,maxmem=32G \
-smp 4,maxcpus=8 \
-numa node,nodeid=0,mem=4G,cpus=0-3 \
-numa node,nodeid=1,mem=0G,cpus=4-7
2. hot-add cpu 3-7
cpu-add [3-7]
2. hot-add memory to nod1
object_add memory-backend-ram,id=ram0,size=1G
device_add pc-dimm,id=dimm0,memdev=ram0,node=1
3. online memory with following order
echo online_movable > memory47/state
echo online > memory40/state
After this, node1 will have its nr_zones equals to (ZONE_NORMAL + 1)
instead of (ZONE_MOVABLE + 1).
Michal said:
"Having an incorrect nr_zones might result in all sorts of problems
which would be quite hard to debug (e.g. reclaim not considering the
movable zone). I do not expect many users would suffer from this it
but still this is trivial and obviously right thing to do so
backporting to the stable tree shouldn't be harmful (last famous
words)"
Link: http://lkml.kernel.org/r/20181117022022.9956-1-richard.weiyang@gmail.com
Fixes: f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online")
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-11-30 22:09:07 +00:00
|
|
|
if (zone_idx > pgdat->nr_zones)
|
|
|
|
pgdat->nr_zones = zone_idx;
|
2005-10-30 01:16:50 +00:00
|
|
|
|
|
|
|
zone->zone_start_pfn = zone_start_pfn;
|
|
|
|
|
2008-07-24 04:26:51 +00:00
|
|
|
mminit_dprintk(MMINIT_TRACE, "memmap_init",
|
|
|
|
"Initialising map node %d zone %lu pfns %lu -> %lu\n",
|
|
|
|
pgdat->node_id,
|
|
|
|
(unsigned long)zone_idx(zone),
|
|
|
|
zone_start_pfn, (zone_start_pfn + size));
|
|
|
|
|
2008-02-05 06:29:26 +00:00
|
|
|
zone_init_free_lists(zone);
|
mm: remove per-zone hashtable of bitlock waitqueues
The per-zone waitqueues exist because of a scalability issue with the
page waitqueues on some NUMA machines, but it turns out that they hurt
normal loads, and now with the vmalloced stacks they also end up
breaking gfs2 that uses a bit_wait on a stack object:
wait_on_bit(&gh->gh_iflags, HIF_WAIT, TASK_UNINTERRUPTIBLE)
where 'gh' can be a reference to the local variable 'mount_gh' on the
stack of fill_super().
The reason the per-zone hash table breaks for this case is that there is
no "zone" for virtual allocations, and trying to look up the physical
page to get at it will fail (with a BUG_ON()).
It turns out that I actually complained to the mm people about the
per-zone hash table for another reason just a month ago: the zone lookup
also hurts the regular use of "unlock_page()" a lot, because the zone
lookup ends up forcing several unnecessary cache misses and generates
horrible code.
As part of that earlier discussion, we had a much better solution for
the NUMA scalability issue - by just making the page lock have a
separate contention bit, the waitqueue doesn't even have to be looked at
for the normal case.
Peter Zijlstra already has a patch for that, but let's see if anybody
even notices. In the meantime, let's fix the actual gfs2 breakage by
simplifying the bitlock waitqueues and removing the per-zone issue.
Reported-by: Andreas Gruenbacher <agruenba@redhat.com>
Tested-by: Bob Peterson <rpeterso@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-26 17:15:30 +00:00
|
|
|
zone->initialized = 1;
|
2005-10-30 01:16:50 +00:00
|
|
|
}
|
|
|
|
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
/**
|
|
|
|
* get_pfn_range_for_nid - Return the start and end page frames for a node
|
2006-10-04 09:15:25 +00:00
|
|
|
* @nid: The nid to return the range for. If MAX_NUMNODES, the min and max PFN are returned.
|
|
|
|
* @start_pfn: Passed by reference. On return, it will have the node start_pfn.
|
|
|
|
* @end_pfn: Passed by reference. On return, it will have the node end_pfn.
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
*
|
|
|
|
* It returns the start and end page frame of a node based on information
|
2014-06-04 23:10:53 +00:00
|
|
|
* provided by memblock_set_node(). If called for a node
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
* with no available memory, a warning is printed and the start and end
|
2006-10-04 09:15:25 +00:00
|
|
|
* PFNs will be 0.
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
*/
|
2018-12-28 08:37:24 +00:00
|
|
|
void __init get_pfn_range_for_nid(unsigned int nid,
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
unsigned long *start_pfn, unsigned long *end_pfn)
|
|
|
|
{
|
2011-07-12 08:46:30 +00:00
|
|
|
unsigned long this_start_pfn, this_end_pfn;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
int i;
|
2011-07-12 08:46:30 +00:00
|
|
|
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
*start_pfn = -1UL;
|
|
|
|
*end_pfn = 0;
|
|
|
|
|
2011-07-12 08:46:30 +00:00
|
|
|
for_each_mem_pfn_range(i, nid, &this_start_pfn, &this_end_pfn, NULL) {
|
|
|
|
*start_pfn = min(*start_pfn, this_start_pfn);
|
|
|
|
*end_pfn = max(*end_pfn, this_end_pfn);
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
|
|
|
|
2007-10-16 08:25:37 +00:00
|
|
|
if (*start_pfn == -1UL)
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
*start_pfn = 0;
|
|
|
|
}
|
|
|
|
|
2007-07-17 11:03:12 +00:00
|
|
|
/*
|
|
|
|
* This finds a zone that can be used for ZONE_MOVABLE pages. The
|
|
|
|
* assumption is made that zones within a node are ordered in monotonic
|
|
|
|
* increasing memory addresses so that the "highest" populated zone is used
|
|
|
|
*/
|
2008-07-24 04:28:12 +00:00
|
|
|
static void __init find_usable_zone_for_movable(void)
|
2007-07-17 11:03:12 +00:00
|
|
|
{
|
|
|
|
int zone_index;
|
|
|
|
for (zone_index = MAX_NR_ZONES - 1; zone_index >= 0; zone_index--) {
|
|
|
|
if (zone_index == ZONE_MOVABLE)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (arch_zone_highest_possible_pfn[zone_index] >
|
|
|
|
arch_zone_lowest_possible_pfn[zone_index])
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
VM_BUG_ON(zone_index == -1);
|
|
|
|
movable_zone = zone_index;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The zone ranges provided by the architecture do not include ZONE_MOVABLE
|
2011-03-31 01:57:33 +00:00
|
|
|
* because it is sized independent of architecture. Unlike the other zones,
|
2007-07-17 11:03:12 +00:00
|
|
|
* the starting point for ZONE_MOVABLE is not fixed. It may be different
|
|
|
|
* in each node depending on the size of each node and how evenly kernelcore
|
|
|
|
* is distributed. This helper function adjusts the zone ranges
|
|
|
|
* provided by the architecture for a given node by using the end of the
|
|
|
|
* highest usable zone for ZONE_MOVABLE. This preserves the assumption that
|
|
|
|
* zones within a node are in order of monotonic increases memory addresses
|
|
|
|
*/
|
2018-12-28 08:37:24 +00:00
|
|
|
static void __init adjust_zone_range_for_zone_movable(int nid,
|
2007-07-17 11:03:12 +00:00
|
|
|
unsigned long zone_type,
|
|
|
|
unsigned long node_start_pfn,
|
|
|
|
unsigned long node_end_pfn,
|
|
|
|
unsigned long *zone_start_pfn,
|
|
|
|
unsigned long *zone_end_pfn)
|
|
|
|
{
|
|
|
|
/* Only adjust if ZONE_MOVABLE is on this node */
|
|
|
|
if (zone_movable_pfn[nid]) {
|
|
|
|
/* Size ZONE_MOVABLE */
|
|
|
|
if (zone_type == ZONE_MOVABLE) {
|
|
|
|
*zone_start_pfn = zone_movable_pfn[nid];
|
|
|
|
*zone_end_pfn = min(node_end_pfn,
|
|
|
|
arch_zone_highest_possible_pfn[movable_zone]);
|
|
|
|
|
mem-hotplug: fix node spanned pages when we have a movable node
Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror
option") rewrote the calculation of node spanned pages. But when we
have a movable node, the size of node spanned pages is double added.
That's because we have an empty normal zone, the present pages is zero,
but its spanned pages is not zero.
e.g.
Zone ranges:
DMA [mem 0x0000000000001000-0x0000000000ffffff]
DMA32 [mem 0x0000000001000000-0x00000000ffffffff]
Normal [mem 0x0000000100000000-0x0000007c7fffffff]
Movable zone start for each node
Node 1: 0x0000001080000000
Node 2: 0x0000002080000000
Node 3: 0x0000003080000000
Node 4: 0x0000003c80000000
Node 5: 0x0000004c80000000
Node 6: 0x0000005c80000000
Early memory node ranges
node 0: [mem 0x0000000000001000-0x000000000009ffff]
node 0: [mem 0x0000000000100000-0x000000007552afff]
node 0: [mem 0x000000007bd46000-0x000000007bd46fff]
node 0: [mem 0x000000007bdcd000-0x000000007bffffff]
node 0: [mem 0x0000000100000000-0x000000107fffffff]
node 1: [mem 0x0000001080000000-0x000000207fffffff]
node 2: [mem 0x0000002080000000-0x000000307fffffff]
node 3: [mem 0x0000003080000000-0x0000003c7fffffff]
node 4: [mem 0x0000003c80000000-0x0000004c7fffffff]
node 5: [mem 0x0000004c80000000-0x0000005c7fffffff]
node 6: [mem 0x0000005c80000000-0x0000006c7fffffff]
node 7: [mem 0x0000006c80000000-0x0000007c7fffffff]
node1:
Normal, start=0x1080000, present=0x0, spanned=0x1000000
Movable, start=0x1080000, present=0x1000000, spanned=0x1000000
pgdat, start=0x1080000, present=0x1000000, spanned=0x2000000
After this patch, the problem is fixed.
node1:
Normal, start=0x0, present=0x0, spanned=0x0
Movable, start=0x1080000, present=0x1000000, spanned=0x1000000
pgdat, start=0x1080000, present=0x1000000, spanned=0x1000000
Link: http://lkml.kernel.org/r/57A325E8.6070100@huawei.com
Signed-off-by: Xishi Qiu <qiuxishi@huawei.com>
Cc: Taku Izumi <izumi.taku@jp.fujitsu.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-07 23:58:06 +00:00
|
|
|
/* Adjust for ZONE_MOVABLE starting within this range */
|
|
|
|
} else if (!mirrored_kernelcore &&
|
|
|
|
*zone_start_pfn < zone_movable_pfn[nid] &&
|
|
|
|
*zone_end_pfn > zone_movable_pfn[nid]) {
|
|
|
|
*zone_end_pfn = zone_movable_pfn[nid];
|
|
|
|
|
2007-07-17 11:03:12 +00:00
|
|
|
/* Check if this whole range is within ZONE_MOVABLE */
|
|
|
|
} else if (*zone_start_pfn >= zone_movable_pfn[nid])
|
|
|
|
*zone_start_pfn = *zone_end_pfn;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
/*
|
|
|
|
* Return the number of pages a zone spans in a node, including holes
|
|
|
|
* present_pages = zone_spanned_pages_in_node() - zone_absent_pages_in_node()
|
|
|
|
*/
|
2018-12-28 08:37:24 +00:00
|
|
|
static unsigned long __init zone_spanned_pages_in_node(int nid,
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
unsigned long zone_type,
|
2013-07-08 22:59:52 +00:00
|
|
|
unsigned long node_start_pfn,
|
|
|
|
unsigned long node_end_pfn,
|
2016-03-15 21:55:18 +00:00
|
|
|
unsigned long *zone_start_pfn,
|
2020-06-03 22:58:13 +00:00
|
|
|
unsigned long *zone_end_pfn)
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
{
|
mem-hotplug: fix node spanned pages when we have a node with only ZONE_MOVABLE
342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror option") and
later patches rewrote the calculation of node spanned pages.
e506b99696a2 ("mem-hotplug: fix node spanned pages when we have a movable
node"), but the current code still has problems,
When we have a node with only zone_movable and the node id is not zero,
the size of node spanned pages is double added.
That's because we have an empty normal zone, and zone_start_pfn or
zone_end_pfn is not between arch_zone_lowest_possible_pfn and
arch_zone_highest_possible_pfn, so we need to use clamp to constrain the
range just like the commit <96e907d13602> (bootmem: Reimplement
__absent_pages_in_range() using for_each_mem_pfn_range()).
e.g.
Zone ranges:
DMA [mem 0x0000000000001000-0x0000000000ffffff]
DMA32 [mem 0x0000000001000000-0x00000000ffffffff]
Normal [mem 0x0000000100000000-0x000000023fffffff]
Movable zone start for each node
Node 0: 0x0000000100000000
Node 1: 0x0000000140000000
Early memory node ranges
node 0: [mem 0x0000000000001000-0x000000000009efff]
node 0: [mem 0x0000000000100000-0x00000000bffdffff]
node 0: [mem 0x0000000100000000-0x000000013fffffff]
node 1: [mem 0x0000000140000000-0x000000023fffffff]
node 0 DMA spanned:0xfff present:0xf9e absent:0x61
node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020
node 0 Normal spanned:0 present:0 absent:0
node 0 Movable spanned:0x40000 present:0x40000 absent:0
On node 0 totalpages(node_present_pages): 1048446
node_spanned_pages:1310719
node 1 DMA spanned:0 present:0 absent:0
node 1 DMA32 spanned:0 present:0 absent:0
node 1 Normal spanned:0x100000 present:0x100000 absent:0
node 1 Movable spanned:0x100000 present:0x100000 absent:0
On node 1 totalpages(node_present_pages): 2097152
node_spanned_pages:2097152
Memory: 6967796K/12582392K available (16388K kernel code, 3686K rwdata,
4468K rodata, 2160K init, 10444K bss, 5614596K reserved, 0K
cma-reserved)
It shows that the current memory of node 1 is double added.
After this patch, the problem is fixed.
node 0 DMA spanned:0xfff present:0xf9e absent:0x61
node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020
node 0 Normal spanned:0 present:0 absent:0
node 0 Movable spanned:0x40000 present:0x40000 absent:0
On node 0 totalpages(node_present_pages): 1048446
node_spanned_pages:1310719
node 1 DMA spanned:0 present:0 absent:0
node 1 DMA32 spanned:0 present:0 absent:0
node 1 Normal spanned:0 present:0 absent:0
node 1 Movable spanned:0x100000 present:0x100000 absent:0
On node 1 totalpages(node_present_pages): 1048576
node_spanned_pages:1048576
memory: 6967796K/8388088K available (16388K kernel code, 3686K rwdata,
4468K rodata, 2160K init, 10444K bss, 1420292K reserved, 0K
cma-reserved)
Link: http://lkml.kernel.org/r/1554178276-10372-1-git-send-email-fanglinxu@huawei.com
Signed-off-by: Linxu Fang <fanglinxu@huawei.com>
Cc: Taku Izumi <izumi.taku@jp.fujitsu.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 00:19:17 +00:00
|
|
|
unsigned long zone_low = arch_zone_lowest_possible_pfn[zone_type];
|
|
|
|
unsigned long zone_high = arch_zone_highest_possible_pfn[zone_type];
|
2015-09-08 22:04:16 +00:00
|
|
|
/* When hotadd a new node from cpu_up(), the node should be empty */
|
2015-08-14 22:35:16 +00:00
|
|
|
if (!node_start_pfn && !node_end_pfn)
|
|
|
|
return 0;
|
|
|
|
|
2013-07-08 22:59:52 +00:00
|
|
|
/* Get the start and end of the zone */
|
mem-hotplug: fix node spanned pages when we have a node with only ZONE_MOVABLE
342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror option") and
later patches rewrote the calculation of node spanned pages.
e506b99696a2 ("mem-hotplug: fix node spanned pages when we have a movable
node"), but the current code still has problems,
When we have a node with only zone_movable and the node id is not zero,
the size of node spanned pages is double added.
That's because we have an empty normal zone, and zone_start_pfn or
zone_end_pfn is not between arch_zone_lowest_possible_pfn and
arch_zone_highest_possible_pfn, so we need to use clamp to constrain the
range just like the commit <96e907d13602> (bootmem: Reimplement
__absent_pages_in_range() using for_each_mem_pfn_range()).
e.g.
Zone ranges:
DMA [mem 0x0000000000001000-0x0000000000ffffff]
DMA32 [mem 0x0000000001000000-0x00000000ffffffff]
Normal [mem 0x0000000100000000-0x000000023fffffff]
Movable zone start for each node
Node 0: 0x0000000100000000
Node 1: 0x0000000140000000
Early memory node ranges
node 0: [mem 0x0000000000001000-0x000000000009efff]
node 0: [mem 0x0000000000100000-0x00000000bffdffff]
node 0: [mem 0x0000000100000000-0x000000013fffffff]
node 1: [mem 0x0000000140000000-0x000000023fffffff]
node 0 DMA spanned:0xfff present:0xf9e absent:0x61
node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020
node 0 Normal spanned:0 present:0 absent:0
node 0 Movable spanned:0x40000 present:0x40000 absent:0
On node 0 totalpages(node_present_pages): 1048446
node_spanned_pages:1310719
node 1 DMA spanned:0 present:0 absent:0
node 1 DMA32 spanned:0 present:0 absent:0
node 1 Normal spanned:0x100000 present:0x100000 absent:0
node 1 Movable spanned:0x100000 present:0x100000 absent:0
On node 1 totalpages(node_present_pages): 2097152
node_spanned_pages:2097152
Memory: 6967796K/12582392K available (16388K kernel code, 3686K rwdata,
4468K rodata, 2160K init, 10444K bss, 5614596K reserved, 0K
cma-reserved)
It shows that the current memory of node 1 is double added.
After this patch, the problem is fixed.
node 0 DMA spanned:0xfff present:0xf9e absent:0x61
node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020
node 0 Normal spanned:0 present:0 absent:0
node 0 Movable spanned:0x40000 present:0x40000 absent:0
On node 0 totalpages(node_present_pages): 1048446
node_spanned_pages:1310719
node 1 DMA spanned:0 present:0 absent:0
node 1 DMA32 spanned:0 present:0 absent:0
node 1 Normal spanned:0 present:0 absent:0
node 1 Movable spanned:0x100000 present:0x100000 absent:0
On node 1 totalpages(node_present_pages): 1048576
node_spanned_pages:1048576
memory: 6967796K/8388088K available (16388K kernel code, 3686K rwdata,
4468K rodata, 2160K init, 10444K bss, 1420292K reserved, 0K
cma-reserved)
Link: http://lkml.kernel.org/r/1554178276-10372-1-git-send-email-fanglinxu@huawei.com
Signed-off-by: Linxu Fang <fanglinxu@huawei.com>
Cc: Taku Izumi <izumi.taku@jp.fujitsu.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 00:19:17 +00:00
|
|
|
*zone_start_pfn = clamp(node_start_pfn, zone_low, zone_high);
|
|
|
|
*zone_end_pfn = clamp(node_end_pfn, zone_low, zone_high);
|
2007-07-17 11:03:12 +00:00
|
|
|
adjust_zone_range_for_zone_movable(nid, zone_type,
|
|
|
|
node_start_pfn, node_end_pfn,
|
2016-03-15 21:55:18 +00:00
|
|
|
zone_start_pfn, zone_end_pfn);
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
|
|
|
|
/* Check that this node has pages within the zone's required range */
|
2016-03-15 21:55:18 +00:00
|
|
|
if (*zone_end_pfn < node_start_pfn || *zone_start_pfn > node_end_pfn)
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Move the zone boundaries inside the node if necessary */
|
2016-03-15 21:55:18 +00:00
|
|
|
*zone_end_pfn = min(*zone_end_pfn, node_end_pfn);
|
|
|
|
*zone_start_pfn = max(*zone_start_pfn, node_start_pfn);
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
|
|
|
|
/* Return the spanned pages */
|
2016-03-15 21:55:18 +00:00
|
|
|
return *zone_end_pfn - *zone_start_pfn;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return the number of holes in a range on a node. If nid is MAX_NUMNODES,
|
2006-10-04 09:15:25 +00:00
|
|
|
* then all holes in the requested range will be accounted for.
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
*/
|
2018-12-28 08:37:24 +00:00
|
|
|
unsigned long __init __absent_pages_in_range(int nid,
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
unsigned long range_start_pfn,
|
|
|
|
unsigned long range_end_pfn)
|
|
|
|
{
|
2011-07-12 08:46:29 +00:00
|
|
|
unsigned long nr_absent = range_end_pfn - range_start_pfn;
|
|
|
|
unsigned long start_pfn, end_pfn;
|
|
|
|
int i;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
|
2011-07-12 08:46:29 +00:00
|
|
|
for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
|
|
|
|
start_pfn = clamp(start_pfn, range_start_pfn, range_end_pfn);
|
|
|
|
end_pfn = clamp(end_pfn, range_start_pfn, range_end_pfn);
|
|
|
|
nr_absent -= end_pfn - start_pfn;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
2011-07-12 08:46:29 +00:00
|
|
|
return nr_absent;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* absent_pages_in_range - Return number of page frames in holes within a range
|
|
|
|
* @start_pfn: The start PFN to start searching for holes
|
|
|
|
* @end_pfn: The end PFN to stop searching for holes
|
|
|
|
*
|
2019-03-05 23:48:42 +00:00
|
|
|
* Return: the number of pages frames in memory holes within a range.
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
*/
|
|
|
|
unsigned long __init absent_pages_in_range(unsigned long start_pfn,
|
|
|
|
unsigned long end_pfn)
|
|
|
|
{
|
|
|
|
return __absent_pages_in_range(MAX_NUMNODES, start_pfn, end_pfn);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Return the number of page frames in holes in a zone on a node */
|
2018-12-28 08:37:24 +00:00
|
|
|
static unsigned long __init zone_absent_pages_in_node(int nid,
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
unsigned long zone_type,
|
2013-07-08 22:59:52 +00:00
|
|
|
unsigned long node_start_pfn,
|
2020-06-03 22:58:13 +00:00
|
|
|
unsigned long node_end_pfn)
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
{
|
2011-07-12 08:46:29 +00:00
|
|
|
unsigned long zone_low = arch_zone_lowest_possible_pfn[zone_type];
|
|
|
|
unsigned long zone_high = arch_zone_highest_possible_pfn[zone_type];
|
2006-09-27 08:49:58 +00:00
|
|
|
unsigned long zone_start_pfn, zone_end_pfn;
|
2016-03-15 21:55:22 +00:00
|
|
|
unsigned long nr_absent;
|
2006-09-27 08:49:58 +00:00
|
|
|
|
2015-09-08 22:04:16 +00:00
|
|
|
/* When hotadd a new node from cpu_up(), the node should be empty */
|
2015-08-14 22:35:16 +00:00
|
|
|
if (!node_start_pfn && !node_end_pfn)
|
|
|
|
return 0;
|
|
|
|
|
2011-07-12 08:46:29 +00:00
|
|
|
zone_start_pfn = clamp(node_start_pfn, zone_low, zone_high);
|
|
|
|
zone_end_pfn = clamp(node_end_pfn, zone_low, zone_high);
|
2006-09-27 08:49:58 +00:00
|
|
|
|
2007-07-17 11:03:12 +00:00
|
|
|
adjust_zone_range_for_zone_movable(nid, zone_type,
|
|
|
|
node_start_pfn, node_end_pfn,
|
|
|
|
&zone_start_pfn, &zone_end_pfn);
|
2016-03-15 21:55:22 +00:00
|
|
|
nr_absent = __absent_pages_in_range(nid, zone_start_pfn, zone_end_pfn);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* ZONE_MOVABLE handling.
|
|
|
|
* Treat pages to be ZONE_MOVABLE in ZONE_NORMAL as absent pages
|
|
|
|
* and vice versa.
|
|
|
|
*/
|
mem-hotplug: fix node spanned pages when we have a movable node
Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror
option") rewrote the calculation of node spanned pages. But when we
have a movable node, the size of node spanned pages is double added.
That's because we have an empty normal zone, the present pages is zero,
but its spanned pages is not zero.
e.g.
Zone ranges:
DMA [mem 0x0000000000001000-0x0000000000ffffff]
DMA32 [mem 0x0000000001000000-0x00000000ffffffff]
Normal [mem 0x0000000100000000-0x0000007c7fffffff]
Movable zone start for each node
Node 1: 0x0000001080000000
Node 2: 0x0000002080000000
Node 3: 0x0000003080000000
Node 4: 0x0000003c80000000
Node 5: 0x0000004c80000000
Node 6: 0x0000005c80000000
Early memory node ranges
node 0: [mem 0x0000000000001000-0x000000000009ffff]
node 0: [mem 0x0000000000100000-0x000000007552afff]
node 0: [mem 0x000000007bd46000-0x000000007bd46fff]
node 0: [mem 0x000000007bdcd000-0x000000007bffffff]
node 0: [mem 0x0000000100000000-0x000000107fffffff]
node 1: [mem 0x0000001080000000-0x000000207fffffff]
node 2: [mem 0x0000002080000000-0x000000307fffffff]
node 3: [mem 0x0000003080000000-0x0000003c7fffffff]
node 4: [mem 0x0000003c80000000-0x0000004c7fffffff]
node 5: [mem 0x0000004c80000000-0x0000005c7fffffff]
node 6: [mem 0x0000005c80000000-0x0000006c7fffffff]
node 7: [mem 0x0000006c80000000-0x0000007c7fffffff]
node1:
Normal, start=0x1080000, present=0x0, spanned=0x1000000
Movable, start=0x1080000, present=0x1000000, spanned=0x1000000
pgdat, start=0x1080000, present=0x1000000, spanned=0x2000000
After this patch, the problem is fixed.
node1:
Normal, start=0x0, present=0x0, spanned=0x0
Movable, start=0x1080000, present=0x1000000, spanned=0x1000000
pgdat, start=0x1080000, present=0x1000000, spanned=0x1000000
Link: http://lkml.kernel.org/r/57A325E8.6070100@huawei.com
Signed-off-by: Xishi Qiu <qiuxishi@huawei.com>
Cc: Taku Izumi <izumi.taku@jp.fujitsu.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-07 23:58:06 +00:00
|
|
|
if (mirrored_kernelcore && zone_movable_pfn[nid]) {
|
|
|
|
unsigned long start_pfn, end_pfn;
|
|
|
|
struct memblock_region *r;
|
|
|
|
|
2020-10-13 23:58:30 +00:00
|
|
|
for_each_mem_region(r) {
|
mem-hotplug: fix node spanned pages when we have a movable node
Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror
option") rewrote the calculation of node spanned pages. But when we
have a movable node, the size of node spanned pages is double added.
That's because we have an empty normal zone, the present pages is zero,
but its spanned pages is not zero.
e.g.
Zone ranges:
DMA [mem 0x0000000000001000-0x0000000000ffffff]
DMA32 [mem 0x0000000001000000-0x00000000ffffffff]
Normal [mem 0x0000000100000000-0x0000007c7fffffff]
Movable zone start for each node
Node 1: 0x0000001080000000
Node 2: 0x0000002080000000
Node 3: 0x0000003080000000
Node 4: 0x0000003c80000000
Node 5: 0x0000004c80000000
Node 6: 0x0000005c80000000
Early memory node ranges
node 0: [mem 0x0000000000001000-0x000000000009ffff]
node 0: [mem 0x0000000000100000-0x000000007552afff]
node 0: [mem 0x000000007bd46000-0x000000007bd46fff]
node 0: [mem 0x000000007bdcd000-0x000000007bffffff]
node 0: [mem 0x0000000100000000-0x000000107fffffff]
node 1: [mem 0x0000001080000000-0x000000207fffffff]
node 2: [mem 0x0000002080000000-0x000000307fffffff]
node 3: [mem 0x0000003080000000-0x0000003c7fffffff]
node 4: [mem 0x0000003c80000000-0x0000004c7fffffff]
node 5: [mem 0x0000004c80000000-0x0000005c7fffffff]
node 6: [mem 0x0000005c80000000-0x0000006c7fffffff]
node 7: [mem 0x0000006c80000000-0x0000007c7fffffff]
node1:
Normal, start=0x1080000, present=0x0, spanned=0x1000000
Movable, start=0x1080000, present=0x1000000, spanned=0x1000000
pgdat, start=0x1080000, present=0x1000000, spanned=0x2000000
After this patch, the problem is fixed.
node1:
Normal, start=0x0, present=0x0, spanned=0x0
Movable, start=0x1080000, present=0x1000000, spanned=0x1000000
pgdat, start=0x1080000, present=0x1000000, spanned=0x1000000
Link: http://lkml.kernel.org/r/57A325E8.6070100@huawei.com
Signed-off-by: Xishi Qiu <qiuxishi@huawei.com>
Cc: Taku Izumi <izumi.taku@jp.fujitsu.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-10-07 23:58:06 +00:00
|
|
|
start_pfn = clamp(memblock_region_memory_base_pfn(r),
|
|
|
|
zone_start_pfn, zone_end_pfn);
|
|
|
|
end_pfn = clamp(memblock_region_memory_end_pfn(r),
|
|
|
|
zone_start_pfn, zone_end_pfn);
|
|
|
|
|
|
|
|
if (zone_type == ZONE_MOVABLE &&
|
|
|
|
memblock_is_mirror(r))
|
|
|
|
nr_absent += end_pfn - start_pfn;
|
|
|
|
|
|
|
|
if (zone_type == ZONE_NORMAL &&
|
|
|
|
!memblock_is_mirror(r))
|
|
|
|
nr_absent += end_pfn - start_pfn;
|
2016-03-15 21:55:22 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return nr_absent;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
2006-09-27 08:49:56 +00:00
|
|
|
|
2018-12-28 08:37:24 +00:00
|
|
|
static void __init calculate_node_totalpages(struct pglist_data *pgdat,
|
2013-07-08 22:59:52 +00:00
|
|
|
unsigned long node_start_pfn,
|
2020-06-03 22:58:13 +00:00
|
|
|
unsigned long node_end_pfn)
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
{
|
2015-06-24 23:57:02 +00:00
|
|
|
unsigned long realtotalpages = 0, totalpages = 0;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
enum zone_type i;
|
|
|
|
|
2015-06-24 23:57:02 +00:00
|
|
|
for (i = 0; i < MAX_NR_ZONES; i++) {
|
|
|
|
struct zone *zone = pgdat->node_zones + i;
|
2016-03-15 21:55:18 +00:00
|
|
|
unsigned long zone_start_pfn, zone_end_pfn;
|
2020-06-03 22:57:02 +00:00
|
|
|
unsigned long spanned, absent;
|
2015-06-24 23:57:02 +00:00
|
|
|
unsigned long size, real_size;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
|
2020-06-03 22:58:13 +00:00
|
|
|
spanned = zone_spanned_pages_in_node(pgdat->node_id, i,
|
|
|
|
node_start_pfn,
|
|
|
|
node_end_pfn,
|
|
|
|
&zone_start_pfn,
|
|
|
|
&zone_end_pfn);
|
|
|
|
absent = zone_absent_pages_in_node(pgdat->node_id, i,
|
|
|
|
node_start_pfn,
|
|
|
|
node_end_pfn);
|
2020-06-03 22:57:02 +00:00
|
|
|
|
|
|
|
size = spanned;
|
|
|
|
real_size = size - absent;
|
|
|
|
|
2016-03-15 21:55:18 +00:00
|
|
|
if (size)
|
|
|
|
zone->zone_start_pfn = zone_start_pfn;
|
|
|
|
else
|
|
|
|
zone->zone_start_pfn = 0;
|
2015-06-24 23:57:02 +00:00
|
|
|
zone->spanned_pages = size;
|
|
|
|
zone->present_pages = real_size;
|
|
|
|
|
|
|
|
totalpages += size;
|
|
|
|
realtotalpages += real_size;
|
|
|
|
}
|
|
|
|
|
|
|
|
pgdat->node_spanned_pages = totalpages;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
pgdat->node_present_pages = realtotalpages;
|
|
|
|
printk(KERN_DEBUG "On node %d totalpages: %lu\n", pgdat->node_id,
|
|
|
|
realtotalpages);
|
|
|
|
}
|
|
|
|
|
Add a bitmap that is used to track flags affecting a block of pages
Here is the latest revision of the anti-fragmentation patches. Of particular
note in this version is special treatment of high-order atomic allocations.
Care is taken to group them together and avoid grouping pages of other types
near them. Artifical tests imply that it works. I'm trying to get the
hardware together that would allow setting up of a "real" test. If anyone
already has a setup and test that can trigger the atomic-allocation problem,
I'd appreciate a test of these patches and a report. The second major change
is that these patches will apply cleanly with patches that implement
anti-fragmentation through zones.
kernbench shows effectively no performance difference varying between -0.2%
and +2% on a variety of test machines. Success rates for huge page allocation
are dramatically increased. For example, on a ppc64 machine, the vanilla
kernel was only able to allocate 1% of memory as a hugepage and this was due
to a single hugepage reserved as min_free_kbytes. With these patches applied,
17% was allocatable as superpages. With reclaim-related fixes from Andy
Whitcroft, it was 40% and further reclaim-related improvements should increase
this further.
Changelog Since V28
o Group high-order atomic allocations together
o It is no longer required to set min_free_kbytes to 10% of memory. A value
of 16384 in most cases will be sufficient
o Now applied with zone-based anti-fragmentation
o Fix incorrect VM_BUG_ON within buffered_rmqueue()
o Reorder the stack so later patches do not back out work from earlier patches
o Fix bug were journal pages were being treated as movable
o Bias placement of non-movable pages to lower PFNs
o More agressive clustering of reclaimable pages in reactions to workloads
like updatedb that flood the size of inode caches
Changelog Since V27
o Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving
the mistaken impression that it was the 100% solution for high order
allocations. Instead, it greatly increases the chances high-order
allocations will succeed and lays the foundation for defragmentation and
memory hot-remove to work properly
o Redefine page groupings based on ability to migrate or reclaim instead of
basing on reclaimability alone
o Get rid of spurious inits
o Per-cpu lists are no longer split up per-type. Instead the per-cpu list is
searched for a page of the appropriate type
o Added more explanation commentary
o Fix up bug in pageblock code where bitmap was used before being initalised
Changelog Since V26
o Fix double init of lists in setup_pageset
Changelog Since V25
o Fix loop order of for_each_rclmtype_order so that order of loop matches args
o gfpflags_to_rclmtype uses gfp_t instead of unsigned long
o Rename get_pageblock_type() to get_page_rclmtype()
o Fix alignment problem in move_freepages()
o Add mechanism for assigning flags to blocks of pages instead of page->flags
o On fallback, do not examine the preferred list of free pages a second time
The purpose of these patches is to reduce external fragmentation by grouping
pages of related types together. When pages are migrated (or reclaimed under
memory pressure), large contiguous pages will be freed.
This patch works by categorising allocations by their ability to migrate;
Movable - The pages may be moved with the page migration mechanism. These are
generally userspace pages.
Reclaimable - These are allocations for some kernel caches that are
reclaimable or allocations that are known to be very short-lived.
Unmovable - These are pages that are allocated by the kernel that
are not trivially reclaimed. For example, the memory allocated for a
loaded module would be in this category. By default, allocations are
considered to be of this type
HighAtomic - These are high-order allocations belonging to callers that
cannot sleep or perform any IO. In practice, this is restricted to
jumbo frame allocation for network receive. It is assumed that the
allocations are short-lived
Instead of having one MAX_ORDER-sized array of free lists in struct free_area,
there is one for each type of reclaimability. Once a 2^MAX_ORDER block of
pages is split for a type of allocation, it is added to the free-lists for
that type, in effect reserving it. Hence, over time, pages of the different
types can be clustered together.
When the preferred freelists are expired, the largest possible block is taken
from an alternative list. Buddies that are split from that large block are
placed on the preferred allocation-type freelists to mitigate fragmentation.
This implementation gives best-effort for low fragmentation in all zones.
Ideally, min_free_kbytes needs to be set to a value equal to 4 * (1 <<
(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for
example.
Our tests show that about 60-70% of physical memory can be allocated on a
desktop after a few days uptime. In benchmarks and stress tests, we are
finding that 80% of memory is available as contiguous blocks at the end of the
test. To compare, a standard kernel was getting < 1% of memory as large pages
on a desktop and about 8-12% of memory as large pages at the end of stress
tests.
Following this email are 12 patches that implement thie page grouping feature.
The first patch introduces a mechanism for storing flags related to a whole
block of pages. Then allocations are split between movable and all other
allocations. Following that are patches to deal with per-cpu pages and make
the mechanism configurable. The next patch moves free pages between lists
when partially allocated blocks are used for pages of another migrate type.
The second last patch groups reclaimable kernel allocations such as inode
caches together. The final patch related to groupings keeps high-order atomic
allocations.
The last two patches are more concerned with control of fragmentation. The
second last patch biases placement of non-movable allocations towards the
start of memory. This is with a view of supporting memory hot-remove of DIMMs
with higher PFNs in the future. The biasing could be enforced a lot heavier
but it would cost. The last patch agressively clusters reclaimable pages like
inode caches together.
The fragmentation reduction strategy needs to track if pages within a block
can be moved or reclaimed so that pages are freed to the appropriate list.
This patch adds a bitmap for flags affecting a whole a MAX_ORDER block of
pages.
In non-SPARSEMEM configurations, the bitmap is stored in the struct zone and
allocated during initialisation. SPARSEMEM statically allocates the bitmap in
a struct mem_section so that bitmaps do not have to be resized during memory
hotadd. This wastes a small amount of memory per unused section (usually
sizeof(unsigned long)) but the complexity of dynamically allocating the memory
is quite high.
Additional credit to Andy Whitcroft who reviewed up an earlier implementation
of the mechanism an suggested how to make it a *lot* cleaner.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:47 +00:00
|
|
|
#ifndef CONFIG_SPARSEMEM
|
|
|
|
/*
|
|
|
|
* Calculate the size of the zone->blockflags rounded to an unsigned long
|
2007-10-16 08:26:01 +00:00
|
|
|
* Start by making sure zonesize is a multiple of pageblock_order by rounding
|
|
|
|
* up. Then use 1 NR_PAGEBLOCK_BITS worth of bits per pageblock, finally
|
Add a bitmap that is used to track flags affecting a block of pages
Here is the latest revision of the anti-fragmentation patches. Of particular
note in this version is special treatment of high-order atomic allocations.
Care is taken to group them together and avoid grouping pages of other types
near them. Artifical tests imply that it works. I'm trying to get the
hardware together that would allow setting up of a "real" test. If anyone
already has a setup and test that can trigger the atomic-allocation problem,
I'd appreciate a test of these patches and a report. The second major change
is that these patches will apply cleanly with patches that implement
anti-fragmentation through zones.
kernbench shows effectively no performance difference varying between -0.2%
and +2% on a variety of test machines. Success rates for huge page allocation
are dramatically increased. For example, on a ppc64 machine, the vanilla
kernel was only able to allocate 1% of memory as a hugepage and this was due
to a single hugepage reserved as min_free_kbytes. With these patches applied,
17% was allocatable as superpages. With reclaim-related fixes from Andy
Whitcroft, it was 40% and further reclaim-related improvements should increase
this further.
Changelog Since V28
o Group high-order atomic allocations together
o It is no longer required to set min_free_kbytes to 10% of memory. A value
of 16384 in most cases will be sufficient
o Now applied with zone-based anti-fragmentation
o Fix incorrect VM_BUG_ON within buffered_rmqueue()
o Reorder the stack so later patches do not back out work from earlier patches
o Fix bug were journal pages were being treated as movable
o Bias placement of non-movable pages to lower PFNs
o More agressive clustering of reclaimable pages in reactions to workloads
like updatedb that flood the size of inode caches
Changelog Since V27
o Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving
the mistaken impression that it was the 100% solution for high order
allocations. Instead, it greatly increases the chances high-order
allocations will succeed and lays the foundation for defragmentation and
memory hot-remove to work properly
o Redefine page groupings based on ability to migrate or reclaim instead of
basing on reclaimability alone
o Get rid of spurious inits
o Per-cpu lists are no longer split up per-type. Instead the per-cpu list is
searched for a page of the appropriate type
o Added more explanation commentary
o Fix up bug in pageblock code where bitmap was used before being initalised
Changelog Since V26
o Fix double init of lists in setup_pageset
Changelog Since V25
o Fix loop order of for_each_rclmtype_order so that order of loop matches args
o gfpflags_to_rclmtype uses gfp_t instead of unsigned long
o Rename get_pageblock_type() to get_page_rclmtype()
o Fix alignment problem in move_freepages()
o Add mechanism for assigning flags to blocks of pages instead of page->flags
o On fallback, do not examine the preferred list of free pages a second time
The purpose of these patches is to reduce external fragmentation by grouping
pages of related types together. When pages are migrated (or reclaimed under
memory pressure), large contiguous pages will be freed.
This patch works by categorising allocations by their ability to migrate;
Movable - The pages may be moved with the page migration mechanism. These are
generally userspace pages.
Reclaimable - These are allocations for some kernel caches that are
reclaimable or allocations that are known to be very short-lived.
Unmovable - These are pages that are allocated by the kernel that
are not trivially reclaimed. For example, the memory allocated for a
loaded module would be in this category. By default, allocations are
considered to be of this type
HighAtomic - These are high-order allocations belonging to callers that
cannot sleep or perform any IO. In practice, this is restricted to
jumbo frame allocation for network receive. It is assumed that the
allocations are short-lived
Instead of having one MAX_ORDER-sized array of free lists in struct free_area,
there is one for each type of reclaimability. Once a 2^MAX_ORDER block of
pages is split for a type of allocation, it is added to the free-lists for
that type, in effect reserving it. Hence, over time, pages of the different
types can be clustered together.
When the preferred freelists are expired, the largest possible block is taken
from an alternative list. Buddies that are split from that large block are
placed on the preferred allocation-type freelists to mitigate fragmentation.
This implementation gives best-effort for low fragmentation in all zones.
Ideally, min_free_kbytes needs to be set to a value equal to 4 * (1 <<
(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for
example.
Our tests show that about 60-70% of physical memory can be allocated on a
desktop after a few days uptime. In benchmarks and stress tests, we are
finding that 80% of memory is available as contiguous blocks at the end of the
test. To compare, a standard kernel was getting < 1% of memory as large pages
on a desktop and about 8-12% of memory as large pages at the end of stress
tests.
Following this email are 12 patches that implement thie page grouping feature.
The first patch introduces a mechanism for storing flags related to a whole
block of pages. Then allocations are split between movable and all other
allocations. Following that are patches to deal with per-cpu pages and make
the mechanism configurable. The next patch moves free pages between lists
when partially allocated blocks are used for pages of another migrate type.
The second last patch groups reclaimable kernel allocations such as inode
caches together. The final patch related to groupings keeps high-order atomic
allocations.
The last two patches are more concerned with control of fragmentation. The
second last patch biases placement of non-movable allocations towards the
start of memory. This is with a view of supporting memory hot-remove of DIMMs
with higher PFNs in the future. The biasing could be enforced a lot heavier
but it would cost. The last patch agressively clusters reclaimable pages like
inode caches together.
The fragmentation reduction strategy needs to track if pages within a block
can be moved or reclaimed so that pages are freed to the appropriate list.
This patch adds a bitmap for flags affecting a whole a MAX_ORDER block of
pages.
In non-SPARSEMEM configurations, the bitmap is stored in the struct zone and
allocated during initialisation. SPARSEMEM statically allocates the bitmap in
a struct mem_section so that bitmaps do not have to be resized during memory
hotadd. This wastes a small amount of memory per unused section (usually
sizeof(unsigned long)) but the complexity of dynamically allocating the memory
is quite high.
Additional credit to Andy Whitcroft who reviewed up an earlier implementation
of the mechanism an suggested how to make it a *lot* cleaner.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:47 +00:00
|
|
|
* round what is now in bits to nearest long in bits, then return it in
|
|
|
|
* bytes.
|
|
|
|
*/
|
2013-02-18 17:58:02 +00:00
|
|
|
static unsigned long __init usemap_size(unsigned long zone_start_pfn, unsigned long zonesize)
|
Add a bitmap that is used to track flags affecting a block of pages
Here is the latest revision of the anti-fragmentation patches. Of particular
note in this version is special treatment of high-order atomic allocations.
Care is taken to group them together and avoid grouping pages of other types
near them. Artifical tests imply that it works. I'm trying to get the
hardware together that would allow setting up of a "real" test. If anyone
already has a setup and test that can trigger the atomic-allocation problem,
I'd appreciate a test of these patches and a report. The second major change
is that these patches will apply cleanly with patches that implement
anti-fragmentation through zones.
kernbench shows effectively no performance difference varying between -0.2%
and +2% on a variety of test machines. Success rates for huge page allocation
are dramatically increased. For example, on a ppc64 machine, the vanilla
kernel was only able to allocate 1% of memory as a hugepage and this was due
to a single hugepage reserved as min_free_kbytes. With these patches applied,
17% was allocatable as superpages. With reclaim-related fixes from Andy
Whitcroft, it was 40% and further reclaim-related improvements should increase
this further.
Changelog Since V28
o Group high-order atomic allocations together
o It is no longer required to set min_free_kbytes to 10% of memory. A value
of 16384 in most cases will be sufficient
o Now applied with zone-based anti-fragmentation
o Fix incorrect VM_BUG_ON within buffered_rmqueue()
o Reorder the stack so later patches do not back out work from earlier patches
o Fix bug were journal pages were being treated as movable
o Bias placement of non-movable pages to lower PFNs
o More agressive clustering of reclaimable pages in reactions to workloads
like updatedb that flood the size of inode caches
Changelog Since V27
o Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving
the mistaken impression that it was the 100% solution for high order
allocations. Instead, it greatly increases the chances high-order
allocations will succeed and lays the foundation for defragmentation and
memory hot-remove to work properly
o Redefine page groupings based on ability to migrate or reclaim instead of
basing on reclaimability alone
o Get rid of spurious inits
o Per-cpu lists are no longer split up per-type. Instead the per-cpu list is
searched for a page of the appropriate type
o Added more explanation commentary
o Fix up bug in pageblock code where bitmap was used before being initalised
Changelog Since V26
o Fix double init of lists in setup_pageset
Changelog Since V25
o Fix loop order of for_each_rclmtype_order so that order of loop matches args
o gfpflags_to_rclmtype uses gfp_t instead of unsigned long
o Rename get_pageblock_type() to get_page_rclmtype()
o Fix alignment problem in move_freepages()
o Add mechanism for assigning flags to blocks of pages instead of page->flags
o On fallback, do not examine the preferred list of free pages a second time
The purpose of these patches is to reduce external fragmentation by grouping
pages of related types together. When pages are migrated (or reclaimed under
memory pressure), large contiguous pages will be freed.
This patch works by categorising allocations by their ability to migrate;
Movable - The pages may be moved with the page migration mechanism. These are
generally userspace pages.
Reclaimable - These are allocations for some kernel caches that are
reclaimable or allocations that are known to be very short-lived.
Unmovable - These are pages that are allocated by the kernel that
are not trivially reclaimed. For example, the memory allocated for a
loaded module would be in this category. By default, allocations are
considered to be of this type
HighAtomic - These are high-order allocations belonging to callers that
cannot sleep or perform any IO. In practice, this is restricted to
jumbo frame allocation for network receive. It is assumed that the
allocations are short-lived
Instead of having one MAX_ORDER-sized array of free lists in struct free_area,
there is one for each type of reclaimability. Once a 2^MAX_ORDER block of
pages is split for a type of allocation, it is added to the free-lists for
that type, in effect reserving it. Hence, over time, pages of the different
types can be clustered together.
When the preferred freelists are expired, the largest possible block is taken
from an alternative list. Buddies that are split from that large block are
placed on the preferred allocation-type freelists to mitigate fragmentation.
This implementation gives best-effort for low fragmentation in all zones.
Ideally, min_free_kbytes needs to be set to a value equal to 4 * (1 <<
(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for
example.
Our tests show that about 60-70% of physical memory can be allocated on a
desktop after a few days uptime. In benchmarks and stress tests, we are
finding that 80% of memory is available as contiguous blocks at the end of the
test. To compare, a standard kernel was getting < 1% of memory as large pages
on a desktop and about 8-12% of memory as large pages at the end of stress
tests.
Following this email are 12 patches that implement thie page grouping feature.
The first patch introduces a mechanism for storing flags related to a whole
block of pages. Then allocations are split between movable and all other
allocations. Following that are patches to deal with per-cpu pages and make
the mechanism configurable. The next patch moves free pages between lists
when partially allocated blocks are used for pages of another migrate type.
The second last patch groups reclaimable kernel allocations such as inode
caches together. The final patch related to groupings keeps high-order atomic
allocations.
The last two patches are more concerned with control of fragmentation. The
second last patch biases placement of non-movable allocations towards the
start of memory. This is with a view of supporting memory hot-remove of DIMMs
with higher PFNs in the future. The biasing could be enforced a lot heavier
but it would cost. The last patch agressively clusters reclaimable pages like
inode caches together.
The fragmentation reduction strategy needs to track if pages within a block
can be moved or reclaimed so that pages are freed to the appropriate list.
This patch adds a bitmap for flags affecting a whole a MAX_ORDER block of
pages.
In non-SPARSEMEM configurations, the bitmap is stored in the struct zone and
allocated during initialisation. SPARSEMEM statically allocates the bitmap in
a struct mem_section so that bitmaps do not have to be resized during memory
hotadd. This wastes a small amount of memory per unused section (usually
sizeof(unsigned long)) but the complexity of dynamically allocating the memory
is quite high.
Additional credit to Andy Whitcroft who reviewed up an earlier implementation
of the mechanism an suggested how to make it a *lot* cleaner.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:47 +00:00
|
|
|
{
|
|
|
|
unsigned long usemapsize;
|
|
|
|
|
2013-02-18 17:58:02 +00:00
|
|
|
zonesize += zone_start_pfn & (pageblock_nr_pages-1);
|
2007-10-16 08:26:01 +00:00
|
|
|
usemapsize = roundup(zonesize, pageblock_nr_pages);
|
|
|
|
usemapsize = usemapsize >> pageblock_order;
|
Add a bitmap that is used to track flags affecting a block of pages
Here is the latest revision of the anti-fragmentation patches. Of particular
note in this version is special treatment of high-order atomic allocations.
Care is taken to group them together and avoid grouping pages of other types
near them. Artifical tests imply that it works. I'm trying to get the
hardware together that would allow setting up of a "real" test. If anyone
already has a setup and test that can trigger the atomic-allocation problem,
I'd appreciate a test of these patches and a report. The second major change
is that these patches will apply cleanly with patches that implement
anti-fragmentation through zones.
kernbench shows effectively no performance difference varying between -0.2%
and +2% on a variety of test machines. Success rates for huge page allocation
are dramatically increased. For example, on a ppc64 machine, the vanilla
kernel was only able to allocate 1% of memory as a hugepage and this was due
to a single hugepage reserved as min_free_kbytes. With these patches applied,
17% was allocatable as superpages. With reclaim-related fixes from Andy
Whitcroft, it was 40% and further reclaim-related improvements should increase
this further.
Changelog Since V28
o Group high-order atomic allocations together
o It is no longer required to set min_free_kbytes to 10% of memory. A value
of 16384 in most cases will be sufficient
o Now applied with zone-based anti-fragmentation
o Fix incorrect VM_BUG_ON within buffered_rmqueue()
o Reorder the stack so later patches do not back out work from earlier patches
o Fix bug were journal pages were being treated as movable
o Bias placement of non-movable pages to lower PFNs
o More agressive clustering of reclaimable pages in reactions to workloads
like updatedb that flood the size of inode caches
Changelog Since V27
o Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving
the mistaken impression that it was the 100% solution for high order
allocations. Instead, it greatly increases the chances high-order
allocations will succeed and lays the foundation for defragmentation and
memory hot-remove to work properly
o Redefine page groupings based on ability to migrate or reclaim instead of
basing on reclaimability alone
o Get rid of spurious inits
o Per-cpu lists are no longer split up per-type. Instead the per-cpu list is
searched for a page of the appropriate type
o Added more explanation commentary
o Fix up bug in pageblock code where bitmap was used before being initalised
Changelog Since V26
o Fix double init of lists in setup_pageset
Changelog Since V25
o Fix loop order of for_each_rclmtype_order so that order of loop matches args
o gfpflags_to_rclmtype uses gfp_t instead of unsigned long
o Rename get_pageblock_type() to get_page_rclmtype()
o Fix alignment problem in move_freepages()
o Add mechanism for assigning flags to blocks of pages instead of page->flags
o On fallback, do not examine the preferred list of free pages a second time
The purpose of these patches is to reduce external fragmentation by grouping
pages of related types together. When pages are migrated (or reclaimed under
memory pressure), large contiguous pages will be freed.
This patch works by categorising allocations by their ability to migrate;
Movable - The pages may be moved with the page migration mechanism. These are
generally userspace pages.
Reclaimable - These are allocations for some kernel caches that are
reclaimable or allocations that are known to be very short-lived.
Unmovable - These are pages that are allocated by the kernel that
are not trivially reclaimed. For example, the memory allocated for a
loaded module would be in this category. By default, allocations are
considered to be of this type
HighAtomic - These are high-order allocations belonging to callers that
cannot sleep or perform any IO. In practice, this is restricted to
jumbo frame allocation for network receive. It is assumed that the
allocations are short-lived
Instead of having one MAX_ORDER-sized array of free lists in struct free_area,
there is one for each type of reclaimability. Once a 2^MAX_ORDER block of
pages is split for a type of allocation, it is added to the free-lists for
that type, in effect reserving it. Hence, over time, pages of the different
types can be clustered together.
When the preferred freelists are expired, the largest possible block is taken
from an alternative list. Buddies that are split from that large block are
placed on the preferred allocation-type freelists to mitigate fragmentation.
This implementation gives best-effort for low fragmentation in all zones.
Ideally, min_free_kbytes needs to be set to a value equal to 4 * (1 <<
(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for
example.
Our tests show that about 60-70% of physical memory can be allocated on a
desktop after a few days uptime. In benchmarks and stress tests, we are
finding that 80% of memory is available as contiguous blocks at the end of the
test. To compare, a standard kernel was getting < 1% of memory as large pages
on a desktop and about 8-12% of memory as large pages at the end of stress
tests.
Following this email are 12 patches that implement thie page grouping feature.
The first patch introduces a mechanism for storing flags related to a whole
block of pages. Then allocations are split between movable and all other
allocations. Following that are patches to deal with per-cpu pages and make
the mechanism configurable. The next patch moves free pages between lists
when partially allocated blocks are used for pages of another migrate type.
The second last patch groups reclaimable kernel allocations such as inode
caches together. The final patch related to groupings keeps high-order atomic
allocations.
The last two patches are more concerned with control of fragmentation. The
second last patch biases placement of non-movable allocations towards the
start of memory. This is with a view of supporting memory hot-remove of DIMMs
with higher PFNs in the future. The biasing could be enforced a lot heavier
but it would cost. The last patch agressively clusters reclaimable pages like
inode caches together.
The fragmentation reduction strategy needs to track if pages within a block
can be moved or reclaimed so that pages are freed to the appropriate list.
This patch adds a bitmap for flags affecting a whole a MAX_ORDER block of
pages.
In non-SPARSEMEM configurations, the bitmap is stored in the struct zone and
allocated during initialisation. SPARSEMEM statically allocates the bitmap in
a struct mem_section so that bitmaps do not have to be resized during memory
hotadd. This wastes a small amount of memory per unused section (usually
sizeof(unsigned long)) but the complexity of dynamically allocating the memory
is quite high.
Additional credit to Andy Whitcroft who reviewed up an earlier implementation
of the mechanism an suggested how to make it a *lot* cleaner.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:47 +00:00
|
|
|
usemapsize *= NR_PAGEBLOCK_BITS;
|
|
|
|
usemapsize = roundup(usemapsize, 8 * sizeof(unsigned long));
|
|
|
|
|
|
|
|
return usemapsize / 8;
|
|
|
|
}
|
|
|
|
|
2021-02-24 20:06:20 +00:00
|
|
|
static void __ref setup_usemap(struct zone *zone)
|
Add a bitmap that is used to track flags affecting a block of pages
Here is the latest revision of the anti-fragmentation patches. Of particular
note in this version is special treatment of high-order atomic allocations.
Care is taken to group them together and avoid grouping pages of other types
near them. Artifical tests imply that it works. I'm trying to get the
hardware together that would allow setting up of a "real" test. If anyone
already has a setup and test that can trigger the atomic-allocation problem,
I'd appreciate a test of these patches and a report. The second major change
is that these patches will apply cleanly with patches that implement
anti-fragmentation through zones.
kernbench shows effectively no performance difference varying between -0.2%
and +2% on a variety of test machines. Success rates for huge page allocation
are dramatically increased. For example, on a ppc64 machine, the vanilla
kernel was only able to allocate 1% of memory as a hugepage and this was due
to a single hugepage reserved as min_free_kbytes. With these patches applied,
17% was allocatable as superpages. With reclaim-related fixes from Andy
Whitcroft, it was 40% and further reclaim-related improvements should increase
this further.
Changelog Since V28
o Group high-order atomic allocations together
o It is no longer required to set min_free_kbytes to 10% of memory. A value
of 16384 in most cases will be sufficient
o Now applied with zone-based anti-fragmentation
o Fix incorrect VM_BUG_ON within buffered_rmqueue()
o Reorder the stack so later patches do not back out work from earlier patches
o Fix bug were journal pages were being treated as movable
o Bias placement of non-movable pages to lower PFNs
o More agressive clustering of reclaimable pages in reactions to workloads
like updatedb that flood the size of inode caches
Changelog Since V27
o Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving
the mistaken impression that it was the 100% solution for high order
allocations. Instead, it greatly increases the chances high-order
allocations will succeed and lays the foundation for defragmentation and
memory hot-remove to work properly
o Redefine page groupings based on ability to migrate or reclaim instead of
basing on reclaimability alone
o Get rid of spurious inits
o Per-cpu lists are no longer split up per-type. Instead the per-cpu list is
searched for a page of the appropriate type
o Added more explanation commentary
o Fix up bug in pageblock code where bitmap was used before being initalised
Changelog Since V26
o Fix double init of lists in setup_pageset
Changelog Since V25
o Fix loop order of for_each_rclmtype_order so that order of loop matches args
o gfpflags_to_rclmtype uses gfp_t instead of unsigned long
o Rename get_pageblock_type() to get_page_rclmtype()
o Fix alignment problem in move_freepages()
o Add mechanism for assigning flags to blocks of pages instead of page->flags
o On fallback, do not examine the preferred list of free pages a second time
The purpose of these patches is to reduce external fragmentation by grouping
pages of related types together. When pages are migrated (or reclaimed under
memory pressure), large contiguous pages will be freed.
This patch works by categorising allocations by their ability to migrate;
Movable - The pages may be moved with the page migration mechanism. These are
generally userspace pages.
Reclaimable - These are allocations for some kernel caches that are
reclaimable or allocations that are known to be very short-lived.
Unmovable - These are pages that are allocated by the kernel that
are not trivially reclaimed. For example, the memory allocated for a
loaded module would be in this category. By default, allocations are
considered to be of this type
HighAtomic - These are high-order allocations belonging to callers that
cannot sleep or perform any IO. In practice, this is restricted to
jumbo frame allocation for network receive. It is assumed that the
allocations are short-lived
Instead of having one MAX_ORDER-sized array of free lists in struct free_area,
there is one for each type of reclaimability. Once a 2^MAX_ORDER block of
pages is split for a type of allocation, it is added to the free-lists for
that type, in effect reserving it. Hence, over time, pages of the different
types can be clustered together.
When the preferred freelists are expired, the largest possible block is taken
from an alternative list. Buddies that are split from that large block are
placed on the preferred allocation-type freelists to mitigate fragmentation.
This implementation gives best-effort for low fragmentation in all zones.
Ideally, min_free_kbytes needs to be set to a value equal to 4 * (1 <<
(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for
example.
Our tests show that about 60-70% of physical memory can be allocated on a
desktop after a few days uptime. In benchmarks and stress tests, we are
finding that 80% of memory is available as contiguous blocks at the end of the
test. To compare, a standard kernel was getting < 1% of memory as large pages
on a desktop and about 8-12% of memory as large pages at the end of stress
tests.
Following this email are 12 patches that implement thie page grouping feature.
The first patch introduces a mechanism for storing flags related to a whole
block of pages. Then allocations are split between movable and all other
allocations. Following that are patches to deal with per-cpu pages and make
the mechanism configurable. The next patch moves free pages between lists
when partially allocated blocks are used for pages of another migrate type.
The second last patch groups reclaimable kernel allocations such as inode
caches together. The final patch related to groupings keeps high-order atomic
allocations.
The last two patches are more concerned with control of fragmentation. The
second last patch biases placement of non-movable allocations towards the
start of memory. This is with a view of supporting memory hot-remove of DIMMs
with higher PFNs in the future. The biasing could be enforced a lot heavier
but it would cost. The last patch agressively clusters reclaimable pages like
inode caches together.
The fragmentation reduction strategy needs to track if pages within a block
can be moved or reclaimed so that pages are freed to the appropriate list.
This patch adds a bitmap for flags affecting a whole a MAX_ORDER block of
pages.
In non-SPARSEMEM configurations, the bitmap is stored in the struct zone and
allocated during initialisation. SPARSEMEM statically allocates the bitmap in
a struct mem_section so that bitmaps do not have to be resized during memory
hotadd. This wastes a small amount of memory per unused section (usually
sizeof(unsigned long)) but the complexity of dynamically allocating the memory
is quite high.
Additional credit to Andy Whitcroft who reviewed up an earlier implementation
of the mechanism an suggested how to make it a *lot* cleaner.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:47 +00:00
|
|
|
{
|
2021-02-24 20:06:20 +00:00
|
|
|
unsigned long usemapsize = usemap_size(zone->zone_start_pfn,
|
|
|
|
zone->spanned_pages);
|
Add a bitmap that is used to track flags affecting a block of pages
Here is the latest revision of the anti-fragmentation patches. Of particular
note in this version is special treatment of high-order atomic allocations.
Care is taken to group them together and avoid grouping pages of other types
near them. Artifical tests imply that it works. I'm trying to get the
hardware together that would allow setting up of a "real" test. If anyone
already has a setup and test that can trigger the atomic-allocation problem,
I'd appreciate a test of these patches and a report. The second major change
is that these patches will apply cleanly with patches that implement
anti-fragmentation through zones.
kernbench shows effectively no performance difference varying between -0.2%
and +2% on a variety of test machines. Success rates for huge page allocation
are dramatically increased. For example, on a ppc64 machine, the vanilla
kernel was only able to allocate 1% of memory as a hugepage and this was due
to a single hugepage reserved as min_free_kbytes. With these patches applied,
17% was allocatable as superpages. With reclaim-related fixes from Andy
Whitcroft, it was 40% and further reclaim-related improvements should increase
this further.
Changelog Since V28
o Group high-order atomic allocations together
o It is no longer required to set min_free_kbytes to 10% of memory. A value
of 16384 in most cases will be sufficient
o Now applied with zone-based anti-fragmentation
o Fix incorrect VM_BUG_ON within buffered_rmqueue()
o Reorder the stack so later patches do not back out work from earlier patches
o Fix bug were journal pages were being treated as movable
o Bias placement of non-movable pages to lower PFNs
o More agressive clustering of reclaimable pages in reactions to workloads
like updatedb that flood the size of inode caches
Changelog Since V27
o Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving
the mistaken impression that it was the 100% solution for high order
allocations. Instead, it greatly increases the chances high-order
allocations will succeed and lays the foundation for defragmentation and
memory hot-remove to work properly
o Redefine page groupings based on ability to migrate or reclaim instead of
basing on reclaimability alone
o Get rid of spurious inits
o Per-cpu lists are no longer split up per-type. Instead the per-cpu list is
searched for a page of the appropriate type
o Added more explanation commentary
o Fix up bug in pageblock code where bitmap was used before being initalised
Changelog Since V26
o Fix double init of lists in setup_pageset
Changelog Since V25
o Fix loop order of for_each_rclmtype_order so that order of loop matches args
o gfpflags_to_rclmtype uses gfp_t instead of unsigned long
o Rename get_pageblock_type() to get_page_rclmtype()
o Fix alignment problem in move_freepages()
o Add mechanism for assigning flags to blocks of pages instead of page->flags
o On fallback, do not examine the preferred list of free pages a second time
The purpose of these patches is to reduce external fragmentation by grouping
pages of related types together. When pages are migrated (or reclaimed under
memory pressure), large contiguous pages will be freed.
This patch works by categorising allocations by their ability to migrate;
Movable - The pages may be moved with the page migration mechanism. These are
generally userspace pages.
Reclaimable - These are allocations for some kernel caches that are
reclaimable or allocations that are known to be very short-lived.
Unmovable - These are pages that are allocated by the kernel that
are not trivially reclaimed. For example, the memory allocated for a
loaded module would be in this category. By default, allocations are
considered to be of this type
HighAtomic - These are high-order allocations belonging to callers that
cannot sleep or perform any IO. In practice, this is restricted to
jumbo frame allocation for network receive. It is assumed that the
allocations are short-lived
Instead of having one MAX_ORDER-sized array of free lists in struct free_area,
there is one for each type of reclaimability. Once a 2^MAX_ORDER block of
pages is split for a type of allocation, it is added to the free-lists for
that type, in effect reserving it. Hence, over time, pages of the different
types can be clustered together.
When the preferred freelists are expired, the largest possible block is taken
from an alternative list. Buddies that are split from that large block are
placed on the preferred allocation-type freelists to mitigate fragmentation.
This implementation gives best-effort for low fragmentation in all zones.
Ideally, min_free_kbytes needs to be set to a value equal to 4 * (1 <<
(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for
example.
Our tests show that about 60-70% of physical memory can be allocated on a
desktop after a few days uptime. In benchmarks and stress tests, we are
finding that 80% of memory is available as contiguous blocks at the end of the
test. To compare, a standard kernel was getting < 1% of memory as large pages
on a desktop and about 8-12% of memory as large pages at the end of stress
tests.
Following this email are 12 patches that implement thie page grouping feature.
The first patch introduces a mechanism for storing flags related to a whole
block of pages. Then allocations are split between movable and all other
allocations. Following that are patches to deal with per-cpu pages and make
the mechanism configurable. The next patch moves free pages between lists
when partially allocated blocks are used for pages of another migrate type.
The second last patch groups reclaimable kernel allocations such as inode
caches together. The final patch related to groupings keeps high-order atomic
allocations.
The last two patches are more concerned with control of fragmentation. The
second last patch biases placement of non-movable allocations towards the
start of memory. This is with a view of supporting memory hot-remove of DIMMs
with higher PFNs in the future. The biasing could be enforced a lot heavier
but it would cost. The last patch agressively clusters reclaimable pages like
inode caches together.
The fragmentation reduction strategy needs to track if pages within a block
can be moved or reclaimed so that pages are freed to the appropriate list.
This patch adds a bitmap for flags affecting a whole a MAX_ORDER block of
pages.
In non-SPARSEMEM configurations, the bitmap is stored in the struct zone and
allocated during initialisation. SPARSEMEM statically allocates the bitmap in
a struct mem_section so that bitmaps do not have to be resized during memory
hotadd. This wastes a small amount of memory per unused section (usually
sizeof(unsigned long)) but the complexity of dynamically allocating the memory
is quite high.
Additional credit to Andy Whitcroft who reviewed up an earlier implementation
of the mechanism an suggested how to make it a *lot* cleaner.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:47 +00:00
|
|
|
zone->pageblock_flags = NULL;
|
2019-03-05 23:46:43 +00:00
|
|
|
if (usemapsize) {
|
2014-01-21 23:50:25 +00:00
|
|
|
zone->pageblock_flags =
|
2019-03-12 06:30:42 +00:00
|
|
|
memblock_alloc_node(usemapsize, SMP_CACHE_BYTES,
|
2021-02-24 20:06:20 +00:00
|
|
|
zone_to_nid(zone));
|
2019-03-05 23:46:43 +00:00
|
|
|
if (!zone->pageblock_flags)
|
|
|
|
panic("Failed to allocate %ld bytes for zone %s pageblock flags on node %d\n",
|
2021-02-24 20:06:20 +00:00
|
|
|
usemapsize, zone->name, zone_to_nid(zone));
|
2019-03-05 23:46:43 +00:00
|
|
|
}
|
Add a bitmap that is used to track flags affecting a block of pages
Here is the latest revision of the anti-fragmentation patches. Of particular
note in this version is special treatment of high-order atomic allocations.
Care is taken to group them together and avoid grouping pages of other types
near them. Artifical tests imply that it works. I'm trying to get the
hardware together that would allow setting up of a "real" test. If anyone
already has a setup and test that can trigger the atomic-allocation problem,
I'd appreciate a test of these patches and a report. The second major change
is that these patches will apply cleanly with patches that implement
anti-fragmentation through zones.
kernbench shows effectively no performance difference varying between -0.2%
and +2% on a variety of test machines. Success rates for huge page allocation
are dramatically increased. For example, on a ppc64 machine, the vanilla
kernel was only able to allocate 1% of memory as a hugepage and this was due
to a single hugepage reserved as min_free_kbytes. With these patches applied,
17% was allocatable as superpages. With reclaim-related fixes from Andy
Whitcroft, it was 40% and further reclaim-related improvements should increase
this further.
Changelog Since V28
o Group high-order atomic allocations together
o It is no longer required to set min_free_kbytes to 10% of memory. A value
of 16384 in most cases will be sufficient
o Now applied with zone-based anti-fragmentation
o Fix incorrect VM_BUG_ON within buffered_rmqueue()
o Reorder the stack so later patches do not back out work from earlier patches
o Fix bug were journal pages were being treated as movable
o Bias placement of non-movable pages to lower PFNs
o More agressive clustering of reclaimable pages in reactions to workloads
like updatedb that flood the size of inode caches
Changelog Since V27
o Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving
the mistaken impression that it was the 100% solution for high order
allocations. Instead, it greatly increases the chances high-order
allocations will succeed and lays the foundation for defragmentation and
memory hot-remove to work properly
o Redefine page groupings based on ability to migrate or reclaim instead of
basing on reclaimability alone
o Get rid of spurious inits
o Per-cpu lists are no longer split up per-type. Instead the per-cpu list is
searched for a page of the appropriate type
o Added more explanation commentary
o Fix up bug in pageblock code where bitmap was used before being initalised
Changelog Since V26
o Fix double init of lists in setup_pageset
Changelog Since V25
o Fix loop order of for_each_rclmtype_order so that order of loop matches args
o gfpflags_to_rclmtype uses gfp_t instead of unsigned long
o Rename get_pageblock_type() to get_page_rclmtype()
o Fix alignment problem in move_freepages()
o Add mechanism for assigning flags to blocks of pages instead of page->flags
o On fallback, do not examine the preferred list of free pages a second time
The purpose of these patches is to reduce external fragmentation by grouping
pages of related types together. When pages are migrated (or reclaimed under
memory pressure), large contiguous pages will be freed.
This patch works by categorising allocations by their ability to migrate;
Movable - The pages may be moved with the page migration mechanism. These are
generally userspace pages.
Reclaimable - These are allocations for some kernel caches that are
reclaimable or allocations that are known to be very short-lived.
Unmovable - These are pages that are allocated by the kernel that
are not trivially reclaimed. For example, the memory allocated for a
loaded module would be in this category. By default, allocations are
considered to be of this type
HighAtomic - These are high-order allocations belonging to callers that
cannot sleep or perform any IO. In practice, this is restricted to
jumbo frame allocation for network receive. It is assumed that the
allocations are short-lived
Instead of having one MAX_ORDER-sized array of free lists in struct free_area,
there is one for each type of reclaimability. Once a 2^MAX_ORDER block of
pages is split for a type of allocation, it is added to the free-lists for
that type, in effect reserving it. Hence, over time, pages of the different
types can be clustered together.
When the preferred freelists are expired, the largest possible block is taken
from an alternative list. Buddies that are split from that large block are
placed on the preferred allocation-type freelists to mitigate fragmentation.
This implementation gives best-effort for low fragmentation in all zones.
Ideally, min_free_kbytes needs to be set to a value equal to 4 * (1 <<
(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for
example.
Our tests show that about 60-70% of physical memory can be allocated on a
desktop after a few days uptime. In benchmarks and stress tests, we are
finding that 80% of memory is available as contiguous blocks at the end of the
test. To compare, a standard kernel was getting < 1% of memory as large pages
on a desktop and about 8-12% of memory as large pages at the end of stress
tests.
Following this email are 12 patches that implement thie page grouping feature.
The first patch introduces a mechanism for storing flags related to a whole
block of pages. Then allocations are split between movable and all other
allocations. Following that are patches to deal with per-cpu pages and make
the mechanism configurable. The next patch moves free pages between lists
when partially allocated blocks are used for pages of another migrate type.
The second last patch groups reclaimable kernel allocations such as inode
caches together. The final patch related to groupings keeps high-order atomic
allocations.
The last two patches are more concerned with control of fragmentation. The
second last patch biases placement of non-movable allocations towards the
start of memory. This is with a view of supporting memory hot-remove of DIMMs
with higher PFNs in the future. The biasing could be enforced a lot heavier
but it would cost. The last patch agressively clusters reclaimable pages like
inode caches together.
The fragmentation reduction strategy needs to track if pages within a block
can be moved or reclaimed so that pages are freed to the appropriate list.
This patch adds a bitmap for flags affecting a whole a MAX_ORDER block of
pages.
In non-SPARSEMEM configurations, the bitmap is stored in the struct zone and
allocated during initialisation. SPARSEMEM statically allocates the bitmap in
a struct mem_section so that bitmaps do not have to be resized during memory
hotadd. This wastes a small amount of memory per unused section (usually
sizeof(unsigned long)) but the complexity of dynamically allocating the memory
is quite high.
Additional credit to Andy Whitcroft who reviewed up an earlier implementation
of the mechanism an suggested how to make it a *lot* cleaner.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:47 +00:00
|
|
|
}
|
|
|
|
#else
|
2021-02-24 20:06:20 +00:00
|
|
|
static inline void setup_usemap(struct zone *zone) {}
|
Add a bitmap that is used to track flags affecting a block of pages
Here is the latest revision of the anti-fragmentation patches. Of particular
note in this version is special treatment of high-order atomic allocations.
Care is taken to group them together and avoid grouping pages of other types
near them. Artifical tests imply that it works. I'm trying to get the
hardware together that would allow setting up of a "real" test. If anyone
already has a setup and test that can trigger the atomic-allocation problem,
I'd appreciate a test of these patches and a report. The second major change
is that these patches will apply cleanly with patches that implement
anti-fragmentation through zones.
kernbench shows effectively no performance difference varying between -0.2%
and +2% on a variety of test machines. Success rates for huge page allocation
are dramatically increased. For example, on a ppc64 machine, the vanilla
kernel was only able to allocate 1% of memory as a hugepage and this was due
to a single hugepage reserved as min_free_kbytes. With these patches applied,
17% was allocatable as superpages. With reclaim-related fixes from Andy
Whitcroft, it was 40% and further reclaim-related improvements should increase
this further.
Changelog Since V28
o Group high-order atomic allocations together
o It is no longer required to set min_free_kbytes to 10% of memory. A value
of 16384 in most cases will be sufficient
o Now applied with zone-based anti-fragmentation
o Fix incorrect VM_BUG_ON within buffered_rmqueue()
o Reorder the stack so later patches do not back out work from earlier patches
o Fix bug were journal pages were being treated as movable
o Bias placement of non-movable pages to lower PFNs
o More agressive clustering of reclaimable pages in reactions to workloads
like updatedb that flood the size of inode caches
Changelog Since V27
o Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving
the mistaken impression that it was the 100% solution for high order
allocations. Instead, it greatly increases the chances high-order
allocations will succeed and lays the foundation for defragmentation and
memory hot-remove to work properly
o Redefine page groupings based on ability to migrate or reclaim instead of
basing on reclaimability alone
o Get rid of spurious inits
o Per-cpu lists are no longer split up per-type. Instead the per-cpu list is
searched for a page of the appropriate type
o Added more explanation commentary
o Fix up bug in pageblock code where bitmap was used before being initalised
Changelog Since V26
o Fix double init of lists in setup_pageset
Changelog Since V25
o Fix loop order of for_each_rclmtype_order so that order of loop matches args
o gfpflags_to_rclmtype uses gfp_t instead of unsigned long
o Rename get_pageblock_type() to get_page_rclmtype()
o Fix alignment problem in move_freepages()
o Add mechanism for assigning flags to blocks of pages instead of page->flags
o On fallback, do not examine the preferred list of free pages a second time
The purpose of these patches is to reduce external fragmentation by grouping
pages of related types together. When pages are migrated (or reclaimed under
memory pressure), large contiguous pages will be freed.
This patch works by categorising allocations by their ability to migrate;
Movable - The pages may be moved with the page migration mechanism. These are
generally userspace pages.
Reclaimable - These are allocations for some kernel caches that are
reclaimable or allocations that are known to be very short-lived.
Unmovable - These are pages that are allocated by the kernel that
are not trivially reclaimed. For example, the memory allocated for a
loaded module would be in this category. By default, allocations are
considered to be of this type
HighAtomic - These are high-order allocations belonging to callers that
cannot sleep or perform any IO. In practice, this is restricted to
jumbo frame allocation for network receive. It is assumed that the
allocations are short-lived
Instead of having one MAX_ORDER-sized array of free lists in struct free_area,
there is one for each type of reclaimability. Once a 2^MAX_ORDER block of
pages is split for a type of allocation, it is added to the free-lists for
that type, in effect reserving it. Hence, over time, pages of the different
types can be clustered together.
When the preferred freelists are expired, the largest possible block is taken
from an alternative list. Buddies that are split from that large block are
placed on the preferred allocation-type freelists to mitigate fragmentation.
This implementation gives best-effort for low fragmentation in all zones.
Ideally, min_free_kbytes needs to be set to a value equal to 4 * (1 <<
(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for
example.
Our tests show that about 60-70% of physical memory can be allocated on a
desktop after a few days uptime. In benchmarks and stress tests, we are
finding that 80% of memory is available as contiguous blocks at the end of the
test. To compare, a standard kernel was getting < 1% of memory as large pages
on a desktop and about 8-12% of memory as large pages at the end of stress
tests.
Following this email are 12 patches that implement thie page grouping feature.
The first patch introduces a mechanism for storing flags related to a whole
block of pages. Then allocations are split between movable and all other
allocations. Following that are patches to deal with per-cpu pages and make
the mechanism configurable. The next patch moves free pages between lists
when partially allocated blocks are used for pages of another migrate type.
The second last patch groups reclaimable kernel allocations such as inode
caches together. The final patch related to groupings keeps high-order atomic
allocations.
The last two patches are more concerned with control of fragmentation. The
second last patch biases placement of non-movable allocations towards the
start of memory. This is with a view of supporting memory hot-remove of DIMMs
with higher PFNs in the future. The biasing could be enforced a lot heavier
but it would cost. The last patch agressively clusters reclaimable pages like
inode caches together.
The fragmentation reduction strategy needs to track if pages within a block
can be moved or reclaimed so that pages are freed to the appropriate list.
This patch adds a bitmap for flags affecting a whole a MAX_ORDER block of
pages.
In non-SPARSEMEM configurations, the bitmap is stored in the struct zone and
allocated during initialisation. SPARSEMEM statically allocates the bitmap in
a struct mem_section so that bitmaps do not have to be resized during memory
hotadd. This wastes a small amount of memory per unused section (usually
sizeof(unsigned long)) but the complexity of dynamically allocating the memory
is quite high.
Additional credit to Andy Whitcroft who reviewed up an earlier implementation
of the mechanism an suggested how to make it a *lot* cleaner.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:25:47 +00:00
|
|
|
#endif /* CONFIG_SPARSEMEM */
|
|
|
|
|
2007-10-16 08:26:01 +00:00
|
|
|
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
|
2007-11-29 00:21:13 +00:00
|
|
|
|
2007-10-16 08:26:01 +00:00
|
|
|
/* Initialise the number of pages represented by NR_PAGEBLOCK_BITS */
|
mm/page_alloc: Introduce free_area_init_core_hotplug
Currently, whenever a new node is created/re-used from the memhotplug
path, we call free_area_init_node()->free_area_init_core(). But there is
some code that we do not really need to run when we are coming from such
path.
free_area_init_core() performs the following actions:
1) Initializes pgdat internals, such as spinlock, waitqueues and more.
2) Account # nr_all_pages and # nr_kernel_pages. These values are used later on
when creating hash tables.
3) Account number of managed_pages per zone, substracting dma_reserved and
memmap pages.
4) Initializes some fields of the zone structure data
5) Calls init_currently_empty_zone to initialize all the freelists
6) Calls memmap_init to initialize all pages belonging to certain zone
When called from memhotplug path, free_area_init_core() only performs
actions #1 and #4.
Action #2 is pointless as the zones do not have any pages since either the
node was freed, or we are re-using it, eitherway all zones belonging to
this node should have 0 pages. For the same reason, action #3 results
always in manages_pages being 0.
Action #5 and #6 are performed later on when onlining the pages:
online_pages()->move_pfn_range_to_zone()->init_currently_empty_zone()
online_pages()->move_pfn_range_to_zone()->memmap_init_zone()
This patch does two things:
First, moves the node/zone initializtion to their own function, so it
allows us to create a small version of free_area_init_core, where we only
perform:
1) Initialization of pgdat internals, such as spinlock, waitqueues and more
4) Initialization of some fields of the zone structure data
These two functions are: pgdat_init_internals() and zone_init_internals().
The second thing this patch does, is to introduce
free_area_init_core_hotplug(), the memhotplug version of
free_area_init_core():
Currently, we call free_area_init_node() from the memhotplug path. In
there, we set some pgdat's fields, and call calculate_node_totalpages().
calculate_node_totalpages() calculates the # of pages the node has.
Since the node is either new, or we are re-using it, the zones belonging
to this node should not have any pages, so there is no point to calculate
this now.
Actually, we re-set these values to 0 later on with the calls to:
reset_node_managed_pages()
reset_node_present_pages()
The # of pages per node and the # of pages per zone will be calculated when
onlining the pages:
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_zone_range()
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_pgdat_range()
Also, since free_area_init_core/free_area_init_node will now only get called during early init, let us replace
__paginginit with __init, so their code gets freed up.
[osalvador@techadventures.net: fix section usage]
Link: http://lkml.kernel.org/r/20180731101752.GA473@techadventures.net
[osalvador@suse.de: v6]
Link: http://lkml.kernel.org/r/20180801122348.21588-6-osalvador@techadventures.net
Link: http://lkml.kernel.org/r/20180730101757.28058-5-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 04:53:43 +00:00
|
|
|
void __init set_pageblock_order(void)
|
2007-10-16 08:26:01 +00:00
|
|
|
{
|
2012-05-29 22:06:31 +00:00
|
|
|
unsigned int order;
|
|
|
|
|
2007-10-16 08:26:01 +00:00
|
|
|
/* Check that pageblock_nr_pages has not already been setup */
|
|
|
|
if (pageblock_order)
|
|
|
|
return;
|
|
|
|
|
2012-05-29 22:06:31 +00:00
|
|
|
if (HPAGE_SHIFT > PAGE_SHIFT)
|
|
|
|
order = HUGETLB_PAGE_ORDER;
|
|
|
|
else
|
|
|
|
order = MAX_ORDER - 1;
|
|
|
|
|
2007-10-16 08:26:01 +00:00
|
|
|
/*
|
|
|
|
* Assume the largest contiguous order of interest is a huge page.
|
2012-05-29 22:06:31 +00:00
|
|
|
* This value may be variable depending on boot parameters on IA64 and
|
|
|
|
* powerpc.
|
2007-10-16 08:26:01 +00:00
|
|
|
*/
|
|
|
|
pageblock_order = order;
|
|
|
|
}
|
|
|
|
#else /* CONFIG_HUGETLB_PAGE_SIZE_VARIABLE */
|
|
|
|
|
2007-11-29 00:21:13 +00:00
|
|
|
/*
|
|
|
|
* When CONFIG_HUGETLB_PAGE_SIZE_VARIABLE is not set, set_pageblock_order()
|
2012-05-29 22:06:31 +00:00
|
|
|
* is unused as pageblock_order is set at compile-time. See
|
|
|
|
* include/linux/pageblock-flags.h for the values of pageblock_order based on
|
|
|
|
* the kernel config
|
2007-11-29 00:21:13 +00:00
|
|
|
*/
|
mm/page_alloc: Introduce free_area_init_core_hotplug
Currently, whenever a new node is created/re-used from the memhotplug
path, we call free_area_init_node()->free_area_init_core(). But there is
some code that we do not really need to run when we are coming from such
path.
free_area_init_core() performs the following actions:
1) Initializes pgdat internals, such as spinlock, waitqueues and more.
2) Account # nr_all_pages and # nr_kernel_pages. These values are used later on
when creating hash tables.
3) Account number of managed_pages per zone, substracting dma_reserved and
memmap pages.
4) Initializes some fields of the zone structure data
5) Calls init_currently_empty_zone to initialize all the freelists
6) Calls memmap_init to initialize all pages belonging to certain zone
When called from memhotplug path, free_area_init_core() only performs
actions #1 and #4.
Action #2 is pointless as the zones do not have any pages since either the
node was freed, or we are re-using it, eitherway all zones belonging to
this node should have 0 pages. For the same reason, action #3 results
always in manages_pages being 0.
Action #5 and #6 are performed later on when onlining the pages:
online_pages()->move_pfn_range_to_zone()->init_currently_empty_zone()
online_pages()->move_pfn_range_to_zone()->memmap_init_zone()
This patch does two things:
First, moves the node/zone initializtion to their own function, so it
allows us to create a small version of free_area_init_core, where we only
perform:
1) Initialization of pgdat internals, such as spinlock, waitqueues and more
4) Initialization of some fields of the zone structure data
These two functions are: pgdat_init_internals() and zone_init_internals().
The second thing this patch does, is to introduce
free_area_init_core_hotplug(), the memhotplug version of
free_area_init_core():
Currently, we call free_area_init_node() from the memhotplug path. In
there, we set some pgdat's fields, and call calculate_node_totalpages().
calculate_node_totalpages() calculates the # of pages the node has.
Since the node is either new, or we are re-using it, the zones belonging
to this node should not have any pages, so there is no point to calculate
this now.
Actually, we re-set these values to 0 later on with the calls to:
reset_node_managed_pages()
reset_node_present_pages()
The # of pages per node and the # of pages per zone will be calculated when
onlining the pages:
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_zone_range()
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_pgdat_range()
Also, since free_area_init_core/free_area_init_node will now only get called during early init, let us replace
__paginginit with __init, so their code gets freed up.
[osalvador@techadventures.net: fix section usage]
Link: http://lkml.kernel.org/r/20180731101752.GA473@techadventures.net
[osalvador@suse.de: v6]
Link: http://lkml.kernel.org/r/20180801122348.21588-6-osalvador@techadventures.net
Link: http://lkml.kernel.org/r/20180730101757.28058-5-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 04:53:43 +00:00
|
|
|
void __init set_pageblock_order(void)
|
2007-11-29 00:21:13 +00:00
|
|
|
{
|
|
|
|
}
|
2007-10-16 08:26:01 +00:00
|
|
|
|
|
|
|
#endif /* CONFIG_HUGETLB_PAGE_SIZE_VARIABLE */
|
|
|
|
|
mm/page_alloc: Introduce free_area_init_core_hotplug
Currently, whenever a new node is created/re-used from the memhotplug
path, we call free_area_init_node()->free_area_init_core(). But there is
some code that we do not really need to run when we are coming from such
path.
free_area_init_core() performs the following actions:
1) Initializes pgdat internals, such as spinlock, waitqueues and more.
2) Account # nr_all_pages and # nr_kernel_pages. These values are used later on
when creating hash tables.
3) Account number of managed_pages per zone, substracting dma_reserved and
memmap pages.
4) Initializes some fields of the zone structure data
5) Calls init_currently_empty_zone to initialize all the freelists
6) Calls memmap_init to initialize all pages belonging to certain zone
When called from memhotplug path, free_area_init_core() only performs
actions #1 and #4.
Action #2 is pointless as the zones do not have any pages since either the
node was freed, or we are re-using it, eitherway all zones belonging to
this node should have 0 pages. For the same reason, action #3 results
always in manages_pages being 0.
Action #5 and #6 are performed later on when onlining the pages:
online_pages()->move_pfn_range_to_zone()->init_currently_empty_zone()
online_pages()->move_pfn_range_to_zone()->memmap_init_zone()
This patch does two things:
First, moves the node/zone initializtion to their own function, so it
allows us to create a small version of free_area_init_core, where we only
perform:
1) Initialization of pgdat internals, such as spinlock, waitqueues and more
4) Initialization of some fields of the zone structure data
These two functions are: pgdat_init_internals() and zone_init_internals().
The second thing this patch does, is to introduce
free_area_init_core_hotplug(), the memhotplug version of
free_area_init_core():
Currently, we call free_area_init_node() from the memhotplug path. In
there, we set some pgdat's fields, and call calculate_node_totalpages().
calculate_node_totalpages() calculates the # of pages the node has.
Since the node is either new, or we are re-using it, the zones belonging
to this node should not have any pages, so there is no point to calculate
this now.
Actually, we re-set these values to 0 later on with the calls to:
reset_node_managed_pages()
reset_node_present_pages()
The # of pages per node and the # of pages per zone will be calculated when
onlining the pages:
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_zone_range()
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_pgdat_range()
Also, since free_area_init_core/free_area_init_node will now only get called during early init, let us replace
__paginginit with __init, so their code gets freed up.
[osalvador@techadventures.net: fix section usage]
Link: http://lkml.kernel.org/r/20180731101752.GA473@techadventures.net
[osalvador@suse.de: v6]
Link: http://lkml.kernel.org/r/20180801122348.21588-6-osalvador@techadventures.net
Link: http://lkml.kernel.org/r/20180730101757.28058-5-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 04:53:43 +00:00
|
|
|
static unsigned long __init calc_memmap_size(unsigned long spanned_pages,
|
2018-08-22 04:53:36 +00:00
|
|
|
unsigned long present_pages)
|
2012-12-12 21:52:19 +00:00
|
|
|
{
|
|
|
|
unsigned long pages = spanned_pages;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Provide a more accurate estimation if there are holes within
|
|
|
|
* the zone and SPARSEMEM is in use. If there are holes within the
|
|
|
|
* zone, each populated memory region may cost us one or two extra
|
|
|
|
* memmap pages due to alignment because memmap pages for each
|
2017-02-27 22:29:01 +00:00
|
|
|
* populated regions may not be naturally aligned on page boundary.
|
2012-12-12 21:52:19 +00:00
|
|
|
* So the (present_pages >> 4) heuristic is a tradeoff for that.
|
|
|
|
*/
|
|
|
|
if (spanned_pages > present_pages + (present_pages >> 4) &&
|
|
|
|
IS_ENABLED(CONFIG_SPARSEMEM))
|
|
|
|
pages = present_pages;
|
|
|
|
|
|
|
|
return PAGE_ALIGN(pages * sizeof(struct page)) >> PAGE_SHIFT;
|
|
|
|
}
|
|
|
|
|
2018-08-22 04:53:29 +00:00
|
|
|
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
|
|
|
|
static void pgdat_init_split_queue(struct pglist_data *pgdat)
|
|
|
|
{
|
2019-09-23 22:38:06 +00:00
|
|
|
struct deferred_split *ds_queue = &pgdat->deferred_split_queue;
|
|
|
|
|
|
|
|
spin_lock_init(&ds_queue->split_queue_lock);
|
|
|
|
INIT_LIST_HEAD(&ds_queue->split_queue);
|
|
|
|
ds_queue->split_queue_len = 0;
|
2018-08-22 04:53:29 +00:00
|
|
|
}
|
|
|
|
#else
|
|
|
|
static void pgdat_init_split_queue(struct pglist_data *pgdat) {}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef CONFIG_COMPACTION
|
|
|
|
static void pgdat_init_kcompactd(struct pglist_data *pgdat)
|
|
|
|
{
|
|
|
|
init_waitqueue_head(&pgdat->kcompactd_wait);
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static void pgdat_init_kcompactd(struct pglist_data *pgdat) {}
|
|
|
|
#endif
|
|
|
|
|
mm/page_alloc: Introduce free_area_init_core_hotplug
Currently, whenever a new node is created/re-used from the memhotplug
path, we call free_area_init_node()->free_area_init_core(). But there is
some code that we do not really need to run when we are coming from such
path.
free_area_init_core() performs the following actions:
1) Initializes pgdat internals, such as spinlock, waitqueues and more.
2) Account # nr_all_pages and # nr_kernel_pages. These values are used later on
when creating hash tables.
3) Account number of managed_pages per zone, substracting dma_reserved and
memmap pages.
4) Initializes some fields of the zone structure data
5) Calls init_currently_empty_zone to initialize all the freelists
6) Calls memmap_init to initialize all pages belonging to certain zone
When called from memhotplug path, free_area_init_core() only performs
actions #1 and #4.
Action #2 is pointless as the zones do not have any pages since either the
node was freed, or we are re-using it, eitherway all zones belonging to
this node should have 0 pages. For the same reason, action #3 results
always in manages_pages being 0.
Action #5 and #6 are performed later on when onlining the pages:
online_pages()->move_pfn_range_to_zone()->init_currently_empty_zone()
online_pages()->move_pfn_range_to_zone()->memmap_init_zone()
This patch does two things:
First, moves the node/zone initializtion to their own function, so it
allows us to create a small version of free_area_init_core, where we only
perform:
1) Initialization of pgdat internals, such as spinlock, waitqueues and more
4) Initialization of some fields of the zone structure data
These two functions are: pgdat_init_internals() and zone_init_internals().
The second thing this patch does, is to introduce
free_area_init_core_hotplug(), the memhotplug version of
free_area_init_core():
Currently, we call free_area_init_node() from the memhotplug path. In
there, we set some pgdat's fields, and call calculate_node_totalpages().
calculate_node_totalpages() calculates the # of pages the node has.
Since the node is either new, or we are re-using it, the zones belonging
to this node should not have any pages, so there is no point to calculate
this now.
Actually, we re-set these values to 0 later on with the calls to:
reset_node_managed_pages()
reset_node_present_pages()
The # of pages per node and the # of pages per zone will be calculated when
onlining the pages:
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_zone_range()
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_pgdat_range()
Also, since free_area_init_core/free_area_init_node will now only get called during early init, let us replace
__paginginit with __init, so their code gets freed up.
[osalvador@techadventures.net: fix section usage]
Link: http://lkml.kernel.org/r/20180731101752.GA473@techadventures.net
[osalvador@suse.de: v6]
Link: http://lkml.kernel.org/r/20180801122348.21588-6-osalvador@techadventures.net
Link: http://lkml.kernel.org/r/20180730101757.28058-5-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 04:53:43 +00:00
|
|
|
static void __meminit pgdat_init_internals(struct pglist_data *pgdat)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2005-10-30 01:16:52 +00:00
|
|
|
pgdat_resize_init(pgdat);
|
2018-08-22 04:53:29 +00:00
|
|
|
|
|
|
|
pgdat_init_split_queue(pgdat);
|
|
|
|
pgdat_init_kcompactd(pgdat);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
init_waitqueue_head(&pgdat->kswapd_wait);
|
2012-07-31 23:44:35 +00:00
|
|
|
init_waitqueue_head(&pgdat->pfmemalloc_wait);
|
2018-08-22 04:53:29 +00:00
|
|
|
|
mm/page_ext: resurrect struct page extending code for debugging
When we debug something, we'd like to insert some information to every
page. For this purpose, we sometimes modify struct page itself. But,
this has drawbacks. First, it requires re-compile. This makes us
hesitate to use the powerful debug feature so development process is
slowed down. And, second, sometimes it is impossible to rebuild the
kernel due to third party module dependency. At third, system behaviour
would be largely different after re-compile, because it changes size of
struct page greatly and this structure is accessed by every part of
kernel. Keeping this as it is would be better to reproduce errornous
situation.
This feature is intended to overcome above mentioned problems. This
feature allocates memory for extended data per page in certain place
rather than the struct page itself. This memory can be accessed by the
accessor functions provided by this code. During the boot process, it
checks whether allocation of huge chunk of memory is needed or not. If
not, it avoids allocating memory at all. With this advantage, we can
include this feature into the kernel in default and can avoid rebuild and
solve related problems.
Until now, memcg uses this technique. But, now, memcg decides to embed
their variable to struct page itself and it's code to extend struct page
has been removed. I'd like to use this code to develop debug feature, so
this patch resurrect it.
To help these things to work well, this patch introduces two callbacks for
clients. One is the need callback which is mandatory if user wants to
avoid useless memory allocation at boot-time. The other is optional, init
callback, which is used to do proper initialization after memory is
allocated. Detailed explanation about purpose of these functions is in
code comment. Please refer it.
Others are completely same with previous extension code in memcg.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dave Hansen <dave@sr71.net>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Jungsoo Son <jungsoo.son@lge.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:55:46 +00:00
|
|
|
pgdat_page_ext_init(pgdat);
|
2019-12-01 01:55:34 +00:00
|
|
|
lruvec_init(&pgdat->__lruvec);
|
mm/page_alloc: Introduce free_area_init_core_hotplug
Currently, whenever a new node is created/re-used from the memhotplug
path, we call free_area_init_node()->free_area_init_core(). But there is
some code that we do not really need to run when we are coming from such
path.
free_area_init_core() performs the following actions:
1) Initializes pgdat internals, such as spinlock, waitqueues and more.
2) Account # nr_all_pages and # nr_kernel_pages. These values are used later on
when creating hash tables.
3) Account number of managed_pages per zone, substracting dma_reserved and
memmap pages.
4) Initializes some fields of the zone structure data
5) Calls init_currently_empty_zone to initialize all the freelists
6) Calls memmap_init to initialize all pages belonging to certain zone
When called from memhotplug path, free_area_init_core() only performs
actions #1 and #4.
Action #2 is pointless as the zones do not have any pages since either the
node was freed, or we are re-using it, eitherway all zones belonging to
this node should have 0 pages. For the same reason, action #3 results
always in manages_pages being 0.
Action #5 and #6 are performed later on when onlining the pages:
online_pages()->move_pfn_range_to_zone()->init_currently_empty_zone()
online_pages()->move_pfn_range_to_zone()->memmap_init_zone()
This patch does two things:
First, moves the node/zone initializtion to their own function, so it
allows us to create a small version of free_area_init_core, where we only
perform:
1) Initialization of pgdat internals, such as spinlock, waitqueues and more
4) Initialization of some fields of the zone structure data
These two functions are: pgdat_init_internals() and zone_init_internals().
The second thing this patch does, is to introduce
free_area_init_core_hotplug(), the memhotplug version of
free_area_init_core():
Currently, we call free_area_init_node() from the memhotplug path. In
there, we set some pgdat's fields, and call calculate_node_totalpages().
calculate_node_totalpages() calculates the # of pages the node has.
Since the node is either new, or we are re-using it, the zones belonging
to this node should not have any pages, so there is no point to calculate
this now.
Actually, we re-set these values to 0 later on with the calls to:
reset_node_managed_pages()
reset_node_present_pages()
The # of pages per node and the # of pages per zone will be calculated when
onlining the pages:
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_zone_range()
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_pgdat_range()
Also, since free_area_init_core/free_area_init_node will now only get called during early init, let us replace
__paginginit with __init, so their code gets freed up.
[osalvador@techadventures.net: fix section usage]
Link: http://lkml.kernel.org/r/20180731101752.GA473@techadventures.net
[osalvador@suse.de: v6]
Link: http://lkml.kernel.org/r/20180801122348.21588-6-osalvador@techadventures.net
Link: http://lkml.kernel.org/r/20180730101757.28058-5-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 04:53:43 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void __meminit zone_init_internals(struct zone *zone, enum zone_type idx, int nid,
|
|
|
|
unsigned long remaining_pages)
|
|
|
|
{
|
2018-12-28 08:34:24 +00:00
|
|
|
atomic_long_set(&zone->managed_pages, remaining_pages);
|
mm/page_alloc: Introduce free_area_init_core_hotplug
Currently, whenever a new node is created/re-used from the memhotplug
path, we call free_area_init_node()->free_area_init_core(). But there is
some code that we do not really need to run when we are coming from such
path.
free_area_init_core() performs the following actions:
1) Initializes pgdat internals, such as spinlock, waitqueues and more.
2) Account # nr_all_pages and # nr_kernel_pages. These values are used later on
when creating hash tables.
3) Account number of managed_pages per zone, substracting dma_reserved and
memmap pages.
4) Initializes some fields of the zone structure data
5) Calls init_currently_empty_zone to initialize all the freelists
6) Calls memmap_init to initialize all pages belonging to certain zone
When called from memhotplug path, free_area_init_core() only performs
actions #1 and #4.
Action #2 is pointless as the zones do not have any pages since either the
node was freed, or we are re-using it, eitherway all zones belonging to
this node should have 0 pages. For the same reason, action #3 results
always in manages_pages being 0.
Action #5 and #6 are performed later on when onlining the pages:
online_pages()->move_pfn_range_to_zone()->init_currently_empty_zone()
online_pages()->move_pfn_range_to_zone()->memmap_init_zone()
This patch does two things:
First, moves the node/zone initializtion to their own function, so it
allows us to create a small version of free_area_init_core, where we only
perform:
1) Initialization of pgdat internals, such as spinlock, waitqueues and more
4) Initialization of some fields of the zone structure data
These two functions are: pgdat_init_internals() and zone_init_internals().
The second thing this patch does, is to introduce
free_area_init_core_hotplug(), the memhotplug version of
free_area_init_core():
Currently, we call free_area_init_node() from the memhotplug path. In
there, we set some pgdat's fields, and call calculate_node_totalpages().
calculate_node_totalpages() calculates the # of pages the node has.
Since the node is either new, or we are re-using it, the zones belonging
to this node should not have any pages, so there is no point to calculate
this now.
Actually, we re-set these values to 0 later on with the calls to:
reset_node_managed_pages()
reset_node_present_pages()
The # of pages per node and the # of pages per zone will be calculated when
onlining the pages:
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_zone_range()
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_pgdat_range()
Also, since free_area_init_core/free_area_init_node will now only get called during early init, let us replace
__paginginit with __init, so their code gets freed up.
[osalvador@techadventures.net: fix section usage]
Link: http://lkml.kernel.org/r/20180731101752.GA473@techadventures.net
[osalvador@suse.de: v6]
Link: http://lkml.kernel.org/r/20180801122348.21588-6-osalvador@techadventures.net
Link: http://lkml.kernel.org/r/20180730101757.28058-5-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 04:53:43 +00:00
|
|
|
zone_set_nid(zone, nid);
|
|
|
|
zone->name = zone_names[idx];
|
|
|
|
zone->zone_pgdat = NODE_DATA(nid);
|
|
|
|
spin_lock_init(&zone->lock);
|
|
|
|
zone_seqlock_init(zone);
|
|
|
|
zone_pcp_init(zone);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set up the zone data structures
|
|
|
|
* - init pgdat internals
|
|
|
|
* - init all zones belonging to this node
|
|
|
|
*
|
|
|
|
* NOTE: this function is only called during memory hotplug
|
|
|
|
*/
|
|
|
|
#ifdef CONFIG_MEMORY_HOTPLUG
|
|
|
|
void __ref free_area_init_core_hotplug(int nid)
|
|
|
|
{
|
|
|
|
enum zone_type z;
|
|
|
|
pg_data_t *pgdat = NODE_DATA(nid);
|
|
|
|
|
|
|
|
pgdat_init_internals(pgdat);
|
|
|
|
for (z = 0; z < MAX_NR_ZONES; z++)
|
|
|
|
zone_init_internals(&pgdat->node_zones[z], z, nid, 0);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set up the zone data structures:
|
|
|
|
* - mark all pages reserved
|
|
|
|
* - mark all memory queues empty
|
|
|
|
* - clear the memory bitmaps
|
|
|
|
*
|
|
|
|
* NOTE: pgdat should get zeroed by caller.
|
|
|
|
* NOTE: this function is only called during early init.
|
|
|
|
*/
|
|
|
|
static void __init free_area_init_core(struct pglist_data *pgdat)
|
|
|
|
{
|
|
|
|
enum zone_type j;
|
|
|
|
int nid = pgdat->node_id;
|
2012-01-11 14:16:11 +00:00
|
|
|
|
mm/page_alloc: Introduce free_area_init_core_hotplug
Currently, whenever a new node is created/re-used from the memhotplug
path, we call free_area_init_node()->free_area_init_core(). But there is
some code that we do not really need to run when we are coming from such
path.
free_area_init_core() performs the following actions:
1) Initializes pgdat internals, such as spinlock, waitqueues and more.
2) Account # nr_all_pages and # nr_kernel_pages. These values are used later on
when creating hash tables.
3) Account number of managed_pages per zone, substracting dma_reserved and
memmap pages.
4) Initializes some fields of the zone structure data
5) Calls init_currently_empty_zone to initialize all the freelists
6) Calls memmap_init to initialize all pages belonging to certain zone
When called from memhotplug path, free_area_init_core() only performs
actions #1 and #4.
Action #2 is pointless as the zones do not have any pages since either the
node was freed, or we are re-using it, eitherway all zones belonging to
this node should have 0 pages. For the same reason, action #3 results
always in manages_pages being 0.
Action #5 and #6 are performed later on when onlining the pages:
online_pages()->move_pfn_range_to_zone()->init_currently_empty_zone()
online_pages()->move_pfn_range_to_zone()->memmap_init_zone()
This patch does two things:
First, moves the node/zone initializtion to their own function, so it
allows us to create a small version of free_area_init_core, where we only
perform:
1) Initialization of pgdat internals, such as spinlock, waitqueues and more
4) Initialization of some fields of the zone structure data
These two functions are: pgdat_init_internals() and zone_init_internals().
The second thing this patch does, is to introduce
free_area_init_core_hotplug(), the memhotplug version of
free_area_init_core():
Currently, we call free_area_init_node() from the memhotplug path. In
there, we set some pgdat's fields, and call calculate_node_totalpages().
calculate_node_totalpages() calculates the # of pages the node has.
Since the node is either new, or we are re-using it, the zones belonging
to this node should not have any pages, so there is no point to calculate
this now.
Actually, we re-set these values to 0 later on with the calls to:
reset_node_managed_pages()
reset_node_present_pages()
The # of pages per node and the # of pages per zone will be calculated when
onlining the pages:
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_zone_range()
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_pgdat_range()
Also, since free_area_init_core/free_area_init_node will now only get called during early init, let us replace
__paginginit with __init, so their code gets freed up.
[osalvador@techadventures.net: fix section usage]
Link: http://lkml.kernel.org/r/20180731101752.GA473@techadventures.net
[osalvador@suse.de: v6]
Link: http://lkml.kernel.org/r/20180801122348.21588-6-osalvador@techadventures.net
Link: http://lkml.kernel.org/r/20180730101757.28058-5-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 04:53:43 +00:00
|
|
|
pgdat_init_internals(pgdat);
|
mm: vmstat: move slab statistics from zone to node counters
Patch series "mm: per-lruvec slab stats"
Josef is working on a new approach to balancing slab caches and the page
cache. For this to work, he needs slab cache statistics on the lruvec
level. These patches implement that by adding infrastructure that
allows updating and reading generic VM stat items per lruvec, then
switches some existing VM accounting sites, including the slab
accounting ones, to this new cgroup-aware API.
I'll follow up with more patches on this, because there is actually
substantial simplification that can be done to the memory controller
when we replace private memcg accounting with making the existing VM
accounting sites cgroup-aware. But this is enough for Josef to base his
slab reclaim work on, so here goes.
This patch (of 5):
To re-implement slab cache vs. page cache balancing, we'll need the
slab counters at the lruvec level, which, ever since lru reclaim was
moved from the zone to the node, is the intersection of the node, not
the zone, and the memcg.
We could retain the per-zone counters for when the page allocator dumps
its memory information on failures, and have counters on both levels -
which on all but NUMA node 0 is usually redundant. But let's keep it
simple for now and just move them. If anybody complains we can restore
the per-zone counters.
[hannes@cmpxchg.org: fix oops]
Link: http://lkml.kernel.org/r/20170605183511.GA8915@cmpxchg.org
Link: http://lkml.kernel.org/r/20170530181724.27197-3-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-06 22:40:43 +00:00
|
|
|
pgdat->per_cpu_nodestats = &boot_nodestats;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
for (j = 0; j < MAX_NR_ZONES; j++) {
|
|
|
|
struct zone *zone = pgdat->node_zones + j;
|
2018-06-08 00:06:04 +00:00
|
|
|
unsigned long size, freesize, memmap_pages;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2015-06-24 23:57:02 +00:00
|
|
|
size = zone->spanned_pages;
|
2018-06-08 00:06:04 +00:00
|
|
|
freesize = zone->present_pages;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-09-27 08:49:56 +00:00
|
|
|
/*
|
2012-12-12 21:52:12 +00:00
|
|
|
* Adjust freesize so that it accounts for how much memory
|
2006-09-27 08:49:56 +00:00
|
|
|
* is used by this zone for memmap. This affects the watermark
|
|
|
|
* and per-cpu initialisations
|
|
|
|
*/
|
2018-06-08 00:06:04 +00:00
|
|
|
memmap_pages = calc_memmap_size(size, freesize);
|
2014-12-13 00:56:21 +00:00
|
|
|
if (!is_highmem_idx(j)) {
|
|
|
|
if (freesize >= memmap_pages) {
|
|
|
|
freesize -= memmap_pages;
|
|
|
|
if (memmap_pages)
|
|
|
|
printk(KERN_DEBUG
|
|
|
|
" %s zone: %lu pages used for memmap\n",
|
|
|
|
zone_names[j], memmap_pages);
|
|
|
|
} else
|
2016-03-17 21:19:50 +00:00
|
|
|
pr_warn(" %s zone: %lu pages exceeds freesize %lu\n",
|
2014-12-13 00:56:21 +00:00
|
|
|
zone_names[j], memmap_pages, freesize);
|
|
|
|
}
|
2006-09-27 08:49:56 +00:00
|
|
|
|
2007-02-10 09:43:07 +00:00
|
|
|
/* Account for reserved pages */
|
2012-12-12 21:52:12 +00:00
|
|
|
if (j == 0 && freesize > dma_reserve) {
|
|
|
|
freesize -= dma_reserve;
|
2008-10-19 03:27:06 +00:00
|
|
|
printk(KERN_DEBUG " %s zone: %lu pages reserved\n",
|
2007-02-10 09:43:07 +00:00
|
|
|
zone_names[0], dma_reserve);
|
2006-09-27 08:49:56 +00:00
|
|
|
}
|
|
|
|
|
2006-09-26 06:31:12 +00:00
|
|
|
if (!is_highmem_idx(j))
|
2012-12-12 21:52:12 +00:00
|
|
|
nr_kernel_pages += freesize;
|
2012-12-12 21:52:19 +00:00
|
|
|
/* Charge for highmem memmap if there are enough kernel pages */
|
|
|
|
else if (nr_kernel_pages > memmap_pages * 2)
|
|
|
|
nr_kernel_pages -= memmap_pages;
|
2012-12-12 21:52:12 +00:00
|
|
|
nr_all_pages += freesize;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2012-12-12 21:52:12 +00:00
|
|
|
/*
|
|
|
|
* Set an approximate value for lowmem here, it will be adjusted
|
|
|
|
* when the bootmem allocator frees pages into the buddy system.
|
|
|
|
* And all highmem pages will be managed by the buddy system.
|
|
|
|
*/
|
mm/page_alloc: Introduce free_area_init_core_hotplug
Currently, whenever a new node is created/re-used from the memhotplug
path, we call free_area_init_node()->free_area_init_core(). But there is
some code that we do not really need to run when we are coming from such
path.
free_area_init_core() performs the following actions:
1) Initializes pgdat internals, such as spinlock, waitqueues and more.
2) Account # nr_all_pages and # nr_kernel_pages. These values are used later on
when creating hash tables.
3) Account number of managed_pages per zone, substracting dma_reserved and
memmap pages.
4) Initializes some fields of the zone structure data
5) Calls init_currently_empty_zone to initialize all the freelists
6) Calls memmap_init to initialize all pages belonging to certain zone
When called from memhotplug path, free_area_init_core() only performs
actions #1 and #4.
Action #2 is pointless as the zones do not have any pages since either the
node was freed, or we are re-using it, eitherway all zones belonging to
this node should have 0 pages. For the same reason, action #3 results
always in manages_pages being 0.
Action #5 and #6 are performed later on when onlining the pages:
online_pages()->move_pfn_range_to_zone()->init_currently_empty_zone()
online_pages()->move_pfn_range_to_zone()->memmap_init_zone()
This patch does two things:
First, moves the node/zone initializtion to their own function, so it
allows us to create a small version of free_area_init_core, where we only
perform:
1) Initialization of pgdat internals, such as spinlock, waitqueues and more
4) Initialization of some fields of the zone structure data
These two functions are: pgdat_init_internals() and zone_init_internals().
The second thing this patch does, is to introduce
free_area_init_core_hotplug(), the memhotplug version of
free_area_init_core():
Currently, we call free_area_init_node() from the memhotplug path. In
there, we set some pgdat's fields, and call calculate_node_totalpages().
calculate_node_totalpages() calculates the # of pages the node has.
Since the node is either new, or we are re-using it, the zones belonging
to this node should not have any pages, so there is no point to calculate
this now.
Actually, we re-set these values to 0 later on with the calls to:
reset_node_managed_pages()
reset_node_present_pages()
The # of pages per node and the # of pages per zone will be calculated when
onlining the pages:
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_zone_range()
online_pages()->move_pfn_range()->move_pfn_range_to_zone()->resize_pgdat_range()
Also, since free_area_init_core/free_area_init_node will now only get called during early init, let us replace
__paginginit with __init, so their code gets freed up.
[osalvador@techadventures.net: fix section usage]
Link: http://lkml.kernel.org/r/20180731101752.GA473@techadventures.net
[osalvador@suse.de: v6]
Link: http://lkml.kernel.org/r/20180801122348.21588-6-osalvador@techadventures.net
Link: http://lkml.kernel.org/r/20180730101757.28058-5-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Pasha Tatashin <Pavel.Tatashin@microsoft.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-22 04:53:43 +00:00
|
|
|
zone_init_internals(zone, j, nid, freesize);
|
mm: page_alloc: fair zone allocator policy
Each zone that holds userspace pages of one workload must be aged at a
speed proportional to the zone size. Otherwise, the time an individual
page gets to stay in memory depends on the zone it happened to be
allocated in. Asymmetry in the zone aging creates rather unpredictable
aging behavior and results in the wrong pages being reclaimed, activated
etc.
But exactly this happens right now because of the way the page allocator
and kswapd interact. The page allocator uses per-node lists of all zones
in the system, ordered by preference, when allocating a new page. When
the first iteration does not yield any results, kswapd is woken up and the
allocator retries. Due to the way kswapd reclaims zones below the high
watermark while a zone can be allocated from when it is above the low
watermark, the allocator may keep kswapd running while kswapd reclaim
ensures that the page allocator can keep allocating from the first zone in
the zonelist for extended periods of time. Meanwhile the other zones
rarely see new allocations and thus get aged much slower in comparison.
The result is that the occasional page placed in lower zones gets
relatively more time in memory, even gets promoted to the active list
after its peers have long been evicted. Meanwhile, the bulk of the
working set may be thrashing on the preferred zone even though there may
be significant amounts of memory available in the lower zones.
Even the most basic test -- repeatedly reading a file slightly bigger than
memory -- shows how broken the zone aging is. In this scenario, no single
page should be able stay in memory long enough to get referenced twice and
activated, but activation happens in spades:
$ grep active_file /proc/zoneinfo
nr_inactive_file 0
nr_active_file 0
nr_inactive_file 0
nr_active_file 8
nr_inactive_file 1582
nr_active_file 11994
$ cat data data data data >/dev/null
$ grep active_file /proc/zoneinfo
nr_inactive_file 0
nr_active_file 70
nr_inactive_file 258753
nr_active_file 443214
nr_inactive_file 149793
nr_active_file 12021
Fix this with a very simple round robin allocator. Each zone is allowed a
batch of allocations that is proportional to the zone's size, after which
it is treated as full. The batch counters are reset when all zones have
been tried and the allocator enters the slowpath and kicks off kswapd
reclaim. Allocation and reclaim is now fairly spread out to all
available/allowable zones:
$ grep active_file /proc/zoneinfo
nr_inactive_file 0
nr_active_file 0
nr_inactive_file 174
nr_active_file 4865
nr_inactive_file 53
nr_active_file 860
$ cat data data data data >/dev/null
$ grep active_file /proc/zoneinfo
nr_inactive_file 0
nr_active_file 0
nr_inactive_file 666622
nr_active_file 4988
nr_inactive_file 190969
nr_active_file 937
When zone_reclaim_mode is enabled, allocations will now spread out to all
zones on the local node, not just the first preferred zone (which on a 4G
node might be a tiny Normal zone).
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Paul Bolle <paul.bollee@gmail.com>
Cc: Zlatko Calusic <zcalusic@bitsync.net>
Tested-by: Kevin Hilman <khilman@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-11 21:20:47 +00:00
|
|
|
|
2018-05-23 01:18:21 +00:00
|
|
|
if (!size)
|
2005-04-16 22:20:36 +00:00
|
|
|
continue;
|
|
|
|
|
2012-05-29 22:06:31 +00:00
|
|
|
set_pageblock_order();
|
2021-02-24 20:06:20 +00:00
|
|
|
setup_usemap(zone);
|
2021-02-24 20:06:24 +00:00
|
|
|
init_currently_empty_zone(zone, zone->zone_start_pfn, size);
|
2021-02-24 20:06:17 +00:00
|
|
|
memmap_init_zone(zone);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-11-16 01:39:18 +00:00
|
|
|
#ifdef CONFIG_FLAT_NODE_MEM_MAP
|
2016-08-02 21:03:33 +00:00
|
|
|
static void __ref alloc_node_mem_map(struct pglist_data *pgdat)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2015-11-10 18:09:47 +00:00
|
|
|
unsigned long __maybe_unused start = 0;
|
2015-11-06 02:48:46 +00:00
|
|
|
unsigned long __maybe_unused offset = 0;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* Skip empty nodes */
|
|
|
|
if (!pgdat->node_spanned_pages)
|
|
|
|
return;
|
|
|
|
|
2015-11-10 18:09:47 +00:00
|
|
|
start = pgdat->node_start_pfn & ~(MAX_ORDER_NR_PAGES - 1);
|
|
|
|
offset = pgdat->node_start_pfn - start;
|
2005-04-16 22:20:36 +00:00
|
|
|
/* ia64 gets its own node_mem_map, before this, without bootmem */
|
|
|
|
if (!pgdat->node_mem_map) {
|
2015-11-10 18:09:47 +00:00
|
|
|
unsigned long size, end;
|
[PATCH] sparsemem memory model
Sparsemem abstracts the use of discontiguous mem_maps[]. This kind of
mem_map[] is needed by discontiguous memory machines (like in the old
CONFIG_DISCONTIGMEM case) as well as memory hotplug systems. Sparsemem
replaces DISCONTIGMEM when enabled, and it is hoped that it can eventually
become a complete replacement.
A significant advantage over DISCONTIGMEM is that it's completely separated
from CONFIG_NUMA. When producing this patch, it became apparent in that NUMA
and DISCONTIG are often confused.
Another advantage is that sparse doesn't require each NUMA node's ranges to be
contiguous. It can handle overlapping ranges between nodes with no problems,
where DISCONTIGMEM currently throws away that memory.
Sparsemem uses an array to provide different pfn_to_page() translations for
each SECTION_SIZE area of physical memory. This is what allows the mem_map[]
to be chopped up.
In order to do quick pfn_to_page() operations, the section number of the page
is encoded in page->flags. Part of the sparsemem infrastructure enables
sharing of these bits more dynamically (at compile-time) between the
page_zone() and sparsemem operations. However, on 32-bit architectures, the
number of bits is quite limited, and may require growing the size of the
page->flags type in certain conditions. Several things might force this to
occur: a decrease in the SECTION_SIZE (if you want to hotplug smaller areas of
memory), an increase in the physical address space, or an increase in the
number of used page->flags.
One thing to note is that, once sparsemem is present, the NUMA node
information no longer needs to be stored in the page->flags. It might provide
speed increases on certain platforms and will be stored there if there is
room. But, if out of room, an alternate (theoretically slower) mechanism is
used.
This patch introduces CONFIG_FLATMEM. It is used in almost all cases where
there used to be an #ifndef DISCONTIG, because SPARSEMEM and DISCONTIGMEM
often have to compile out the same areas of code.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Martin Bligh <mbligh@aracnet.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:07:54 +00:00
|
|
|
struct page *map;
|
|
|
|
|
2006-05-20 22:00:31 +00:00
|
|
|
/*
|
|
|
|
* The zone's endpoints aren't required to be MAX_ORDER
|
|
|
|
* aligned but the node_mem_map endpoints must be in order
|
|
|
|
* for the buddy allocator to function correctly.
|
|
|
|
*/
|
2013-02-23 00:35:23 +00:00
|
|
|
end = pgdat_end_pfn(pgdat);
|
2006-05-20 22:00:31 +00:00
|
|
|
end = ALIGN(end, MAX_ORDER_NR_PAGES);
|
|
|
|
size = (end - start) * sizeof(struct page);
|
2019-03-12 06:30:42 +00:00
|
|
|
map = memblock_alloc_node(size, SMP_CACHE_BYTES,
|
|
|
|
pgdat->node_id);
|
2019-03-05 23:46:43 +00:00
|
|
|
if (!map)
|
|
|
|
panic("Failed to allocate %ld bytes for node %d memory map\n",
|
|
|
|
size, pgdat->node_id);
|
2015-11-06 02:48:46 +00:00
|
|
|
pgdat->node_mem_map = map + offset;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2017-11-16 01:39:18 +00:00
|
|
|
pr_debug("%s: node %d, pgdat %08lx, node_mem_map %08lx\n",
|
|
|
|
__func__, pgdat->node_id, (unsigned long)pgdat,
|
|
|
|
(unsigned long)pgdat->node_mem_map);
|
2007-05-31 07:40:54 +00:00
|
|
|
#ifndef CONFIG_NEED_MULTIPLE_NODES
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* With no DISCONTIG, the global mem_map is just set as node 0's
|
|
|
|
*/
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
if (pgdat == NODE_DATA(0)) {
|
2005-04-16 22:20:36 +00:00
|
|
|
mem_map = NODE_DATA(0)->node_mem_map;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
if (page_to_pfn(mem_map) != pgdat->node_start_pfn)
|
2015-11-06 02:48:46 +00:00
|
|
|
mem_map -= offset;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
#endif
|
|
|
|
}
|
2017-11-16 01:39:18 +00:00
|
|
|
#else
|
|
|
|
static void __ref alloc_node_mem_map(struct pglist_data *pgdat) { }
|
|
|
|
#endif /* CONFIG_FLAT_NODE_MEM_MAP */
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2018-08-22 04:53:39 +00:00
|
|
|
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
|
|
|
|
static inline void pgdat_set_deferred_range(pg_data_t *pgdat)
|
|
|
|
{
|
|
|
|
pgdat->first_deferred_pfn = ULONG_MAX;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static inline void pgdat_set_deferred_range(pg_data_t *pgdat) {}
|
|
|
|
#endif
|
|
|
|
|
2020-06-03 22:58:13 +00:00
|
|
|
static void __init free_area_init_node(int nid)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2008-07-24 04:27:20 +00:00
|
|
|
pg_data_t *pgdat = NODE_DATA(nid);
|
2013-07-08 22:59:52 +00:00
|
|
|
unsigned long start_pfn = 0;
|
|
|
|
unsigned long end_pfn = 0;
|
2008-07-24 04:27:20 +00:00
|
|
|
|
2012-07-31 23:46:14 +00:00
|
|
|
/* pg_data_t should be reset to zero when it's allocated */
|
2020-06-03 22:59:01 +00:00
|
|
|
WARN_ON(pgdat->nr_zones || pgdat->kswapd_highest_zoneidx);
|
2012-07-31 23:46:14 +00:00
|
|
|
|
2020-06-03 22:58:13 +00:00
|
|
|
get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
|
2012-07-31 23:46:14 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
pgdat->node_id = nid;
|
2020-06-03 22:58:13 +00:00
|
|
|
pgdat->node_start_pfn = start_pfn;
|
mm, vmstat: add infrastructure for per-node vmstats
Patchset: "Move LRU page reclaim from zones to nodes v9"
This series moves LRUs from the zones to the node. While this is a
current rebase, the test results were based on mmotm as of June 23rd.
Conceptually, this series is simple but there are a lot of details.
Some of the broad motivations for this are;
1. The residency of a page partially depends on what zone the page was
allocated from. This is partially combatted by the fair zone allocation
policy but that is a partial solution that introduces overhead in the
page allocator paths.
2. Currently, reclaim on node 0 behaves slightly different to node 1. For
example, direct reclaim scans in zonelist order and reclaims even if
the zone is over the high watermark regardless of the age of pages
in that LRU. Kswapd on the other hand starts reclaim on the highest
unbalanced zone. A difference in distribution of file/anon pages due
to when they were allocated results can result in a difference in
again. While the fair zone allocation policy mitigates some of the
problems here, the page reclaim results on a multi-zone node will
always be different to a single-zone node.
it was scheduled on as a result.
3. kswapd and the page allocator scan zones in the opposite order to
avoid interfering with each other but it's sensitive to timing. This
mitigates the page allocator using pages that were allocated very recently
in the ideal case but it's sensitive to timing. When kswapd is allocating
from lower zones then it's great but during the rebalancing of the highest
zone, the page allocator and kswapd interfere with each other. It's worse
if the highest zone is small and difficult to balance.
4. slab shrinkers are node-based which makes it harder to identify the exact
relationship between slab reclaim and LRU reclaim.
The reason we have zone-based reclaim is that we used to have
large highmem zones in common configurations and it was necessary
to quickly find ZONE_NORMAL pages for reclaim. Today, this is much
less of a concern as machines with lots of memory will (or should) use
64-bit kernels. Combinations of 32-bit hardware and 64-bit hardware are
rare. Machines that do use highmem should have relatively low highmem:lowmem
ratios than we worried about in the past.
Conceptually, moving to node LRUs should be easier to understand. The
page allocator plays fewer tricks to game reclaim and reclaim behaves
similarly on all nodes.
The series has been tested on a 16 core UMA machine and a 2-socket 48
core NUMA machine. The UMA results are presented in most cases as the NUMA
machine behaved similarly.
pagealloc
---------
This is a microbenchmark that shows the benefit of removing the fair zone
allocation policy. It was tested uip to order-4 but only orders 0 and 1 are
shown as the other orders were comparable.
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v9
Min total-odr0-1 490.00 ( 0.00%) 457.00 ( 6.73%)
Min total-odr0-2 347.00 ( 0.00%) 329.00 ( 5.19%)
Min total-odr0-4 288.00 ( 0.00%) 273.00 ( 5.21%)
Min total-odr0-8 251.00 ( 0.00%) 239.00 ( 4.78%)
Min total-odr0-16 234.00 ( 0.00%) 222.00 ( 5.13%)
Min total-odr0-32 223.00 ( 0.00%) 211.00 ( 5.38%)
Min total-odr0-64 217.00 ( 0.00%) 208.00 ( 4.15%)
Min total-odr0-128 214.00 ( 0.00%) 204.00 ( 4.67%)
Min total-odr0-256 250.00 ( 0.00%) 230.00 ( 8.00%)
Min total-odr0-512 271.00 ( 0.00%) 269.00 ( 0.74%)
Min total-odr0-1024 291.00 ( 0.00%) 282.00 ( 3.09%)
Min total-odr0-2048 303.00 ( 0.00%) 296.00 ( 2.31%)
Min total-odr0-4096 311.00 ( 0.00%) 309.00 ( 0.64%)
Min total-odr0-8192 316.00 ( 0.00%) 314.00 ( 0.63%)
Min total-odr0-16384 317.00 ( 0.00%) 315.00 ( 0.63%)
Min total-odr1-1 742.00 ( 0.00%) 712.00 ( 4.04%)
Min total-odr1-2 562.00 ( 0.00%) 530.00 ( 5.69%)
Min total-odr1-4 457.00 ( 0.00%) 433.00 ( 5.25%)
Min total-odr1-8 411.00 ( 0.00%) 381.00 ( 7.30%)
Min total-odr1-16 381.00 ( 0.00%) 356.00 ( 6.56%)
Min total-odr1-32 372.00 ( 0.00%) 346.00 ( 6.99%)
Min total-odr1-64 372.00 ( 0.00%) 343.00 ( 7.80%)
Min total-odr1-128 375.00 ( 0.00%) 351.00 ( 6.40%)
Min total-odr1-256 379.00 ( 0.00%) 351.00 ( 7.39%)
Min total-odr1-512 385.00 ( 0.00%) 355.00 ( 7.79%)
Min total-odr1-1024 386.00 ( 0.00%) 358.00 ( 7.25%)
Min total-odr1-2048 390.00 ( 0.00%) 362.00 ( 7.18%)
Min total-odr1-4096 390.00 ( 0.00%) 362.00 ( 7.18%)
Min total-odr1-8192 388.00 ( 0.00%) 363.00 ( 6.44%)
This shows a steady improvement throughout. The primary benefit is from
reduced system CPU usage which is obvious from the overall times;
4.7.0-rc4 4.7.0-rc4
mmotm-20160623nodelru-v8
User 189.19 191.80
System 2604.45 2533.56
Elapsed 2855.30 2786.39
The vmstats also showed that the fair zone allocation policy was definitely
removed as can be seen here;
4.7.0-rc3 4.7.0-rc3
mmotm-20160623 nodelru-v8
DMA32 allocs 28794729769 0
Normal allocs 48432501431 77227309877
Movable allocs 0 0
tiobench on ext4
----------------
tiobench is a benchmark that artifically benefits if old pages remain resident
while new pages get reclaimed. The fair zone allocation policy mitigates this
problem so pages age fairly. While the benchmark has problems, it is important
that tiobench performance remains constant as it implies that page aging
problems that the fair zone allocation policy fixes are not re-introduced.
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v9
Min PotentialReadSpeed 89.65 ( 0.00%) 90.21 ( 0.62%)
Min SeqRead-MB/sec-1 82.68 ( 0.00%) 82.01 ( -0.81%)
Min SeqRead-MB/sec-2 72.76 ( 0.00%) 72.07 ( -0.95%)
Min SeqRead-MB/sec-4 75.13 ( 0.00%) 74.92 ( -0.28%)
Min SeqRead-MB/sec-8 64.91 ( 0.00%) 65.19 ( 0.43%)
Min SeqRead-MB/sec-16 62.24 ( 0.00%) 62.22 ( -0.03%)
Min RandRead-MB/sec-1 0.88 ( 0.00%) 0.88 ( 0.00%)
Min RandRead-MB/sec-2 0.95 ( 0.00%) 0.92 ( -3.16%)
Min RandRead-MB/sec-4 1.43 ( 0.00%) 1.34 ( -6.29%)
Min RandRead-MB/sec-8 1.61 ( 0.00%) 1.60 ( -0.62%)
Min RandRead-MB/sec-16 1.80 ( 0.00%) 1.90 ( 5.56%)
Min SeqWrite-MB/sec-1 76.41 ( 0.00%) 76.85 ( 0.58%)
Min SeqWrite-MB/sec-2 74.11 ( 0.00%) 73.54 ( -0.77%)
Min SeqWrite-MB/sec-4 80.05 ( 0.00%) 80.13 ( 0.10%)
Min SeqWrite-MB/sec-8 72.88 ( 0.00%) 73.20 ( 0.44%)
Min SeqWrite-MB/sec-16 75.91 ( 0.00%) 76.44 ( 0.70%)
Min RandWrite-MB/sec-1 1.18 ( 0.00%) 1.14 ( -3.39%)
Min RandWrite-MB/sec-2 1.02 ( 0.00%) 1.03 ( 0.98%)
Min RandWrite-MB/sec-4 1.05 ( 0.00%) 0.98 ( -6.67%)
Min RandWrite-MB/sec-8 0.89 ( 0.00%) 0.92 ( 3.37%)
Min RandWrite-MB/sec-16 0.92 ( 0.00%) 0.93 ( 1.09%)
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 approx-v9
User 645.72 525.90
System 403.85 331.75
Elapsed 6795.36 6783.67
This shows that the series has little or not impact on tiobench which is
desirable and a reduction in system CPU usage. It indicates that the fair
zone allocation policy was removed in a manner that didn't reintroduce
one class of page aging bug. There were only minor differences in overall
reclaim activity
4.7.0-rc4 4.7.0-rc4
mmotm-20160623nodelru-v8
Minor Faults 645838 647465
Major Faults 573 640
Swap Ins 0 0
Swap Outs 0 0
DMA allocs 0 0
DMA32 allocs 46041453 44190646
Normal allocs 78053072 79887245
Movable allocs 0 0
Allocation stalls 24 67
Stall zone DMA 0 0
Stall zone DMA32 0 0
Stall zone Normal 0 2
Stall zone HighMem 0 0
Stall zone Movable 0 65
Direct pages scanned 10969 30609
Kswapd pages scanned 93375144 93492094
Kswapd pages reclaimed 93372243 93489370
Direct pages reclaimed 10969 30609
Kswapd efficiency 99% 99%
Kswapd velocity 13741.015 13781.934
Direct efficiency 100% 100%
Direct velocity 1.614 4.512
Percentage direct scans 0% 0%
kswapd activity was roughly comparable. There were differences in direct
reclaim activity but negligible in the context of the overall workload
(velocity of 4 pages per second with the patches applied, 1.6 pages per
second in the baseline kernel).
pgbench read-only large configuration on ext4
---------------------------------------------
pgbench is a database benchmark that can be sensitive to page reclaim
decisions. This also checks if removing the fair zone allocation policy
is safe
pgbench Transactions
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v8
Hmean 1 188.26 ( 0.00%) 189.78 ( 0.81%)
Hmean 5 330.66 ( 0.00%) 328.69 ( -0.59%)
Hmean 12 370.32 ( 0.00%) 380.72 ( 2.81%)
Hmean 21 368.89 ( 0.00%) 369.00 ( 0.03%)
Hmean 30 382.14 ( 0.00%) 360.89 ( -5.56%)
Hmean 32 428.87 ( 0.00%) 432.96 ( 0.95%)
Negligible differences again. As with tiobench, overall reclaim activity
was comparable.
bonnie++ on ext4
----------------
No interesting performance difference, negligible differences on reclaim
stats.
paralleldd on ext4
------------------
This workload uses varying numbers of dd instances to read large amounts of
data from disk.
4.7.0-rc3 4.7.0-rc3
mmotm-20160623 nodelru-v9
Amean Elapsd-1 186.04 ( 0.00%) 189.41 ( -1.82%)
Amean Elapsd-3 192.27 ( 0.00%) 191.38 ( 0.46%)
Amean Elapsd-5 185.21 ( 0.00%) 182.75 ( 1.33%)
Amean Elapsd-7 183.71 ( 0.00%) 182.11 ( 0.87%)
Amean Elapsd-12 180.96 ( 0.00%) 181.58 ( -0.35%)
Amean Elapsd-16 181.36 ( 0.00%) 183.72 ( -1.30%)
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v9
User 1548.01 1552.44
System 8609.71 8515.08
Elapsed 3587.10 3594.54
There is little or no change in performance but some drop in system CPU usage.
4.7.0-rc3 4.7.0-rc3
mmotm-20160623 nodelru-v9
Minor Faults 362662 367360
Major Faults 1204 1143
Swap Ins 22 0
Swap Outs 2855 1029
DMA allocs 0 0
DMA32 allocs 31409797 28837521
Normal allocs 46611853 49231282
Movable allocs 0 0
Direct pages scanned 0 0
Kswapd pages scanned 40845270 40869088
Kswapd pages reclaimed 40830976 40855294
Direct pages reclaimed 0 0
Kswapd efficiency 99% 99%
Kswapd velocity 11386.711 11369.769
Direct efficiency 100% 100%
Direct velocity 0.000 0.000
Percentage direct scans 0% 0%
Page writes by reclaim 2855 1029
Page writes file 0 0
Page writes anon 2855 1029
Page reclaim immediate 771 1628
Sector Reads 293312636 293536360
Sector Writes 18213568 18186480
Page rescued immediate 0 0
Slabs scanned 128257 132747
Direct inode steals 181 56
Kswapd inode steals 59 1131
It basically shows that kswapd was active at roughly the same rate in
both kernels. There was also comparable slab scanning activity and direct
reclaim was avoided in both cases. There appears to be a large difference
in numbers of inodes reclaimed but the workload has few active inodes and
is likely a timing artifact.
stutter
-------
stutter simulates a simple workload. One part uses a lot of anonymous
memory, a second measures mmap latency and a third copies a large file.
The primary metric is checking for mmap latency.
stutter
4.7.0-rc4 4.7.0-rc4
mmotm-20160623 nodelru-v8
Min mmap 16.6283 ( 0.00%) 13.4258 ( 19.26%)
1st-qrtle mmap 54.7570 ( 0.00%) 34.9121 ( 36.24%)
2nd-qrtle mmap 57.3163 ( 0.00%) 46.1147 ( 19.54%)
3rd-qrtle mmap 58.9976 ( 0.00%) 47.1882 ( 20.02%)
Max-90% mmap 59.7433 ( 0.00%) 47.4453 ( 20.58%)
Max-93% mmap 60.1298 ( 0.00%) 47.6037 ( 20.83%)
Max-95% mmap 73.4112 ( 0.00%) 82.8719 (-12.89%)
Max-99% mmap 92.8542 ( 0.00%) 88.8870 ( 4.27%)
Max mmap 1440.6569 ( 0.00%) 121.4201 ( 91.57%)
Mean mmap 59.3493 ( 0.00%) 42.2991 ( 28.73%)
Best99%Mean mmap 57.2121 ( 0.00%) 41.8207 ( 26.90%)
Best95%Mean mmap 55.9113 ( 0.00%) 39.9620 ( 28.53%)
Best90%Mean mmap 55.6199 ( 0.00%) 39.3124 ( 29.32%)
Best50%Mean mmap 53.2183 ( 0.00%) 33.1307 ( 37.75%)
Best10%Mean mmap 45.9842 ( 0.00%) 20.4040 ( 55.63%)
Best5%Mean mmap 43.2256 ( 0.00%) 17.9654 ( 58.44%)
Best1%Mean mmap 32.9388 ( 0.00%) 16.6875 ( 49.34%)
This shows a number of improvements with the worst-case outlier greatly
improved.
Some of the vmstats are interesting
4.7.0-rc4 4.7.0-rc4
mmotm-20160623nodelru-v8
Swap Ins 163 502
Swap Outs 0 0
DMA allocs 0 0
DMA32 allocs 618719206 1381662383
Normal allocs 891235743 564138421
Movable allocs 0 0
Allocation stalls 2603 1
Direct pages scanned 216787 2
Kswapd pages scanned 50719775 41778378
Kswapd pages reclaimed 41541765 41777639
Direct pages reclaimed 209159 0
Kswapd efficiency 81% 99%
Kswapd velocity 16859.554 14329.059
Direct efficiency 96% 0%
Direct velocity 72.061 0.001
Percentage direct scans 0% 0%
Page writes by reclaim 6215049 0
Page writes file 6215049 0
Page writes anon 0 0
Page reclaim immediate 70673 90
Sector Reads 81940800 81680456
Sector Writes 100158984 98816036
Page rescued immediate 0 0
Slabs scanned 1366954 22683
While this is not guaranteed in all cases, this particular test showed
a large reduction in direct reclaim activity. It's also worth noting
that no page writes were issued from reclaim context.
This series is not without its hazards. There are at least three areas
that I'm concerned with even though I could not reproduce any problems in
that area.
1. Reclaim/compaction is going to be affected because the amount of reclaim is
no longer targetted at a specific zone. Compaction works on a per-zone basis
so there is no guarantee that reclaiming a few THP's worth page pages will
have a positive impact on compaction success rates.
2. The Slab/LRU reclaim ratio is affected because the frequency the shrinkers
are called is now different. This may or may not be a problem but if it
is, it'll be because shrinkers are not called enough and some balancing
is required.
3. The anon/file reclaim ratio may be affected. Pages about to be dirtied are
distributed between zones and the fair zone allocation policy used to do
something very similar for anon. The distribution is now different but not
necessarily in any way that matters but it's still worth bearing in mind.
VM statistic counters for reclaim decisions are zone-based. If the kernel
is to reclaim on a per-node basis then we need to track per-node
statistics but there is no infrastructure for that. The most notable
change is that the old node_page_state is renamed to
sum_zone_node_page_state. The new node_page_state takes a pglist_data and
uses per-node stats but none exist yet. There is some renaming such as
vm_stat to vm_zone_stat and the addition of vm_node_stat and the renaming
of mod_state to mod_zone_state. Otherwise, this is mostly a mechanical
patch with no functional change. There is a lot of similarity between the
node and zone helpers which is unfortunate but there was no obvious way of
reusing the code and maintaining type safety.
Link: http://lkml.kernel.org/r/1467970510-21195-2-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Rik van Riel <riel@surriel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-28 22:45:24 +00:00
|
|
|
pgdat->per_cpu_nodestats = NULL;
|
2020-06-03 22:58:13 +00:00
|
|
|
|
2015-02-11 23:26:01 +00:00
|
|
|
pr_info("Initmem setup node %d [mem %#018Lx-%#018Lx]\n", nid,
|
2015-09-08 22:04:19 +00:00
|
|
|
(u64)start_pfn << PAGE_SHIFT,
|
|
|
|
end_pfn ? ((u64)end_pfn << PAGE_SHIFT) - 1 : 0);
|
2020-06-03 22:58:13 +00:00
|
|
|
calculate_node_totalpages(pgdat, start_pfn, end_pfn);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
alloc_node_mem_map(pgdat);
|
2018-08-22 04:53:39 +00:00
|
|
|
pgdat_set_deferred_range(pgdat);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2015-09-08 21:59:50 +00:00
|
|
|
free_area_init_core(pgdat);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2020-06-03 22:58:09 +00:00
|
|
|
void __init free_area_init_memoryless_node(int nid)
|
2020-06-03 22:57:02 +00:00
|
|
|
{
|
2020-06-03 22:58:13 +00:00
|
|
|
free_area_init_node(nid);
|
2020-06-03 22:57:02 +00:00
|
|
|
}
|
|
|
|
|
2018-10-30 22:07:44 +00:00
|
|
|
#if !defined(CONFIG_FLAT_NODE_MEM_MAP)
|
2018-10-26 22:10:21 +00:00
|
|
|
/*
|
2020-02-04 01:33:55 +00:00
|
|
|
* Initialize all valid struct pages in the range [spfn, epfn) and mark them
|
|
|
|
* PageReserved(). Return the number of struct pages that were initialized.
|
2018-10-26 22:10:21 +00:00
|
|
|
*/
|
2021-01-26 18:39:46 +00:00
|
|
|
static u64 __init init_unavailable_range(unsigned long spfn, unsigned long epfn)
|
2018-10-26 22:10:21 +00:00
|
|
|
{
|
2021-01-26 18:39:46 +00:00
|
|
|
unsigned long pfn;
|
2018-10-26 22:10:21 +00:00
|
|
|
u64 pgcnt = 0;
|
|
|
|
|
|
|
|
for (pfn = spfn; pfn < epfn; pfn++) {
|
|
|
|
if (!pfn_valid(ALIGN_DOWN(pfn, pageblock_nr_pages))) {
|
|
|
|
pfn = ALIGN_DOWN(pfn, pageblock_nr_pages)
|
|
|
|
+ pageblock_nr_pages - 1;
|
|
|
|
continue;
|
|
|
|
}
|
2021-01-26 18:39:46 +00:00
|
|
|
/*
|
|
|
|
* Use a fake node/zone (0) for now. Some of these pages
|
|
|
|
* (in memblock.reserved but not in memblock.memory) will
|
|
|
|
* get re-initialized via reserve_bootmem_region() later.
|
|
|
|
*/
|
|
|
|
__init_single_page(pfn_to_page(pfn), pfn, 0, 0);
|
2020-02-04 01:33:55 +00:00
|
|
|
__SetPageReserved(pfn_to_page(pfn));
|
2018-10-26 22:10:21 +00:00
|
|
|
pgcnt++;
|
|
|
|
}
|
|
|
|
|
|
|
|
return pgcnt;
|
|
|
|
}
|
|
|
|
|
mm: zero reserved and unavailable struct pages
Some memory is reserved but unavailable: not present in memblock.memory
(because not backed by physical pages), but present in memblock.reserved.
Such memory has backing struct pages, but they are not initialized by
going through __init_single_page().
In some cases these struct pages are accessed even if they do not
contain any data. One example is page_to_pfn() might access page->flags
if this is where section information is stored (CONFIG_SPARSEMEM,
SECTION_IN_PAGE_FLAGS).
One example of such memory: trim_low_memory_range() unconditionally
reserves from pfn 0, but e820__memblock_setup() might provide the
exiting memory from pfn 1 (i.e. KVM).
Since struct pages are zeroed in __init_single_page(), and not during
allocation time, we must zero such struct pages explicitly.
The patch involves adding a new memblock iterator:
for_each_resv_unavail_range(i, p_start, p_end)
Which iterates through reserved && !memory lists, and we zero struct pages
explicitly by calling mm_zero_struct_page().
===
Here is more detailed example of problem that this patch is addressing:
Run tested on qemu with the following arguments:
-enable-kvm -cpu kvm64 -m 512 -smp 2
This patch reports that there are 98 unavailable pages.
They are: pfn 0 and pfns in range [159, 255].
Note, trim_low_memory_range() reserves only pfns in range [0, 15], it does
not reserve [159, 255] ones.
e820__memblock_setup() reports linux that the following physical ranges are
available:
[1 , 158]
[256, 130783]
Notice, that exactly unavailable pfns are missing!
Now, lets check what we have in zone 0: [1, 131039]
pfn 0, is not part of the zone, but pfns [1, 158], are.
However, the bigger problem we have if we do not initialize these struct
pages is with memory hotplug. Because, that path operates at 2M
boundaries (section_nr). And checks if 2M range of pages is hot
removable. It starts with first pfn from zone, rounds it down to 2M
boundary (sturct pages are allocated at 2M boundaries when vmemmap is
created), and checks if that section is hot removable. In this case
start with pfn 1 and convert it down to pfn 0. Later pfn is converted
to struct page, and some fields are checked. Now, if we do not zero
struct pages, we get unpredictable results.
In fact when CONFIG_VM_DEBUG is enabled, and we explicitly set all
vmemmap memory to ones, the following panic is observed with kernel test
without this patch applied:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: is_pageblock_removable_nolock+0x35/0x90
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT
...
task: ffff88001f4e2900 task.stack: ffffc90000314000
RIP: 0010:is_pageblock_removable_nolock+0x35/0x90
Call Trace:
? is_mem_section_removable+0x5a/0xd0
show_mem_removable+0x6b/0xa0
dev_attr_show+0x1b/0x50
sysfs_kf_seq_show+0xa1/0x100
kernfs_seq_show+0x22/0x30
seq_read+0x1ac/0x3a0
kernfs_fop_read+0x36/0x190
? security_file_permission+0x90/0xb0
__vfs_read+0x16/0x30
vfs_read+0x81/0x130
SyS_read+0x44/0xa0
entry_SYSCALL_64_fastpath+0x1f/0xbd
Link: http://lkml.kernel.org/r/20171013173214.27300-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:31 +00:00
|
|
|
/*
|
2021-01-26 18:39:46 +00:00
|
|
|
* Only struct pages that are backed by physical memory are zeroed and
|
|
|
|
* initialized by going through __init_single_page(). But, there are some
|
|
|
|
* struct pages which are reserved in memblock allocator and their fields
|
|
|
|
* may be accessed (for example page_to_pfn() on some configuration accesses
|
|
|
|
* flags). We must explicitly initialize those struct pages.
|
mm: zero remaining unavailable struct pages
Patch series "mm: Fix for movable_node boot option", v3.
This patch series contains a fix for the movable_node boot option issue
which was introduced by commit 124049decbb1 ("x86/e820: put !E820_TYPE_RAM
regions into memblock.reserved").
The commit breaks the option because it changed the memory gap range to
reserved memblock. So, the node is marked as Normal zone even if the SRAT
has Hot pluggable affinity.
First and second patch fix the original issue which the commit tried to
fix, then revert the commit.
This patch (of 3):
There is a kernel panic that is triggered when reading /proc/kpageflags on
the kernel booted with kernel parameter 'memmap=nn[KMG]!ss[KMG]':
BUG: unable to handle kernel paging request at fffffffffffffffe
PGD 9b20e067 P4D 9b20e067 PUD 9b210067 PMD 0
Oops: 0000 [#1] SMP PTI
CPU: 2 PID: 1728 Comm: page-types Not tainted 4.17.0-rc6-mm1-v4.17-rc6-180605-0816-00236-g2dfb086ef02c+ #160
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.fc28 04/01/2014
RIP: 0010:stable_page_flags+0x27/0x3c0
Code: 00 00 00 0f 1f 44 00 00 48 85 ff 0f 84 a0 03 00 00 41 54 55 49 89 fc 53 48 8b 57 08 48 8b 2f 48 8d 42 ff 83 e2 01 48 0f 44 c7 <48> 8b 00 f6 c4 01 0f 84 10 03 00 00 31 db 49 8b 54 24 08 4c 89 e7
RSP: 0018:ffffbbd44111fde0 EFLAGS: 00010202
RAX: fffffffffffffffe RBX: 00007fffffffeff9 RCX: 0000000000000000
RDX: 0000000000000001 RSI: 0000000000000202 RDI: ffffed1182fff5c0
RBP: ffffffffffffffff R08: 0000000000000001 R09: 0000000000000001
R10: ffffbbd44111fed8 R11: 0000000000000000 R12: ffffed1182fff5c0
R13: 00000000000bffd7 R14: 0000000002fff5c0 R15: ffffbbd44111ff10
FS: 00007efc4335a500(0000) GS:ffff93a5bfc00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: fffffffffffffffe CR3: 00000000b2a58000 CR4: 00000000001406e0
Call Trace:
kpageflags_read+0xc7/0x120
proc_reg_read+0x3c/0x60
__vfs_read+0x36/0x170
vfs_read+0x89/0x130
ksys_pread64+0x71/0x90
do_syscall_64+0x5b/0x160
entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7efc42e75e23
Code: 09 00 ba 9f 01 00 00 e8 ab 81 f4 ff 66 2e 0f 1f 84 00 00 00 00 00 90 83 3d 29 0a 2d 00 00 75 13 49 89 ca b8 11 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 34 c3 48 83 ec 08 e8 db d3 01 00 48 89 04 24
According to kernel bisection, this problem became visible due to commit
f7f99100d8d9 which changes how struct pages are initialized.
Memblock layout affects the pfn ranges covered by node/zone. Consider
that we have a VM with 2 NUMA nodes and each node has 4GB memory, and the
default (no memmap= given) memblock layout is like below:
MEMBLOCK configuration:
memory size = 0x00000001fff75c00 reserved size = 0x000000000300c000
memory.cnt = 0x4
memory[0x0] [0x0000000000001000-0x000000000009efff], 0x000000000009e000 bytes on node 0 flags: 0x0
memory[0x1] [0x0000000000100000-0x00000000bffd6fff], 0x00000000bfed7000 bytes on node 0 flags: 0x0
memory[0x2] [0x0000000100000000-0x000000013fffffff], 0x0000000040000000 bytes on node 0 flags: 0x0
memory[0x3] [0x0000000140000000-0x000000023fffffff], 0x0000000100000000 bytes on node 1 flags: 0x0
...
If you give memmap=1G!4G (so it just covers memory[0x2]),
the range [0x100000000-0x13fffffff] is gone:
MEMBLOCK configuration:
memory size = 0x00000001bff75c00 reserved size = 0x000000000300c000
memory.cnt = 0x3
memory[0x0] [0x0000000000001000-0x000000000009efff], 0x000000000009e000 bytes on node 0 flags: 0x0
memory[0x1] [0x0000000000100000-0x00000000bffd6fff], 0x00000000bfed7000 bytes on node 0 flags: 0x0
memory[0x2] [0x0000000140000000-0x000000023fffffff], 0x0000000100000000 bytes on node 1 flags: 0x0
...
This causes shrinking node 0's pfn range because it is calculated by the
address range of memblock.memory. So some of struct pages in the gap
range are left uninitialized.
We have a function zero_resv_unavail() which does zeroing the struct pages
outside memblock.memory, but currently it covers only the reserved
unavailable range (i.e. memblock.memory && !memblock.reserved). This
patch extends it to cover all unavailable range, which fixes the reported
issue.
Link: http://lkml.kernel.org/r/20181002143821.5112-2-msys.mizuma@gmail.com
Fixes: f7f99100d8d9 ("mm: stop zeroing memory during allocation in vmemmap")
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Tested-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Reviewed-by: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:10:15 +00:00
|
|
|
*
|
2021-01-26 18:39:46 +00:00
|
|
|
* This function also addresses a similar issue where struct pages are left
|
|
|
|
* uninitialized because the physical address range is not covered by
|
|
|
|
* memblock.memory or memblock.reserved. That could happen when memblock
|
|
|
|
* layout is manually configured via memmap=, or when the highest physical
|
|
|
|
* address (max_pfn) does not end on a section boundary.
|
mm: zero reserved and unavailable struct pages
Some memory is reserved but unavailable: not present in memblock.memory
(because not backed by physical pages), but present in memblock.reserved.
Such memory has backing struct pages, but they are not initialized by
going through __init_single_page().
In some cases these struct pages are accessed even if they do not
contain any data. One example is page_to_pfn() might access page->flags
if this is where section information is stored (CONFIG_SPARSEMEM,
SECTION_IN_PAGE_FLAGS).
One example of such memory: trim_low_memory_range() unconditionally
reserves from pfn 0, but e820__memblock_setup() might provide the
exiting memory from pfn 1 (i.e. KVM).
Since struct pages are zeroed in __init_single_page(), and not during
allocation time, we must zero such struct pages explicitly.
The patch involves adding a new memblock iterator:
for_each_resv_unavail_range(i, p_start, p_end)
Which iterates through reserved && !memory lists, and we zero struct pages
explicitly by calling mm_zero_struct_page().
===
Here is more detailed example of problem that this patch is addressing:
Run tested on qemu with the following arguments:
-enable-kvm -cpu kvm64 -m 512 -smp 2
This patch reports that there are 98 unavailable pages.
They are: pfn 0 and pfns in range [159, 255].
Note, trim_low_memory_range() reserves only pfns in range [0, 15], it does
not reserve [159, 255] ones.
e820__memblock_setup() reports linux that the following physical ranges are
available:
[1 , 158]
[256, 130783]
Notice, that exactly unavailable pfns are missing!
Now, lets check what we have in zone 0: [1, 131039]
pfn 0, is not part of the zone, but pfns [1, 158], are.
However, the bigger problem we have if we do not initialize these struct
pages is with memory hotplug. Because, that path operates at 2M
boundaries (section_nr). And checks if 2M range of pages is hot
removable. It starts with first pfn from zone, rounds it down to 2M
boundary (sturct pages are allocated at 2M boundaries when vmemmap is
created), and checks if that section is hot removable. In this case
start with pfn 1 and convert it down to pfn 0. Later pfn is converted
to struct page, and some fields are checked. Now, if we do not zero
struct pages, we get unpredictable results.
In fact when CONFIG_VM_DEBUG is enabled, and we explicitly set all
vmemmap memory to ones, the following panic is observed with kernel test
without this patch applied:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: is_pageblock_removable_nolock+0x35/0x90
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT
...
task: ffff88001f4e2900 task.stack: ffffc90000314000
RIP: 0010:is_pageblock_removable_nolock+0x35/0x90
Call Trace:
? is_mem_section_removable+0x5a/0xd0
show_mem_removable+0x6b/0xa0
dev_attr_show+0x1b/0x50
sysfs_kf_seq_show+0xa1/0x100
kernfs_seq_show+0x22/0x30
seq_read+0x1ac/0x3a0
kernfs_fop_read+0x36/0x190
? security_file_permission+0x90/0xb0
__vfs_read+0x16/0x30
vfs_read+0x81/0x130
SyS_read+0x44/0xa0
entry_SYSCALL_64_fastpath+0x1f/0xbd
Link: http://lkml.kernel.org/r/20171013173214.27300-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:31 +00:00
|
|
|
*/
|
2021-01-26 18:39:46 +00:00
|
|
|
static void __init init_unavailable_mem(void)
|
mm: zero reserved and unavailable struct pages
Some memory is reserved but unavailable: not present in memblock.memory
(because not backed by physical pages), but present in memblock.reserved.
Such memory has backing struct pages, but they are not initialized by
going through __init_single_page().
In some cases these struct pages are accessed even if they do not
contain any data. One example is page_to_pfn() might access page->flags
if this is where section information is stored (CONFIG_SPARSEMEM,
SECTION_IN_PAGE_FLAGS).
One example of such memory: trim_low_memory_range() unconditionally
reserves from pfn 0, but e820__memblock_setup() might provide the
exiting memory from pfn 1 (i.e. KVM).
Since struct pages are zeroed in __init_single_page(), and not during
allocation time, we must zero such struct pages explicitly.
The patch involves adding a new memblock iterator:
for_each_resv_unavail_range(i, p_start, p_end)
Which iterates through reserved && !memory lists, and we zero struct pages
explicitly by calling mm_zero_struct_page().
===
Here is more detailed example of problem that this patch is addressing:
Run tested on qemu with the following arguments:
-enable-kvm -cpu kvm64 -m 512 -smp 2
This patch reports that there are 98 unavailable pages.
They are: pfn 0 and pfns in range [159, 255].
Note, trim_low_memory_range() reserves only pfns in range [0, 15], it does
not reserve [159, 255] ones.
e820__memblock_setup() reports linux that the following physical ranges are
available:
[1 , 158]
[256, 130783]
Notice, that exactly unavailable pfns are missing!
Now, lets check what we have in zone 0: [1, 131039]
pfn 0, is not part of the zone, but pfns [1, 158], are.
However, the bigger problem we have if we do not initialize these struct
pages is with memory hotplug. Because, that path operates at 2M
boundaries (section_nr). And checks if 2M range of pages is hot
removable. It starts with first pfn from zone, rounds it down to 2M
boundary (sturct pages are allocated at 2M boundaries when vmemmap is
created), and checks if that section is hot removable. In this case
start with pfn 1 and convert it down to pfn 0. Later pfn is converted
to struct page, and some fields are checked. Now, if we do not zero
struct pages, we get unpredictable results.
In fact when CONFIG_VM_DEBUG is enabled, and we explicitly set all
vmemmap memory to ones, the following panic is observed with kernel test
without this patch applied:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: is_pageblock_removable_nolock+0x35/0x90
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT
...
task: ffff88001f4e2900 task.stack: ffffc90000314000
RIP: 0010:is_pageblock_removable_nolock+0x35/0x90
Call Trace:
? is_mem_section_removable+0x5a/0xd0
show_mem_removable+0x6b/0xa0
dev_attr_show+0x1b/0x50
sysfs_kf_seq_show+0xa1/0x100
kernfs_seq_show+0x22/0x30
seq_read+0x1ac/0x3a0
kernfs_fop_read+0x36/0x190
? security_file_permission+0x90/0xb0
__vfs_read+0x16/0x30
vfs_read+0x81/0x130
SyS_read+0x44/0xa0
entry_SYSCALL_64_fastpath+0x1f/0xbd
Link: http://lkml.kernel.org/r/20171013173214.27300-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:31 +00:00
|
|
|
{
|
2021-01-26 18:39:46 +00:00
|
|
|
phys_addr_t start, end;
|
|
|
|
u64 i, pgcnt;
|
|
|
|
phys_addr_t next = 0;
|
mm: zero reserved and unavailable struct pages
Some memory is reserved but unavailable: not present in memblock.memory
(because not backed by physical pages), but present in memblock.reserved.
Such memory has backing struct pages, but they are not initialized by
going through __init_single_page().
In some cases these struct pages are accessed even if they do not
contain any data. One example is page_to_pfn() might access page->flags
if this is where section information is stored (CONFIG_SPARSEMEM,
SECTION_IN_PAGE_FLAGS).
One example of such memory: trim_low_memory_range() unconditionally
reserves from pfn 0, but e820__memblock_setup() might provide the
exiting memory from pfn 1 (i.e. KVM).
Since struct pages are zeroed in __init_single_page(), and not during
allocation time, we must zero such struct pages explicitly.
The patch involves adding a new memblock iterator:
for_each_resv_unavail_range(i, p_start, p_end)
Which iterates through reserved && !memory lists, and we zero struct pages
explicitly by calling mm_zero_struct_page().
===
Here is more detailed example of problem that this patch is addressing:
Run tested on qemu with the following arguments:
-enable-kvm -cpu kvm64 -m 512 -smp 2
This patch reports that there are 98 unavailable pages.
They are: pfn 0 and pfns in range [159, 255].
Note, trim_low_memory_range() reserves only pfns in range [0, 15], it does
not reserve [159, 255] ones.
e820__memblock_setup() reports linux that the following physical ranges are
available:
[1 , 158]
[256, 130783]
Notice, that exactly unavailable pfns are missing!
Now, lets check what we have in zone 0: [1, 131039]
pfn 0, is not part of the zone, but pfns [1, 158], are.
However, the bigger problem we have if we do not initialize these struct
pages is with memory hotplug. Because, that path operates at 2M
boundaries (section_nr). And checks if 2M range of pages is hot
removable. It starts with first pfn from zone, rounds it down to 2M
boundary (sturct pages are allocated at 2M boundaries when vmemmap is
created), and checks if that section is hot removable. In this case
start with pfn 1 and convert it down to pfn 0. Later pfn is converted
to struct page, and some fields are checked. Now, if we do not zero
struct pages, we get unpredictable results.
In fact when CONFIG_VM_DEBUG is enabled, and we explicitly set all
vmemmap memory to ones, the following panic is observed with kernel test
without this patch applied:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: is_pageblock_removable_nolock+0x35/0x90
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT
...
task: ffff88001f4e2900 task.stack: ffffc90000314000
RIP: 0010:is_pageblock_removable_nolock+0x35/0x90
Call Trace:
? is_mem_section_removable+0x5a/0xd0
show_mem_removable+0x6b/0xa0
dev_attr_show+0x1b/0x50
sysfs_kf_seq_show+0xa1/0x100
kernfs_seq_show+0x22/0x30
seq_read+0x1ac/0x3a0
kernfs_fop_read+0x36/0x190
? security_file_permission+0x90/0xb0
__vfs_read+0x16/0x30
vfs_read+0x81/0x130
SyS_read+0x44/0xa0
entry_SYSCALL_64_fastpath+0x1f/0xbd
Link: http://lkml.kernel.org/r/20171013173214.27300-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:31 +00:00
|
|
|
|
|
|
|
/*
|
2021-01-26 18:39:46 +00:00
|
|
|
* Loop through unavailable ranges not covered by memblock.memory.
|
mm: zero reserved and unavailable struct pages
Some memory is reserved but unavailable: not present in memblock.memory
(because not backed by physical pages), but present in memblock.reserved.
Such memory has backing struct pages, but they are not initialized by
going through __init_single_page().
In some cases these struct pages are accessed even if they do not
contain any data. One example is page_to_pfn() might access page->flags
if this is where section information is stored (CONFIG_SPARSEMEM,
SECTION_IN_PAGE_FLAGS).
One example of such memory: trim_low_memory_range() unconditionally
reserves from pfn 0, but e820__memblock_setup() might provide the
exiting memory from pfn 1 (i.e. KVM).
Since struct pages are zeroed in __init_single_page(), and not during
allocation time, we must zero such struct pages explicitly.
The patch involves adding a new memblock iterator:
for_each_resv_unavail_range(i, p_start, p_end)
Which iterates through reserved && !memory lists, and we zero struct pages
explicitly by calling mm_zero_struct_page().
===
Here is more detailed example of problem that this patch is addressing:
Run tested on qemu with the following arguments:
-enable-kvm -cpu kvm64 -m 512 -smp 2
This patch reports that there are 98 unavailable pages.
They are: pfn 0 and pfns in range [159, 255].
Note, trim_low_memory_range() reserves only pfns in range [0, 15], it does
not reserve [159, 255] ones.
e820__memblock_setup() reports linux that the following physical ranges are
available:
[1 , 158]
[256, 130783]
Notice, that exactly unavailable pfns are missing!
Now, lets check what we have in zone 0: [1, 131039]
pfn 0, is not part of the zone, but pfns [1, 158], are.
However, the bigger problem we have if we do not initialize these struct
pages is with memory hotplug. Because, that path operates at 2M
boundaries (section_nr). And checks if 2M range of pages is hot
removable. It starts with first pfn from zone, rounds it down to 2M
boundary (sturct pages are allocated at 2M boundaries when vmemmap is
created), and checks if that section is hot removable. In this case
start with pfn 1 and convert it down to pfn 0. Later pfn is converted
to struct page, and some fields are checked. Now, if we do not zero
struct pages, we get unpredictable results.
In fact when CONFIG_VM_DEBUG is enabled, and we explicitly set all
vmemmap memory to ones, the following panic is observed with kernel test
without this patch applied:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: is_pageblock_removable_nolock+0x35/0x90
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT
...
task: ffff88001f4e2900 task.stack: ffffc90000314000
RIP: 0010:is_pageblock_removable_nolock+0x35/0x90
Call Trace:
? is_mem_section_removable+0x5a/0xd0
show_mem_removable+0x6b/0xa0
dev_attr_show+0x1b/0x50
sysfs_kf_seq_show+0xa1/0x100
kernfs_seq_show+0x22/0x30
seq_read+0x1ac/0x3a0
kernfs_fop_read+0x36/0x190
? security_file_permission+0x90/0xb0
__vfs_read+0x16/0x30
vfs_read+0x81/0x130
SyS_read+0x44/0xa0
entry_SYSCALL_64_fastpath+0x1f/0xbd
Link: http://lkml.kernel.org/r/20171013173214.27300-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:31 +00:00
|
|
|
*/
|
|
|
|
pgcnt = 0;
|
2021-01-26 18:39:46 +00:00
|
|
|
for_each_mem_range(i, &start, &end) {
|
2018-10-26 22:10:21 +00:00
|
|
|
if (next < start)
|
2021-01-26 18:39:46 +00:00
|
|
|
pgcnt += init_unavailable_range(PFN_DOWN(next),
|
|
|
|
PFN_UP(start));
|
mm: zero remaining unavailable struct pages
Patch series "mm: Fix for movable_node boot option", v3.
This patch series contains a fix for the movable_node boot option issue
which was introduced by commit 124049decbb1 ("x86/e820: put !E820_TYPE_RAM
regions into memblock.reserved").
The commit breaks the option because it changed the memory gap range to
reserved memblock. So, the node is marked as Normal zone even if the SRAT
has Hot pluggable affinity.
First and second patch fix the original issue which the commit tried to
fix, then revert the commit.
This patch (of 3):
There is a kernel panic that is triggered when reading /proc/kpageflags on
the kernel booted with kernel parameter 'memmap=nn[KMG]!ss[KMG]':
BUG: unable to handle kernel paging request at fffffffffffffffe
PGD 9b20e067 P4D 9b20e067 PUD 9b210067 PMD 0
Oops: 0000 [#1] SMP PTI
CPU: 2 PID: 1728 Comm: page-types Not tainted 4.17.0-rc6-mm1-v4.17-rc6-180605-0816-00236-g2dfb086ef02c+ #160
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.fc28 04/01/2014
RIP: 0010:stable_page_flags+0x27/0x3c0
Code: 00 00 00 0f 1f 44 00 00 48 85 ff 0f 84 a0 03 00 00 41 54 55 49 89 fc 53 48 8b 57 08 48 8b 2f 48 8d 42 ff 83 e2 01 48 0f 44 c7 <48> 8b 00 f6 c4 01 0f 84 10 03 00 00 31 db 49 8b 54 24 08 4c 89 e7
RSP: 0018:ffffbbd44111fde0 EFLAGS: 00010202
RAX: fffffffffffffffe RBX: 00007fffffffeff9 RCX: 0000000000000000
RDX: 0000000000000001 RSI: 0000000000000202 RDI: ffffed1182fff5c0
RBP: ffffffffffffffff R08: 0000000000000001 R09: 0000000000000001
R10: ffffbbd44111fed8 R11: 0000000000000000 R12: ffffed1182fff5c0
R13: 00000000000bffd7 R14: 0000000002fff5c0 R15: ffffbbd44111ff10
FS: 00007efc4335a500(0000) GS:ffff93a5bfc00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: fffffffffffffffe CR3: 00000000b2a58000 CR4: 00000000001406e0
Call Trace:
kpageflags_read+0xc7/0x120
proc_reg_read+0x3c/0x60
__vfs_read+0x36/0x170
vfs_read+0x89/0x130
ksys_pread64+0x71/0x90
do_syscall_64+0x5b/0x160
entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7efc42e75e23
Code: 09 00 ba 9f 01 00 00 e8 ab 81 f4 ff 66 2e 0f 1f 84 00 00 00 00 00 90 83 3d 29 0a 2d 00 00 75 13 49 89 ca b8 11 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 34 c3 48 83 ec 08 e8 db d3 01 00 48 89 04 24
According to kernel bisection, this problem became visible due to commit
f7f99100d8d9 which changes how struct pages are initialized.
Memblock layout affects the pfn ranges covered by node/zone. Consider
that we have a VM with 2 NUMA nodes and each node has 4GB memory, and the
default (no memmap= given) memblock layout is like below:
MEMBLOCK configuration:
memory size = 0x00000001fff75c00 reserved size = 0x000000000300c000
memory.cnt = 0x4
memory[0x0] [0x0000000000001000-0x000000000009efff], 0x000000000009e000 bytes on node 0 flags: 0x0
memory[0x1] [0x0000000000100000-0x00000000bffd6fff], 0x00000000bfed7000 bytes on node 0 flags: 0x0
memory[0x2] [0x0000000100000000-0x000000013fffffff], 0x0000000040000000 bytes on node 0 flags: 0x0
memory[0x3] [0x0000000140000000-0x000000023fffffff], 0x0000000100000000 bytes on node 1 flags: 0x0
...
If you give memmap=1G!4G (so it just covers memory[0x2]),
the range [0x100000000-0x13fffffff] is gone:
MEMBLOCK configuration:
memory size = 0x00000001bff75c00 reserved size = 0x000000000300c000
memory.cnt = 0x3
memory[0x0] [0x0000000000001000-0x000000000009efff], 0x000000000009e000 bytes on node 0 flags: 0x0
memory[0x1] [0x0000000000100000-0x00000000bffd6fff], 0x00000000bfed7000 bytes on node 0 flags: 0x0
memory[0x2] [0x0000000140000000-0x000000023fffffff], 0x0000000100000000 bytes on node 1 flags: 0x0
...
This causes shrinking node 0's pfn range because it is calculated by the
address range of memblock.memory. So some of struct pages in the gap
range are left uninitialized.
We have a function zero_resv_unavail() which does zeroing the struct pages
outside memblock.memory, but currently it covers only the reserved
unavailable range (i.e. memblock.memory && !memblock.reserved). This
patch extends it to cover all unavailable range, which fixes the reported
issue.
Link: http://lkml.kernel.org/r/20181002143821.5112-2-msys.mizuma@gmail.com
Fixes: f7f99100d8d9 ("mm: stop zeroing memory during allocation in vmemmap")
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Tested-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Reviewed-by: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:10:15 +00:00
|
|
|
next = end;
|
|
|
|
}
|
mm/page_alloc.c: fix uninitialized memmaps on a partially populated last section
Patch series "mm: fix max_pfn not falling on section boundary", v2.
Playing with different memory sizes for a x86-64 guest, I discovered that
some memmaps (highest section if max_mem does not fall on the section
boundary) are marked as being valid and online, but contain garbage. We
have to properly initialize these memmaps.
Looking at /proc/kpageflags and friends, I found some more issues,
partially related to this.
This patch (of 3):
If max_pfn is not aligned to a section boundary, we can easily run into
BUGs. This can e.g., be triggered on x86-64 under QEMU by specifying a
memory size that is not a multiple of 128MB (e.g., 4097MB, but also
4160MB). I was told that on real HW, we can easily have this scenario
(esp., one of the main reasons sub-section hotadd of devmem was added).
The issue is, that we have a valid memmap (pfn_valid()) for the whole
section, and the whole section will be marked "online".
pfn_to_online_page() will succeed, but the memmap contains garbage.
E.g., doing a "./page-types -r -a 0x144001" when QEMU was started with "-m
4160M" - (see tools/vm/page-types.c):
[ 200.476376] BUG: unable to handle page fault for address: fffffffffffffffe
[ 200.477500] #PF: supervisor read access in kernel mode
[ 200.478334] #PF: error_code(0x0000) - not-present page
[ 200.479076] PGD 59614067 P4D 59614067 PUD 59616067 PMD 0
[ 200.479557] Oops: 0000 [#4] SMP NOPTI
[ 200.479875] CPU: 0 PID: 603 Comm: page-types Tainted: G D W 5.5.0-rc1-next-20191209 #93
[ 200.480646] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu4
[ 200.481648] RIP: 0010:stable_page_flags+0x4d/0x410
[ 200.482061] Code: f3 ff 41 89 c0 48 b8 00 00 00 00 01 00 00 00 45 84 c0 0f 85 cd 02 00 00 48 8b 53 08 48 8b 2b 48f
[ 200.483644] RSP: 0018:ffffb139401cbe60 EFLAGS: 00010202
[ 200.484091] RAX: fffffffffffffffe RBX: fffffbeec5100040 RCX: 0000000000000000
[ 200.484697] RDX: 0000000000000001 RSI: ffffffff9535c7cd RDI: 0000000000000246
[ 200.485313] RBP: ffffffffffffffff R08: 0000000000000000 R09: 0000000000000000
[ 200.485917] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000144001
[ 200.486523] R13: 00007ffd6ba55f48 R14: 00007ffd6ba55f40 R15: ffffb139401cbf08
[ 200.487130] FS: 00007f68df717580(0000) GS:ffff9ec77fa00000(0000) knlGS:0000000000000000
[ 200.487804] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 200.488295] CR2: fffffffffffffffe CR3: 0000000135d48000 CR4: 00000000000006f0
[ 200.488897] Call Trace:
[ 200.489115] kpageflags_read+0xe9/0x140
[ 200.489447] proc_reg_read+0x3c/0x60
[ 200.489755] vfs_read+0xc2/0x170
[ 200.490037] ksys_pread64+0x65/0xa0
[ 200.490352] do_syscall_64+0x5c/0xa0
[ 200.490665] entry_SYSCALL_64_after_hwframe+0x49/0xbe
But it can be triggered much easier via "cat /proc/kpageflags > /dev/null"
after cold/hot plugging a DIMM to such a system:
[root@localhost ~]# cat /proc/kpageflags > /dev/null
[ 111.517275] BUG: unable to handle page fault for address: fffffffffffffffe
[ 111.517907] #PF: supervisor read access in kernel mode
[ 111.518333] #PF: error_code(0x0000) - not-present page
[ 111.518771] PGD a240e067 P4D a240e067 PUD a2410067 PMD 0
This patch fixes that by at least zero-ing out that memmap (so e.g.,
page_to_pfn() will not crash). Commit 907ec5fca3dc ("mm: zero remaining
unavailable struct pages") tried to fix a similar issue, but forgot to
consider this special case.
After this patch, there are still problems to solve. E.g., not all of
these pages falling into a memory hole will actually get initialized later
and set PageReserved - they are only zeroed out - but at least the
immediate crashes are gone. A follow-up patch will take care of this.
Link: http://lkml.kernel.org/r/20191211163201.17179-2-david@redhat.com
Fixes: f7f99100d8d9 ("mm: stop zeroing memory during allocation in vmemmap")
Signed-off-by: David Hildenbrand <david@redhat.com>
Tested-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Bob Picco <bob.picco@oracle.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: <stable@vger.kernel.org> [4.15+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-02-04 01:33:48 +00:00
|
|
|
|
|
|
|
/*
|
2021-01-26 18:39:46 +00:00
|
|
|
* Early sections always have a fully populated memmap for the whole
|
|
|
|
* section - see pfn_valid(). If the last section has holes at the
|
|
|
|
* end and that section is marked "online", the memmap will be
|
|
|
|
* considered initialized. Make sure that memmap has a well defined
|
|
|
|
* state.
|
mm/page_alloc.c: fix uninitialized memmaps on a partially populated last section
Patch series "mm: fix max_pfn not falling on section boundary", v2.
Playing with different memory sizes for a x86-64 guest, I discovered that
some memmaps (highest section if max_mem does not fall on the section
boundary) are marked as being valid and online, but contain garbage. We
have to properly initialize these memmaps.
Looking at /proc/kpageflags and friends, I found some more issues,
partially related to this.
This patch (of 3):
If max_pfn is not aligned to a section boundary, we can easily run into
BUGs. This can e.g., be triggered on x86-64 under QEMU by specifying a
memory size that is not a multiple of 128MB (e.g., 4097MB, but also
4160MB). I was told that on real HW, we can easily have this scenario
(esp., one of the main reasons sub-section hotadd of devmem was added).
The issue is, that we have a valid memmap (pfn_valid()) for the whole
section, and the whole section will be marked "online".
pfn_to_online_page() will succeed, but the memmap contains garbage.
E.g., doing a "./page-types -r -a 0x144001" when QEMU was started with "-m
4160M" - (see tools/vm/page-types.c):
[ 200.476376] BUG: unable to handle page fault for address: fffffffffffffffe
[ 200.477500] #PF: supervisor read access in kernel mode
[ 200.478334] #PF: error_code(0x0000) - not-present page
[ 200.479076] PGD 59614067 P4D 59614067 PUD 59616067 PMD 0
[ 200.479557] Oops: 0000 [#4] SMP NOPTI
[ 200.479875] CPU: 0 PID: 603 Comm: page-types Tainted: G D W 5.5.0-rc1-next-20191209 #93
[ 200.480646] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu4
[ 200.481648] RIP: 0010:stable_page_flags+0x4d/0x410
[ 200.482061] Code: f3 ff 41 89 c0 48 b8 00 00 00 00 01 00 00 00 45 84 c0 0f 85 cd 02 00 00 48 8b 53 08 48 8b 2b 48f
[ 200.483644] RSP: 0018:ffffb139401cbe60 EFLAGS: 00010202
[ 200.484091] RAX: fffffffffffffffe RBX: fffffbeec5100040 RCX: 0000000000000000
[ 200.484697] RDX: 0000000000000001 RSI: ffffffff9535c7cd RDI: 0000000000000246
[ 200.485313] RBP: ffffffffffffffff R08: 0000000000000000 R09: 0000000000000000
[ 200.485917] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000144001
[ 200.486523] R13: 00007ffd6ba55f48 R14: 00007ffd6ba55f40 R15: ffffb139401cbf08
[ 200.487130] FS: 00007f68df717580(0000) GS:ffff9ec77fa00000(0000) knlGS:0000000000000000
[ 200.487804] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 200.488295] CR2: fffffffffffffffe CR3: 0000000135d48000 CR4: 00000000000006f0
[ 200.488897] Call Trace:
[ 200.489115] kpageflags_read+0xe9/0x140
[ 200.489447] proc_reg_read+0x3c/0x60
[ 200.489755] vfs_read+0xc2/0x170
[ 200.490037] ksys_pread64+0x65/0xa0
[ 200.490352] do_syscall_64+0x5c/0xa0
[ 200.490665] entry_SYSCALL_64_after_hwframe+0x49/0xbe
But it can be triggered much easier via "cat /proc/kpageflags > /dev/null"
after cold/hot plugging a DIMM to such a system:
[root@localhost ~]# cat /proc/kpageflags > /dev/null
[ 111.517275] BUG: unable to handle page fault for address: fffffffffffffffe
[ 111.517907] #PF: supervisor read access in kernel mode
[ 111.518333] #PF: error_code(0x0000) - not-present page
[ 111.518771] PGD a240e067 P4D a240e067 PUD a2410067 PMD 0
This patch fixes that by at least zero-ing out that memmap (so e.g.,
page_to_pfn() will not crash). Commit 907ec5fca3dc ("mm: zero remaining
unavailable struct pages") tried to fix a similar issue, but forgot to
consider this special case.
After this patch, there are still problems to solve. E.g., not all of
these pages falling into a memory hole will actually get initialized later
and set PageReserved - they are only zeroed out - but at least the
immediate crashes are gone. A follow-up patch will take care of this.
Link: http://lkml.kernel.org/r/20191211163201.17179-2-david@redhat.com
Fixes: f7f99100d8d9 ("mm: stop zeroing memory during allocation in vmemmap")
Signed-off-by: David Hildenbrand <david@redhat.com>
Tested-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Bob Picco <bob.picco@oracle.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: <stable@vger.kernel.org> [4.15+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-02-04 01:33:48 +00:00
|
|
|
*/
|
2021-01-26 18:39:46 +00:00
|
|
|
pgcnt += init_unavailable_range(PFN_DOWN(next),
|
|
|
|
round_up(max_pfn, PAGES_PER_SECTION));
|
mm: zero remaining unavailable struct pages
Patch series "mm: Fix for movable_node boot option", v3.
This patch series contains a fix for the movable_node boot option issue
which was introduced by commit 124049decbb1 ("x86/e820: put !E820_TYPE_RAM
regions into memblock.reserved").
The commit breaks the option because it changed the memory gap range to
reserved memblock. So, the node is marked as Normal zone even if the SRAT
has Hot pluggable affinity.
First and second patch fix the original issue which the commit tried to
fix, then revert the commit.
This patch (of 3):
There is a kernel panic that is triggered when reading /proc/kpageflags on
the kernel booted with kernel parameter 'memmap=nn[KMG]!ss[KMG]':
BUG: unable to handle kernel paging request at fffffffffffffffe
PGD 9b20e067 P4D 9b20e067 PUD 9b210067 PMD 0
Oops: 0000 [#1] SMP PTI
CPU: 2 PID: 1728 Comm: page-types Not tainted 4.17.0-rc6-mm1-v4.17-rc6-180605-0816-00236-g2dfb086ef02c+ #160
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.fc28 04/01/2014
RIP: 0010:stable_page_flags+0x27/0x3c0
Code: 00 00 00 0f 1f 44 00 00 48 85 ff 0f 84 a0 03 00 00 41 54 55 49 89 fc 53 48 8b 57 08 48 8b 2f 48 8d 42 ff 83 e2 01 48 0f 44 c7 <48> 8b 00 f6 c4 01 0f 84 10 03 00 00 31 db 49 8b 54 24 08 4c 89 e7
RSP: 0018:ffffbbd44111fde0 EFLAGS: 00010202
RAX: fffffffffffffffe RBX: 00007fffffffeff9 RCX: 0000000000000000
RDX: 0000000000000001 RSI: 0000000000000202 RDI: ffffed1182fff5c0
RBP: ffffffffffffffff R08: 0000000000000001 R09: 0000000000000001
R10: ffffbbd44111fed8 R11: 0000000000000000 R12: ffffed1182fff5c0
R13: 00000000000bffd7 R14: 0000000002fff5c0 R15: ffffbbd44111ff10
FS: 00007efc4335a500(0000) GS:ffff93a5bfc00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: fffffffffffffffe CR3: 00000000b2a58000 CR4: 00000000001406e0
Call Trace:
kpageflags_read+0xc7/0x120
proc_reg_read+0x3c/0x60
__vfs_read+0x36/0x170
vfs_read+0x89/0x130
ksys_pread64+0x71/0x90
do_syscall_64+0x5b/0x160
entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7efc42e75e23
Code: 09 00 ba 9f 01 00 00 e8 ab 81 f4 ff 66 2e 0f 1f 84 00 00 00 00 00 90 83 3d 29 0a 2d 00 00 75 13 49 89 ca b8 11 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 34 c3 48 83 ec 08 e8 db d3 01 00 48 89 04 24
According to kernel bisection, this problem became visible due to commit
f7f99100d8d9 which changes how struct pages are initialized.
Memblock layout affects the pfn ranges covered by node/zone. Consider
that we have a VM with 2 NUMA nodes and each node has 4GB memory, and the
default (no memmap= given) memblock layout is like below:
MEMBLOCK configuration:
memory size = 0x00000001fff75c00 reserved size = 0x000000000300c000
memory.cnt = 0x4
memory[0x0] [0x0000000000001000-0x000000000009efff], 0x000000000009e000 bytes on node 0 flags: 0x0
memory[0x1] [0x0000000000100000-0x00000000bffd6fff], 0x00000000bfed7000 bytes on node 0 flags: 0x0
memory[0x2] [0x0000000100000000-0x000000013fffffff], 0x0000000040000000 bytes on node 0 flags: 0x0
memory[0x3] [0x0000000140000000-0x000000023fffffff], 0x0000000100000000 bytes on node 1 flags: 0x0
...
If you give memmap=1G!4G (so it just covers memory[0x2]),
the range [0x100000000-0x13fffffff] is gone:
MEMBLOCK configuration:
memory size = 0x00000001bff75c00 reserved size = 0x000000000300c000
memory.cnt = 0x3
memory[0x0] [0x0000000000001000-0x000000000009efff], 0x000000000009e000 bytes on node 0 flags: 0x0
memory[0x1] [0x0000000000100000-0x00000000bffd6fff], 0x00000000bfed7000 bytes on node 0 flags: 0x0
memory[0x2] [0x0000000140000000-0x000000023fffffff], 0x0000000100000000 bytes on node 1 flags: 0x0
...
This causes shrinking node 0's pfn range because it is calculated by the
address range of memblock.memory. So some of struct pages in the gap
range are left uninitialized.
We have a function zero_resv_unavail() which does zeroing the struct pages
outside memblock.memory, but currently it covers only the reserved
unavailable range (i.e. memblock.memory && !memblock.reserved). This
patch extends it to cover all unavailable range, which fixes the reported
issue.
Link: http://lkml.kernel.org/r/20181002143821.5112-2-msys.mizuma@gmail.com
Fixes: f7f99100d8d9 ("mm: stop zeroing memory during allocation in vmemmap")
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Tested-by: Oscar Salvador <osalvador@suse.de>
Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Reviewed-by: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:10:15 +00:00
|
|
|
|
mm: zero reserved and unavailable struct pages
Some memory is reserved but unavailable: not present in memblock.memory
(because not backed by physical pages), but present in memblock.reserved.
Such memory has backing struct pages, but they are not initialized by
going through __init_single_page().
In some cases these struct pages are accessed even if they do not
contain any data. One example is page_to_pfn() might access page->flags
if this is where section information is stored (CONFIG_SPARSEMEM,
SECTION_IN_PAGE_FLAGS).
One example of such memory: trim_low_memory_range() unconditionally
reserves from pfn 0, but e820__memblock_setup() might provide the
exiting memory from pfn 1 (i.e. KVM).
Since struct pages are zeroed in __init_single_page(), and not during
allocation time, we must zero such struct pages explicitly.
The patch involves adding a new memblock iterator:
for_each_resv_unavail_range(i, p_start, p_end)
Which iterates through reserved && !memory lists, and we zero struct pages
explicitly by calling mm_zero_struct_page().
===
Here is more detailed example of problem that this patch is addressing:
Run tested on qemu with the following arguments:
-enable-kvm -cpu kvm64 -m 512 -smp 2
This patch reports that there are 98 unavailable pages.
They are: pfn 0 and pfns in range [159, 255].
Note, trim_low_memory_range() reserves only pfns in range [0, 15], it does
not reserve [159, 255] ones.
e820__memblock_setup() reports linux that the following physical ranges are
available:
[1 , 158]
[256, 130783]
Notice, that exactly unavailable pfns are missing!
Now, lets check what we have in zone 0: [1, 131039]
pfn 0, is not part of the zone, but pfns [1, 158], are.
However, the bigger problem we have if we do not initialize these struct
pages is with memory hotplug. Because, that path operates at 2M
boundaries (section_nr). And checks if 2M range of pages is hot
removable. It starts with first pfn from zone, rounds it down to 2M
boundary (sturct pages are allocated at 2M boundaries when vmemmap is
created), and checks if that section is hot removable. In this case
start with pfn 1 and convert it down to pfn 0. Later pfn is converted
to struct page, and some fields are checked. Now, if we do not zero
struct pages, we get unpredictable results.
In fact when CONFIG_VM_DEBUG is enabled, and we explicitly set all
vmemmap memory to ones, the following panic is observed with kernel test
without this patch applied:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: is_pageblock_removable_nolock+0x35/0x90
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT
...
task: ffff88001f4e2900 task.stack: ffffc90000314000
RIP: 0010:is_pageblock_removable_nolock+0x35/0x90
Call Trace:
? is_mem_section_removable+0x5a/0xd0
show_mem_removable+0x6b/0xa0
dev_attr_show+0x1b/0x50
sysfs_kf_seq_show+0xa1/0x100
kernfs_seq_show+0x22/0x30
seq_read+0x1ac/0x3a0
kernfs_fop_read+0x36/0x190
? security_file_permission+0x90/0xb0
__vfs_read+0x16/0x30
vfs_read+0x81/0x130
SyS_read+0x44/0xa0
entry_SYSCALL_64_fastpath+0x1f/0xbd
Link: http://lkml.kernel.org/r/20171013173214.27300-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:31 +00:00
|
|
|
/*
|
|
|
|
* Struct pages that do not have backing memory. This could be because
|
|
|
|
* firmware is using some of this memory, or for some other reasons.
|
|
|
|
*/
|
|
|
|
if (pgcnt)
|
2021-01-26 18:39:46 +00:00
|
|
|
pr_info("Zeroed struct page in unavailable ranges: %lld pages", pgcnt);
|
mm: zero reserved and unavailable struct pages
Some memory is reserved but unavailable: not present in memblock.memory
(because not backed by physical pages), but present in memblock.reserved.
Such memory has backing struct pages, but they are not initialized by
going through __init_single_page().
In some cases these struct pages are accessed even if they do not
contain any data. One example is page_to_pfn() might access page->flags
if this is where section information is stored (CONFIG_SPARSEMEM,
SECTION_IN_PAGE_FLAGS).
One example of such memory: trim_low_memory_range() unconditionally
reserves from pfn 0, but e820__memblock_setup() might provide the
exiting memory from pfn 1 (i.e. KVM).
Since struct pages are zeroed in __init_single_page(), and not during
allocation time, we must zero such struct pages explicitly.
The patch involves adding a new memblock iterator:
for_each_resv_unavail_range(i, p_start, p_end)
Which iterates through reserved && !memory lists, and we zero struct pages
explicitly by calling mm_zero_struct_page().
===
Here is more detailed example of problem that this patch is addressing:
Run tested on qemu with the following arguments:
-enable-kvm -cpu kvm64 -m 512 -smp 2
This patch reports that there are 98 unavailable pages.
They are: pfn 0 and pfns in range [159, 255].
Note, trim_low_memory_range() reserves only pfns in range [0, 15], it does
not reserve [159, 255] ones.
e820__memblock_setup() reports linux that the following physical ranges are
available:
[1 , 158]
[256, 130783]
Notice, that exactly unavailable pfns are missing!
Now, lets check what we have in zone 0: [1, 131039]
pfn 0, is not part of the zone, but pfns [1, 158], are.
However, the bigger problem we have if we do not initialize these struct
pages is with memory hotplug. Because, that path operates at 2M
boundaries (section_nr). And checks if 2M range of pages is hot
removable. It starts with first pfn from zone, rounds it down to 2M
boundary (sturct pages are allocated at 2M boundaries when vmemmap is
created), and checks if that section is hot removable. In this case
start with pfn 1 and convert it down to pfn 0. Later pfn is converted
to struct page, and some fields are checked. Now, if we do not zero
struct pages, we get unpredictable results.
In fact when CONFIG_VM_DEBUG is enabled, and we explicitly set all
vmemmap memory to ones, the following panic is observed with kernel test
without this patch applied:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: is_pageblock_removable_nolock+0x35/0x90
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT
...
task: ffff88001f4e2900 task.stack: ffffc90000314000
RIP: 0010:is_pageblock_removable_nolock+0x35/0x90
Call Trace:
? is_mem_section_removable+0x5a/0xd0
show_mem_removable+0x6b/0xa0
dev_attr_show+0x1b/0x50
sysfs_kf_seq_show+0xa1/0x100
kernfs_seq_show+0x22/0x30
seq_read+0x1ac/0x3a0
kernfs_fop_read+0x36/0x190
? security_file_permission+0x90/0xb0
__vfs_read+0x16/0x30
vfs_read+0x81/0x130
SyS_read+0x44/0xa0
entry_SYSCALL_64_fastpath+0x1f/0xbd
Link: http://lkml.kernel.org/r/20171013173214.27300-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:31 +00:00
|
|
|
}
|
2020-02-04 01:33:55 +00:00
|
|
|
#else
|
|
|
|
static inline void __init init_unavailable_mem(void)
|
|
|
|
{
|
|
|
|
}
|
2018-10-30 22:07:44 +00:00
|
|
|
#endif /* !CONFIG_FLAT_NODE_MEM_MAP */
|
mm: zero reserved and unavailable struct pages
Some memory is reserved but unavailable: not present in memblock.memory
(because not backed by physical pages), but present in memblock.reserved.
Such memory has backing struct pages, but they are not initialized by
going through __init_single_page().
In some cases these struct pages are accessed even if they do not
contain any data. One example is page_to_pfn() might access page->flags
if this is where section information is stored (CONFIG_SPARSEMEM,
SECTION_IN_PAGE_FLAGS).
One example of such memory: trim_low_memory_range() unconditionally
reserves from pfn 0, but e820__memblock_setup() might provide the
exiting memory from pfn 1 (i.e. KVM).
Since struct pages are zeroed in __init_single_page(), and not during
allocation time, we must zero such struct pages explicitly.
The patch involves adding a new memblock iterator:
for_each_resv_unavail_range(i, p_start, p_end)
Which iterates through reserved && !memory lists, and we zero struct pages
explicitly by calling mm_zero_struct_page().
===
Here is more detailed example of problem that this patch is addressing:
Run tested on qemu with the following arguments:
-enable-kvm -cpu kvm64 -m 512 -smp 2
This patch reports that there are 98 unavailable pages.
They are: pfn 0 and pfns in range [159, 255].
Note, trim_low_memory_range() reserves only pfns in range [0, 15], it does
not reserve [159, 255] ones.
e820__memblock_setup() reports linux that the following physical ranges are
available:
[1 , 158]
[256, 130783]
Notice, that exactly unavailable pfns are missing!
Now, lets check what we have in zone 0: [1, 131039]
pfn 0, is not part of the zone, but pfns [1, 158], are.
However, the bigger problem we have if we do not initialize these struct
pages is with memory hotplug. Because, that path operates at 2M
boundaries (section_nr). And checks if 2M range of pages is hot
removable. It starts with first pfn from zone, rounds it down to 2M
boundary (sturct pages are allocated at 2M boundaries when vmemmap is
created), and checks if that section is hot removable. In this case
start with pfn 1 and convert it down to pfn 0. Later pfn is converted
to struct page, and some fields are checked. Now, if we do not zero
struct pages, we get unpredictable results.
In fact when CONFIG_VM_DEBUG is enabled, and we explicitly set all
vmemmap memory to ones, the following panic is observed with kernel test
without this patch applied:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: is_pageblock_removable_nolock+0x35/0x90
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT
...
task: ffff88001f4e2900 task.stack: ffffc90000314000
RIP: 0010:is_pageblock_removable_nolock+0x35/0x90
Call Trace:
? is_mem_section_removable+0x5a/0xd0
show_mem_removable+0x6b/0xa0
dev_attr_show+0x1b/0x50
sysfs_kf_seq_show+0xa1/0x100
kernfs_seq_show+0x22/0x30
seq_read+0x1ac/0x3a0
kernfs_fop_read+0x36/0x190
? security_file_permission+0x90/0xb0
__vfs_read+0x16/0x30
vfs_read+0x81/0x130
SyS_read+0x44/0xa0
entry_SYSCALL_64_fastpath+0x1f/0xbd
Link: http://lkml.kernel.org/r/20171013173214.27300-7-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Reviewed-by: Steven Sistare <steven.sistare@oracle.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Reviewed-by: Bob Picco <bob.picco@oracle.com>
Tested-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-16 01:36:31 +00:00
|
|
|
|
2007-05-23 20:57:55 +00:00
|
|
|
#if MAX_NUMNODES > 1
|
|
|
|
/*
|
|
|
|
* Figure out the number of possible node ids.
|
|
|
|
*/
|
2013-04-29 22:08:01 +00:00
|
|
|
void __init setup_nr_node_ids(void)
|
2007-05-23 20:57:55 +00:00
|
|
|
{
|
2015-09-08 21:59:48 +00:00
|
|
|
unsigned int highest;
|
2007-05-23 20:57:55 +00:00
|
|
|
|
2015-09-08 21:59:48 +00:00
|
|
|
highest = find_last_bit(node_possible_map.bits, MAX_NUMNODES);
|
2007-05-23 20:57:55 +00:00
|
|
|
nr_node_ids = highest + 1;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
x86, numa: Implement pfn -> nid mapping granularity check
SPARSEMEM w/o VMEMMAP and DISCONTIGMEM, both used only on 32bit, use
sections array to map pfn to nid which is limited in granularity. If
NUMA nodes are laid out such that the mapping cannot be accurate, boot
will fail triggering BUG_ON() in mminit_verify_page_links().
On 32bit, it's 512MiB w/ PAE and SPARSEMEM. This seems to have been
granular enough until commit 2706a0bf7b (x86, NUMA: Enable
CONFIG_AMD_NUMA on 32bit too). Apparently, there is a machine which
aligns NUMA nodes to 128MiB and has only AMD NUMA but not SRAT. This
led to the following BUG_ON().
On node 0 totalpages: 2096615
DMA zone: 32 pages used for memmap
DMA zone: 0 pages reserved
DMA zone: 3927 pages, LIFO batch:0
Normal zone: 1740 pages used for memmap
Normal zone: 220978 pages, LIFO batch:31
HighMem zone: 16405 pages used for memmap
HighMem zone: 1853533 pages, LIFO batch:31
BUG: Int 6: CR2 (null)
EDI (null) ESI 00000002 EBP 00000002 ESP c1543ecc
EBX f2400000 EDX 00000006 ECX (null) EAX 00000001
err (null) EIP c16209aa CS 00000060 flg 00010002
Stack: f2400000 00220000 f7200800 c1620613 00220000 01000000 04400000 00238000
(null) f7200000 00000002 f7200b58 f7200800 c1620929 000375fe (null)
f7200b80 c16395f0 00200a02 f7200a80 (null) 000375fe 00000002 (null)
Pid: 0, comm: swapper Not tainted 2.6.39-rc5-00181-g2706a0b #17
Call Trace:
[<c136b1e5>] ? early_fault+0x2e/0x2e
[<c16209aa>] ? mminit_verify_page_links+0x12/0x42
[<c1620613>] ? memmap_init_zone+0xaf/0x10c
[<c1620929>] ? free_area_init_node+0x2b9/0x2e3
[<c1607e99>] ? free_area_init_nodes+0x3f2/0x451
[<c1601d80>] ? paging_init+0x112/0x118
[<c15f578d>] ? setup_arch+0x791/0x82f
[<c15f43d9>] ? start_kernel+0x6a/0x257
This patch implements node_map_pfn_alignment() which determines
maximum internode alignment and update numa_register_memblks() to
reject NUMA configuration if alignment exceeds the pfn -> nid mapping
granularity of the memory model as determined by PAGES_PER_SECTION.
This makes the problematic machine boot w/ flatmem by rejecting the
NUMA config and provides protection against crazy NUMA configurations.
Signed-off-by: Tejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/20110712074534.GB2872@htj.dyndns.org
LKML-Reference: <20110628174613.GP478@escobedo.osrc.amd.com>
Reported-and-Tested-by: Hans Rosenfeld <hans.rosenfeld@amd.com>
Cc: Conny Seidel <conny.seidel@amd.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2011-07-12 07:45:34 +00:00
|
|
|
/**
|
|
|
|
* node_map_pfn_alignment - determine the maximum internode alignment
|
|
|
|
*
|
|
|
|
* This function should be called after node map is populated and sorted.
|
|
|
|
* It calculates the maximum power of two alignment which can distinguish
|
|
|
|
* all the nodes.
|
|
|
|
*
|
|
|
|
* For example, if all nodes are 1GiB and aligned to 1GiB, the return value
|
|
|
|
* would indicate 1GiB alignment with (1 << (30 - PAGE_SHIFT)). If the
|
|
|
|
* nodes are shifted by 256MiB, 256MiB. Note that if only the last node is
|
|
|
|
* shifted, 1GiB is enough and this function will indicate so.
|
|
|
|
*
|
|
|
|
* This is used to test whether pfn -> nid mapping of the chosen memory
|
|
|
|
* model has fine enough granularity to avoid incorrect mapping for the
|
|
|
|
* populated node map.
|
|
|
|
*
|
2019-03-05 23:48:42 +00:00
|
|
|
* Return: the determined alignment in pfn's. 0 if there is no alignment
|
x86, numa: Implement pfn -> nid mapping granularity check
SPARSEMEM w/o VMEMMAP and DISCONTIGMEM, both used only on 32bit, use
sections array to map pfn to nid which is limited in granularity. If
NUMA nodes are laid out such that the mapping cannot be accurate, boot
will fail triggering BUG_ON() in mminit_verify_page_links().
On 32bit, it's 512MiB w/ PAE and SPARSEMEM. This seems to have been
granular enough until commit 2706a0bf7b (x86, NUMA: Enable
CONFIG_AMD_NUMA on 32bit too). Apparently, there is a machine which
aligns NUMA nodes to 128MiB and has only AMD NUMA but not SRAT. This
led to the following BUG_ON().
On node 0 totalpages: 2096615
DMA zone: 32 pages used for memmap
DMA zone: 0 pages reserved
DMA zone: 3927 pages, LIFO batch:0
Normal zone: 1740 pages used for memmap
Normal zone: 220978 pages, LIFO batch:31
HighMem zone: 16405 pages used for memmap
HighMem zone: 1853533 pages, LIFO batch:31
BUG: Int 6: CR2 (null)
EDI (null) ESI 00000002 EBP 00000002 ESP c1543ecc
EBX f2400000 EDX 00000006 ECX (null) EAX 00000001
err (null) EIP c16209aa CS 00000060 flg 00010002
Stack: f2400000 00220000 f7200800 c1620613 00220000 01000000 04400000 00238000
(null) f7200000 00000002 f7200b58 f7200800 c1620929 000375fe (null)
f7200b80 c16395f0 00200a02 f7200a80 (null) 000375fe 00000002 (null)
Pid: 0, comm: swapper Not tainted 2.6.39-rc5-00181-g2706a0b #17
Call Trace:
[<c136b1e5>] ? early_fault+0x2e/0x2e
[<c16209aa>] ? mminit_verify_page_links+0x12/0x42
[<c1620613>] ? memmap_init_zone+0xaf/0x10c
[<c1620929>] ? free_area_init_node+0x2b9/0x2e3
[<c1607e99>] ? free_area_init_nodes+0x3f2/0x451
[<c1601d80>] ? paging_init+0x112/0x118
[<c15f578d>] ? setup_arch+0x791/0x82f
[<c15f43d9>] ? start_kernel+0x6a/0x257
This patch implements node_map_pfn_alignment() which determines
maximum internode alignment and update numa_register_memblks() to
reject NUMA configuration if alignment exceeds the pfn -> nid mapping
granularity of the memory model as determined by PAGES_PER_SECTION.
This makes the problematic machine boot w/ flatmem by rejecting the
NUMA config and provides protection against crazy NUMA configurations.
Signed-off-by: Tejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/20110712074534.GB2872@htj.dyndns.org
LKML-Reference: <20110628174613.GP478@escobedo.osrc.amd.com>
Reported-and-Tested-by: Hans Rosenfeld <hans.rosenfeld@amd.com>
Cc: Conny Seidel <conny.seidel@amd.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2011-07-12 07:45:34 +00:00
|
|
|
* requirement (single node).
|
|
|
|
*/
|
|
|
|
unsigned long __init node_map_pfn_alignment(void)
|
|
|
|
{
|
|
|
|
unsigned long accl_mask = 0, last_end = 0;
|
2011-07-12 08:46:30 +00:00
|
|
|
unsigned long start, end, mask;
|
2019-03-05 23:42:58 +00:00
|
|
|
int last_nid = NUMA_NO_NODE;
|
2011-07-12 08:46:30 +00:00
|
|
|
int i, nid;
|
x86, numa: Implement pfn -> nid mapping granularity check
SPARSEMEM w/o VMEMMAP and DISCONTIGMEM, both used only on 32bit, use
sections array to map pfn to nid which is limited in granularity. If
NUMA nodes are laid out such that the mapping cannot be accurate, boot
will fail triggering BUG_ON() in mminit_verify_page_links().
On 32bit, it's 512MiB w/ PAE and SPARSEMEM. This seems to have been
granular enough until commit 2706a0bf7b (x86, NUMA: Enable
CONFIG_AMD_NUMA on 32bit too). Apparently, there is a machine which
aligns NUMA nodes to 128MiB and has only AMD NUMA but not SRAT. This
led to the following BUG_ON().
On node 0 totalpages: 2096615
DMA zone: 32 pages used for memmap
DMA zone: 0 pages reserved
DMA zone: 3927 pages, LIFO batch:0
Normal zone: 1740 pages used for memmap
Normal zone: 220978 pages, LIFO batch:31
HighMem zone: 16405 pages used for memmap
HighMem zone: 1853533 pages, LIFO batch:31
BUG: Int 6: CR2 (null)
EDI (null) ESI 00000002 EBP 00000002 ESP c1543ecc
EBX f2400000 EDX 00000006 ECX (null) EAX 00000001
err (null) EIP c16209aa CS 00000060 flg 00010002
Stack: f2400000 00220000 f7200800 c1620613 00220000 01000000 04400000 00238000
(null) f7200000 00000002 f7200b58 f7200800 c1620929 000375fe (null)
f7200b80 c16395f0 00200a02 f7200a80 (null) 000375fe 00000002 (null)
Pid: 0, comm: swapper Not tainted 2.6.39-rc5-00181-g2706a0b #17
Call Trace:
[<c136b1e5>] ? early_fault+0x2e/0x2e
[<c16209aa>] ? mminit_verify_page_links+0x12/0x42
[<c1620613>] ? memmap_init_zone+0xaf/0x10c
[<c1620929>] ? free_area_init_node+0x2b9/0x2e3
[<c1607e99>] ? free_area_init_nodes+0x3f2/0x451
[<c1601d80>] ? paging_init+0x112/0x118
[<c15f578d>] ? setup_arch+0x791/0x82f
[<c15f43d9>] ? start_kernel+0x6a/0x257
This patch implements node_map_pfn_alignment() which determines
maximum internode alignment and update numa_register_memblks() to
reject NUMA configuration if alignment exceeds the pfn -> nid mapping
granularity of the memory model as determined by PAGES_PER_SECTION.
This makes the problematic machine boot w/ flatmem by rejecting the
NUMA config and provides protection against crazy NUMA configurations.
Signed-off-by: Tejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/20110712074534.GB2872@htj.dyndns.org
LKML-Reference: <20110628174613.GP478@escobedo.osrc.amd.com>
Reported-and-Tested-by: Hans Rosenfeld <hans.rosenfeld@amd.com>
Cc: Conny Seidel <conny.seidel@amd.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2011-07-12 07:45:34 +00:00
|
|
|
|
2011-07-12 08:46:30 +00:00
|
|
|
for_each_mem_pfn_range(i, MAX_NUMNODES, &start, &end, &nid) {
|
x86, numa: Implement pfn -> nid mapping granularity check
SPARSEMEM w/o VMEMMAP and DISCONTIGMEM, both used only on 32bit, use
sections array to map pfn to nid which is limited in granularity. If
NUMA nodes are laid out such that the mapping cannot be accurate, boot
will fail triggering BUG_ON() in mminit_verify_page_links().
On 32bit, it's 512MiB w/ PAE and SPARSEMEM. This seems to have been
granular enough until commit 2706a0bf7b (x86, NUMA: Enable
CONFIG_AMD_NUMA on 32bit too). Apparently, there is a machine which
aligns NUMA nodes to 128MiB and has only AMD NUMA but not SRAT. This
led to the following BUG_ON().
On node 0 totalpages: 2096615
DMA zone: 32 pages used for memmap
DMA zone: 0 pages reserved
DMA zone: 3927 pages, LIFO batch:0
Normal zone: 1740 pages used for memmap
Normal zone: 220978 pages, LIFO batch:31
HighMem zone: 16405 pages used for memmap
HighMem zone: 1853533 pages, LIFO batch:31
BUG: Int 6: CR2 (null)
EDI (null) ESI 00000002 EBP 00000002 ESP c1543ecc
EBX f2400000 EDX 00000006 ECX (null) EAX 00000001
err (null) EIP c16209aa CS 00000060 flg 00010002
Stack: f2400000 00220000 f7200800 c1620613 00220000 01000000 04400000 00238000
(null) f7200000 00000002 f7200b58 f7200800 c1620929 000375fe (null)
f7200b80 c16395f0 00200a02 f7200a80 (null) 000375fe 00000002 (null)
Pid: 0, comm: swapper Not tainted 2.6.39-rc5-00181-g2706a0b #17
Call Trace:
[<c136b1e5>] ? early_fault+0x2e/0x2e
[<c16209aa>] ? mminit_verify_page_links+0x12/0x42
[<c1620613>] ? memmap_init_zone+0xaf/0x10c
[<c1620929>] ? free_area_init_node+0x2b9/0x2e3
[<c1607e99>] ? free_area_init_nodes+0x3f2/0x451
[<c1601d80>] ? paging_init+0x112/0x118
[<c15f578d>] ? setup_arch+0x791/0x82f
[<c15f43d9>] ? start_kernel+0x6a/0x257
This patch implements node_map_pfn_alignment() which determines
maximum internode alignment and update numa_register_memblks() to
reject NUMA configuration if alignment exceeds the pfn -> nid mapping
granularity of the memory model as determined by PAGES_PER_SECTION.
This makes the problematic machine boot w/ flatmem by rejecting the
NUMA config and provides protection against crazy NUMA configurations.
Signed-off-by: Tejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/20110712074534.GB2872@htj.dyndns.org
LKML-Reference: <20110628174613.GP478@escobedo.osrc.amd.com>
Reported-and-Tested-by: Hans Rosenfeld <hans.rosenfeld@amd.com>
Cc: Conny Seidel <conny.seidel@amd.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2011-07-12 07:45:34 +00:00
|
|
|
if (!start || last_nid < 0 || last_nid == nid) {
|
|
|
|
last_nid = nid;
|
|
|
|
last_end = end;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Start with a mask granular enough to pin-point to the
|
|
|
|
* start pfn and tick off bits one-by-one until it becomes
|
|
|
|
* too coarse to separate the current node from the last.
|
|
|
|
*/
|
|
|
|
mask = ~((1 << __ffs(start)) - 1);
|
|
|
|
while (mask && last_end <= (start & (mask << 1)))
|
|
|
|
mask <<= 1;
|
|
|
|
|
|
|
|
/* accumulate all internode masks */
|
|
|
|
accl_mask |= mask;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* convert mask to number of pages */
|
|
|
|
return ~accl_mask + 1;
|
|
|
|
}
|
|
|
|
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
/**
|
|
|
|
* find_min_pfn_with_active_regions - Find the minimum PFN registered
|
|
|
|
*
|
2019-03-05 23:48:42 +00:00
|
|
|
* Return: the minimum PFN based on information provided via
|
2014-06-04 23:10:53 +00:00
|
|
|
* memblock_set_node().
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
*/
|
|
|
|
unsigned long __init find_min_pfn_with_active_regions(void)
|
|
|
|
{
|
2020-06-03 22:58:18 +00:00
|
|
|
return PHYS_PFN(memblock_start_of_DRAM());
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
|
|
|
|
2007-10-16 08:25:39 +00:00
|
|
|
/*
|
|
|
|
* early_calculate_totalpages()
|
|
|
|
* Sum pages in active regions for movable zone.
|
2012-12-12 21:51:46 +00:00
|
|
|
* Populate N_MEMORY for calculating usable_nodes.
|
2007-10-16 08:25:39 +00:00
|
|
|
*/
|
2007-10-16 08:26:03 +00:00
|
|
|
static unsigned long __init early_calculate_totalpages(void)
|
2007-07-17 11:03:15 +00:00
|
|
|
{
|
|
|
|
unsigned long totalpages = 0;
|
2011-07-12 08:46:30 +00:00
|
|
|
unsigned long start_pfn, end_pfn;
|
|
|
|
int i, nid;
|
|
|
|
|
|
|
|
for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid) {
|
|
|
|
unsigned long pages = end_pfn - start_pfn;
|
2007-07-17 11:03:15 +00:00
|
|
|
|
2007-10-16 08:25:39 +00:00
|
|
|
totalpages += pages;
|
|
|
|
if (pages)
|
2012-12-12 21:51:46 +00:00
|
|
|
node_set_state(nid, N_MEMORY);
|
2007-10-16 08:25:39 +00:00
|
|
|
}
|
2013-09-11 21:20:34 +00:00
|
|
|
return totalpages;
|
2007-07-17 11:03:15 +00:00
|
|
|
}
|
|
|
|
|
2007-07-17 11:03:12 +00:00
|
|
|
/*
|
|
|
|
* Find the PFN the Movable zone begins in each node. Kernel memory
|
|
|
|
* is spread evenly between nodes as long as the nodes have enough
|
|
|
|
* memory. When they don't, some nodes will have more kernelcore than
|
|
|
|
* others
|
|
|
|
*/
|
2012-03-21 23:34:15 +00:00
|
|
|
static void __init find_zone_movable_pfns_for_nodes(void)
|
2007-07-17 11:03:12 +00:00
|
|
|
{
|
|
|
|
int i, nid;
|
|
|
|
unsigned long usable_startpfn;
|
|
|
|
unsigned long kernelcore_node, kernelcore_remaining;
|
2009-06-30 18:41:37 +00:00
|
|
|
/* save the state before borrow the nodemask */
|
2012-12-12 21:51:46 +00:00
|
|
|
nodemask_t saved_node_state = node_states[N_MEMORY];
|
2007-10-16 08:25:39 +00:00
|
|
|
unsigned long totalpages = early_calculate_totalpages();
|
2012-12-12 21:51:46 +00:00
|
|
|
int usable_nodes = nodes_weight(node_states[N_MEMORY]);
|
2014-04-07 22:37:52 +00:00
|
|
|
struct memblock_region *r;
|
2014-01-21 23:49:38 +00:00
|
|
|
|
|
|
|
/* Need to find movable_zone earlier when movable_node is specified. */
|
|
|
|
find_usable_zone_for_movable();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If movable_node is specified, ignore kernelcore and movablecore
|
|
|
|
* options.
|
|
|
|
*/
|
|
|
|
if (movable_node_is_enabled()) {
|
2020-10-13 23:58:30 +00:00
|
|
|
for_each_mem_region(r) {
|
2014-04-07 22:37:52 +00:00
|
|
|
if (!memblock_is_hotpluggable(r))
|
2014-01-21 23:49:38 +00:00
|
|
|
continue;
|
|
|
|
|
2020-06-03 22:56:53 +00:00
|
|
|
nid = memblock_get_region_node(r);
|
2014-01-21 23:49:38 +00:00
|
|
|
|
2014-04-07 22:37:52 +00:00
|
|
|
usable_startpfn = PFN_DOWN(r->base);
|
2014-01-21 23:49:38 +00:00
|
|
|
zone_movable_pfn[nid] = zone_movable_pfn[nid] ?
|
|
|
|
min(usable_startpfn, zone_movable_pfn[nid]) :
|
|
|
|
usable_startpfn;
|
|
|
|
}
|
|
|
|
|
|
|
|
goto out2;
|
|
|
|
}
|
2007-07-17 11:03:12 +00:00
|
|
|
|
2016-03-15 21:55:22 +00:00
|
|
|
/*
|
|
|
|
* If kernelcore=mirror is specified, ignore movablecore option
|
|
|
|
*/
|
|
|
|
if (mirrored_kernelcore) {
|
|
|
|
bool mem_below_4gb_not_mirrored = false;
|
|
|
|
|
2020-10-13 23:58:30 +00:00
|
|
|
for_each_mem_region(r) {
|
2016-03-15 21:55:22 +00:00
|
|
|
if (memblock_is_mirror(r))
|
|
|
|
continue;
|
|
|
|
|
2020-06-03 22:56:53 +00:00
|
|
|
nid = memblock_get_region_node(r);
|
2016-03-15 21:55:22 +00:00
|
|
|
|
|
|
|
usable_startpfn = memblock_region_memory_base_pfn(r);
|
|
|
|
|
|
|
|
if (usable_startpfn < 0x100000) {
|
|
|
|
mem_below_4gb_not_mirrored = true;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
zone_movable_pfn[nid] = zone_movable_pfn[nid] ?
|
|
|
|
min(usable_startpfn, zone_movable_pfn[nid]) :
|
|
|
|
usable_startpfn;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (mem_below_4gb_not_mirrored)
|
2020-06-03 23:00:02 +00:00
|
|
|
pr_warn("This configuration results in unmirrored kernel memory.\n");
|
2016-03-15 21:55:22 +00:00
|
|
|
|
|
|
|
goto out2;
|
|
|
|
}
|
|
|
|
|
2007-07-17 11:03:15 +00:00
|
|
|
/*
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
* If kernelcore=nn% or movablecore=nn% was specified, calculate the
|
|
|
|
* amount of necessary memory.
|
|
|
|
*/
|
|
|
|
if (required_kernelcore_percent)
|
|
|
|
required_kernelcore = (totalpages * 100 * required_kernelcore_percent) /
|
|
|
|
10000UL;
|
|
|
|
if (required_movablecore_percent)
|
|
|
|
required_movablecore = (totalpages * 100 * required_movablecore_percent) /
|
|
|
|
10000UL;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If movablecore= was specified, calculate what size of
|
2007-07-17 11:03:15 +00:00
|
|
|
* kernelcore that corresponds so that memory usable for
|
|
|
|
* any allocation type is evenly spread. If both kernelcore
|
|
|
|
* and movablecore are specified, then the value of kernelcore
|
|
|
|
* will be used for required_kernelcore if it's greater than
|
|
|
|
* what movablecore would have allowed.
|
|
|
|
*/
|
|
|
|
if (required_movablecore) {
|
|
|
|
unsigned long corepages;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Round-up so that ZONE_MOVABLE is at least as large as what
|
|
|
|
* was requested by the user
|
|
|
|
*/
|
|
|
|
required_movablecore =
|
|
|
|
roundup(required_movablecore, MAX_ORDER_NR_PAGES);
|
2015-11-06 02:48:11 +00:00
|
|
|
required_movablecore = min(totalpages, required_movablecore);
|
2007-07-17 11:03:15 +00:00
|
|
|
corepages = totalpages - required_movablecore;
|
|
|
|
|
|
|
|
required_kernelcore = max(required_kernelcore, corepages);
|
|
|
|
}
|
|
|
|
|
2015-11-06 02:48:56 +00:00
|
|
|
/*
|
|
|
|
* If kernelcore was not specified or kernelcore size is larger
|
|
|
|
* than totalpages, there is no ZONE_MOVABLE.
|
|
|
|
*/
|
|
|
|
if (!required_kernelcore || required_kernelcore >= totalpages)
|
2009-06-30 18:41:37 +00:00
|
|
|
goto out;
|
2007-07-17 11:03:12 +00:00
|
|
|
|
|
|
|
/* usable_startpfn is the lowest possible pfn ZONE_MOVABLE can be at */
|
|
|
|
usable_startpfn = arch_zone_lowest_possible_pfn[movable_zone];
|
|
|
|
|
|
|
|
restart:
|
|
|
|
/* Spread kernelcore memory as evenly as possible throughout nodes */
|
|
|
|
kernelcore_node = required_kernelcore / usable_nodes;
|
2012-12-12 21:51:46 +00:00
|
|
|
for_each_node_state(nid, N_MEMORY) {
|
2011-07-12 08:46:30 +00:00
|
|
|
unsigned long start_pfn, end_pfn;
|
|
|
|
|
2007-07-17 11:03:12 +00:00
|
|
|
/*
|
|
|
|
* Recalculate kernelcore_node if the division per node
|
|
|
|
* now exceeds what is necessary to satisfy the requested
|
|
|
|
* amount of memory for the kernel
|
|
|
|
*/
|
|
|
|
if (required_kernelcore < kernelcore_node)
|
|
|
|
kernelcore_node = required_kernelcore / usable_nodes;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* As the map is walked, we track how much memory is usable
|
|
|
|
* by the kernel using kernelcore_remaining. When it is
|
|
|
|
* 0, the rest of the node is usable by ZONE_MOVABLE
|
|
|
|
*/
|
|
|
|
kernelcore_remaining = kernelcore_node;
|
|
|
|
|
|
|
|
/* Go through each range of PFNs within this node */
|
2011-07-12 08:46:30 +00:00
|
|
|
for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
|
2007-07-17 11:03:12 +00:00
|
|
|
unsigned long size_pages;
|
|
|
|
|
2011-07-12 08:46:30 +00:00
|
|
|
start_pfn = max(start_pfn, zone_movable_pfn[nid]);
|
2007-07-17 11:03:12 +00:00
|
|
|
if (start_pfn >= end_pfn)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/* Account for what is only usable for kernelcore */
|
|
|
|
if (start_pfn < usable_startpfn) {
|
|
|
|
unsigned long kernel_pages;
|
|
|
|
kernel_pages = min(end_pfn, usable_startpfn)
|
|
|
|
- start_pfn;
|
|
|
|
|
|
|
|
kernelcore_remaining -= min(kernel_pages,
|
|
|
|
kernelcore_remaining);
|
|
|
|
required_kernelcore -= min(kernel_pages,
|
|
|
|
required_kernelcore);
|
|
|
|
|
|
|
|
/* Continue if range is now fully accounted */
|
|
|
|
if (end_pfn <= usable_startpfn) {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Push zone_movable_pfn to the end so
|
|
|
|
* that if we have to rebalance
|
|
|
|
* kernelcore across nodes, we will
|
|
|
|
* not double account here
|
|
|
|
*/
|
|
|
|
zone_movable_pfn[nid] = end_pfn;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
start_pfn = usable_startpfn;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The usable PFN range for ZONE_MOVABLE is from
|
|
|
|
* start_pfn->end_pfn. Calculate size_pages as the
|
|
|
|
* number of pages used as kernelcore
|
|
|
|
*/
|
|
|
|
size_pages = end_pfn - start_pfn;
|
|
|
|
if (size_pages > kernelcore_remaining)
|
|
|
|
size_pages = kernelcore_remaining;
|
|
|
|
zone_movable_pfn[nid] = start_pfn + size_pages;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Some kernelcore has been met, update counts and
|
|
|
|
* break if the kernelcore for this node has been
|
2013-09-11 21:20:34 +00:00
|
|
|
* satisfied
|
2007-07-17 11:03:12 +00:00
|
|
|
*/
|
|
|
|
required_kernelcore -= min(required_kernelcore,
|
|
|
|
size_pages);
|
|
|
|
kernelcore_remaining -= size_pages;
|
|
|
|
if (!kernelcore_remaining)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If there is still required_kernelcore, we do another pass with one
|
|
|
|
* less node in the count. This will push zone_movable_pfn[nid] further
|
|
|
|
* along on the nodes that still have memory until kernelcore is
|
2013-09-11 21:20:34 +00:00
|
|
|
* satisfied
|
2007-07-17 11:03:12 +00:00
|
|
|
*/
|
|
|
|
usable_nodes--;
|
|
|
|
if (usable_nodes && required_kernelcore > usable_nodes)
|
|
|
|
goto restart;
|
|
|
|
|
2014-01-21 23:49:38 +00:00
|
|
|
out2:
|
2007-07-17 11:03:12 +00:00
|
|
|
/* Align start of ZONE_MOVABLE on all nids to MAX_ORDER_NR_PAGES */
|
|
|
|
for (nid = 0; nid < MAX_NUMNODES; nid++)
|
|
|
|
zone_movable_pfn[nid] =
|
|
|
|
roundup(zone_movable_pfn[nid], MAX_ORDER_NR_PAGES);
|
2009-06-30 18:41:37 +00:00
|
|
|
|
x86, ACPI, mm: Revert movablemem_map support
Tim found:
WARNING: at arch/x86/kernel/smpboot.c:324 topology_sane.isra.2+0x6f/0x80()
Hardware name: S2600CP
sched: CPU #1's llc-sibling CPU #0 is not on the same node! [node: 1 != 0]. Ignoring dependency.
smpboot: Booting Node 1, Processors #1
Modules linked in:
Pid: 0, comm: swapper/1 Not tainted 3.9.0-0-generic #1
Call Trace:
set_cpu_sibling_map+0x279/0x449
start_secondary+0x11d/0x1e5
Don Morris reproduced on a HP z620 workstation, and bisected it to
commit e8d195525809 ("acpi, memory-hotplug: parse SRAT before memblock
is ready")
It turns out movable_map has some problems, and it breaks several things
1. numa_init is called several times, NOT just for srat. so those
nodes_clear(numa_nodes_parsed)
memset(&numa_meminfo, 0, sizeof(numa_meminfo))
can not be just removed. Need to consider sequence is: numaq, srat, amd, dummy.
and make fall back path working.
2. simply split acpi_numa_init to early_parse_srat.
a. that early_parse_srat is NOT called for ia64, so you break ia64.
b. for (i = 0; i < MAX_LOCAL_APIC; i++)
set_apicid_to_node(i, NUMA_NO_NODE)
still left in numa_init. So it will just clear result from early_parse_srat.
it should be moved before that....
c. it breaks ACPI_TABLE_OVERIDE...as the acpi table scan is moved
early before override from INITRD is settled.
3. that patch TITLE is total misleading, there is NO x86 in the title,
but it changes critical x86 code. It caused x86 guys did not
pay attention to find the problem early. Those patches really should
be routed via tip/x86/mm.
4. after that commit, following range can not use movable ram:
a. real_mode code.... well..funny, legacy Node0 [0,1M) could be hot-removed?
b. initrd... it will be freed after booting, so it could be on movable...
c. crashkernel for kdump...: looks like we can not put kdump kernel above 4G
anymore.
d. init_mem_mapping: can not put page table high anymore.
e. initmem_init: vmemmap can not be high local node anymore. That is
not good.
If node is hotplugable, the mem related range like page table and
vmemmap could be on the that node without problem and should be on that
node.
We have workaround patch that could fix some problems, but some can not
be fixed.
So just remove that offending commit and related ones including:
f7210e6c4ac7 ("mm/memblock.c: use CONFIG_HAVE_MEMBLOCK_NODE_MAP to
protect movablecore_map in memblock_overlaps_region().")
01a178a94e8e ("acpi, memory-hotplug: support getting hotplug info from
SRAT")
27168d38fa20 ("acpi, memory-hotplug: extend movablemem_map ranges to
the end of node")
e8d195525809 ("acpi, memory-hotplug: parse SRAT before memblock is
ready")
fb06bc8e5f42 ("page_alloc: bootmem limit with movablecore_map")
42f47e27e761 ("page_alloc: make movablemem_map have higher priority")
6981ec31146c ("page_alloc: introduce zone_movable_limit[] to keep
movable limit for nodes")
34b71f1e04fc ("page_alloc: add movable_memmap kernel parameter")
4d59a75125d5 ("x86: get pg_data_t's memory from other node")
Later we should have patches that will make sure kernel put page table
and vmemmap on local node ram instead of push them down to node0. Also
need to find way to put other kernel used ram to local node ram.
Reported-by: Tim Gardner <tim.gardner@canonical.com>
Reported-by: Don Morris <don.morris@hp.com>
Bisected-by: Don Morris <don.morris@hp.com>
Tested-by: Don Morris <don.morris@hp.com>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Thomas Renninger <trenn@suse.de>
Cc: Tejun Heo <tj@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-03-01 22:51:27 +00:00
|
|
|
out:
|
2009-06-30 18:41:37 +00:00
|
|
|
/* restore the node_state */
|
2012-12-12 21:51:46 +00:00
|
|
|
node_states[N_MEMORY] = saved_node_state;
|
2007-07-17 11:03:12 +00:00
|
|
|
}
|
|
|
|
|
2012-12-12 21:51:46 +00:00
|
|
|
/* Any regular or high memory on that node ? */
|
|
|
|
static void check_for_memory(pg_data_t *pgdat, int nid)
|
2007-10-16 08:25:39 +00:00
|
|
|
{
|
|
|
|
enum zone_type zone_type;
|
|
|
|
|
2012-12-12 21:51:46 +00:00
|
|
|
for (zone_type = 0; zone_type <= ZONE_MOVABLE - 1; zone_type++) {
|
2007-10-16 08:25:39 +00:00
|
|
|
struct zone *zone = &pgdat->node_zones[zone_type];
|
2013-11-12 23:07:20 +00:00
|
|
|
if (populated_zone(zone)) {
|
mm/page_alloc.c: clean up check_for_memory()
check_for_memory() looks a bit confusing. First of all, we have this:
if (N_MEMORY == N_NORMAL_MEMORY)
return;
Checking the ENUM declaration, looks like N_MEMORY canot be equal to
N_NORMAL_MEMORY.
I could not find where N_MEMORY is set to N_NORMAL_MEMORY, or the other
way around either, so unless I am missing something, this condition will
never evaluate to true. It makes sense to get rid of it.
Moving forward, the operations within the loop look a bit confusing as
well.
We set N_HIGH_MEMORY unconditionally, and then we set N_NORMAL_MEMORY in
case we have CONFIG_HIGHMEM (N_NORMAL_MEMORY != N_HIGH_MEMORY) and zone <=
ZONE_NORMAL. (N_HIGH_MEMORY falls back to N_NORMAL_MEMORY on
!CONFIG_HIGHMEM systems, and that is why we can just go ahead and set
N_HIGH_MEMORY unconditionally)
Although this works, it is a bit subtle.
I think that this could be easier to follow:
First, we should only set N_HIGH_MEMORY in case we have CONFIG_HIGHMEM.
And then we should set N_NORMAL_MEMORY in case zone <= ZONE_NORMAL,
without further checking whether we have CONFIG_HIGHMEM or not.
Link: http://lkml.kernel.org/r/20180828210158.4617-1-osalvador@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Michael Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 22:03:58 +00:00
|
|
|
if (IS_ENABLED(CONFIG_HIGHMEM))
|
|
|
|
node_set_state(nid, N_HIGH_MEMORY);
|
|
|
|
if (zone_type <= ZONE_NORMAL)
|
2012-12-12 21:51:46 +00:00
|
|
|
node_set_state(nid, N_NORMAL_MEMORY);
|
2012-01-13 01:19:07 +00:00
|
|
|
break;
|
|
|
|
}
|
2007-10-16 08:25:39 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-06-03 22:58:03 +00:00
|
|
|
/*
|
|
|
|
* Some architecturs, e.g. ARC may have ZONE_HIGHMEM below ZONE_NORMAL. For
|
|
|
|
* such cases we allow max_zone_pfn sorted in the descending order
|
|
|
|
*/
|
|
|
|
bool __weak arch_has_descending_max_zone_pfns(void)
|
|
|
|
{
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
/**
|
2020-06-03 22:57:10 +00:00
|
|
|
* free_area_init - Initialise all pg_data_t and zone data
|
2006-10-04 09:15:25 +00:00
|
|
|
* @max_zone_pfn: an array of max PFNs for each zone
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
*
|
|
|
|
* This will call free_area_init_node() for each active node in the system.
|
2014-06-04 23:10:53 +00:00
|
|
|
* Using the page ranges provided by memblock_set_node(), the size of each
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
* zone in each node and their holes is calculated. If the maximum PFN
|
|
|
|
* between two adjacent zones match, it is assumed that the zone is empty.
|
|
|
|
* For example, if arch_max_dma_pfn == arch_max_dma32_pfn, it is assumed
|
|
|
|
* that arch_max_dma32_pfn has no pages. It is also assumed that a zone
|
|
|
|
* starts where the previous one ended. For example, ZONE_DMA32 starts
|
|
|
|
* at arch_max_dma_pfn.
|
|
|
|
*/
|
2020-06-03 22:57:10 +00:00
|
|
|
void __init free_area_init(unsigned long *max_zone_pfn)
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
{
|
2011-07-12 08:46:30 +00:00
|
|
|
unsigned long start_pfn, end_pfn;
|
2020-06-03 22:58:03 +00:00
|
|
|
int i, nid, zone;
|
|
|
|
bool descending;
|
2007-02-10 09:42:57 +00:00
|
|
|
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
/* Record where the zone boundaries are */
|
|
|
|
memset(arch_zone_lowest_possible_pfn, 0,
|
|
|
|
sizeof(arch_zone_lowest_possible_pfn));
|
|
|
|
memset(arch_zone_highest_possible_pfn, 0,
|
|
|
|
sizeof(arch_zone_highest_possible_pfn));
|
2016-07-26 22:22:17 +00:00
|
|
|
|
|
|
|
start_pfn = find_min_pfn_with_active_regions();
|
2020-06-03 22:58:03 +00:00
|
|
|
descending = arch_has_descending_max_zone_pfns();
|
2016-07-26 22:22:17 +00:00
|
|
|
|
|
|
|
for (i = 0; i < MAX_NR_ZONES; i++) {
|
2020-06-03 22:58:03 +00:00
|
|
|
if (descending)
|
|
|
|
zone = MAX_NR_ZONES - i - 1;
|
|
|
|
else
|
|
|
|
zone = i;
|
|
|
|
|
|
|
|
if (zone == ZONE_MOVABLE)
|
2007-07-17 11:03:12 +00:00
|
|
|
continue;
|
2016-07-26 22:22:17 +00:00
|
|
|
|
2020-06-03 22:58:03 +00:00
|
|
|
end_pfn = max(max_zone_pfn[zone], start_pfn);
|
|
|
|
arch_zone_lowest_possible_pfn[zone] = start_pfn;
|
|
|
|
arch_zone_highest_possible_pfn[zone] = end_pfn;
|
2016-07-26 22:22:17 +00:00
|
|
|
|
|
|
|
start_pfn = end_pfn;
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
2007-07-17 11:03:12 +00:00
|
|
|
|
|
|
|
/* Find the PFNs that ZONE_MOVABLE begins at in each node */
|
|
|
|
memset(zone_movable_pfn, 0, sizeof(zone_movable_pfn));
|
2012-03-21 23:34:15 +00:00
|
|
|
find_zone_movable_pfns_for_nodes();
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
|
|
|
|
/* Print out the zone ranges */
|
2014-12-10 23:42:53 +00:00
|
|
|
pr_info("Zone ranges:\n");
|
2007-07-17 11:03:12 +00:00
|
|
|
for (i = 0; i < MAX_NR_ZONES; i++) {
|
|
|
|
if (i == ZONE_MOVABLE)
|
|
|
|
continue;
|
2014-12-10 23:42:53 +00:00
|
|
|
pr_info(" %-8s ", zone_names[i]);
|
2010-03-05 21:42:14 +00:00
|
|
|
if (arch_zone_lowest_possible_pfn[i] ==
|
|
|
|
arch_zone_highest_possible_pfn[i])
|
2014-12-10 23:42:53 +00:00
|
|
|
pr_cont("empty\n");
|
2010-03-05 21:42:14 +00:00
|
|
|
else
|
2015-02-11 23:26:01 +00:00
|
|
|
pr_cont("[mem %#018Lx-%#018Lx]\n",
|
|
|
|
(u64)arch_zone_lowest_possible_pfn[i]
|
|
|
|
<< PAGE_SHIFT,
|
|
|
|
((u64)arch_zone_highest_possible_pfn[i]
|
2012-05-29 22:06:30 +00:00
|
|
|
<< PAGE_SHIFT) - 1);
|
2007-07-17 11:03:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Print out the PFNs ZONE_MOVABLE begins at in each node */
|
2014-12-10 23:42:53 +00:00
|
|
|
pr_info("Movable zone start for each node\n");
|
2007-07-17 11:03:12 +00:00
|
|
|
for (i = 0; i < MAX_NUMNODES; i++) {
|
|
|
|
if (zone_movable_pfn[i])
|
2015-02-11 23:26:01 +00:00
|
|
|
pr_info(" Node %d: %#018Lx\n", i,
|
|
|
|
(u64)zone_movable_pfn[i] << PAGE_SHIFT);
|
2007-07-17 11:03:12 +00:00
|
|
|
}
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
|
2019-07-18 22:58:04 +00:00
|
|
|
/*
|
|
|
|
* Print out the early node map, and initialize the
|
|
|
|
* subsection-map relative to active online memory ranges to
|
|
|
|
* enable future "sub-section" extensions of the memory map.
|
|
|
|
*/
|
2014-12-10 23:42:53 +00:00
|
|
|
pr_info("Early memory node ranges\n");
|
2019-07-18 22:58:04 +00:00
|
|
|
for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid) {
|
2015-02-11 23:26:01 +00:00
|
|
|
pr_info(" node %3d: [mem %#018Lx-%#018Lx]\n", nid,
|
|
|
|
(u64)start_pfn << PAGE_SHIFT,
|
|
|
|
((u64)end_pfn << PAGE_SHIFT) - 1);
|
2019-07-18 22:58:04 +00:00
|
|
|
subsection_map_init(start_pfn, end_pfn - start_pfn);
|
|
|
|
}
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
|
|
|
|
/* Initialise every node */
|
2008-07-24 04:26:51 +00:00
|
|
|
mminit_verify_pageflags_layout();
|
2007-02-20 21:57:52 +00:00
|
|
|
setup_nr_node_ids();
|
2020-02-04 01:33:55 +00:00
|
|
|
init_unavailable_mem();
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
for_each_online_node(nid) {
|
|
|
|
pg_data_t *pgdat = NODE_DATA(nid);
|
2020-06-03 22:58:13 +00:00
|
|
|
free_area_init_node(nid);
|
2007-10-16 08:25:39 +00:00
|
|
|
|
|
|
|
/* Any memory on that node */
|
|
|
|
if (pgdat->node_present_pages)
|
2012-12-12 21:51:46 +00:00
|
|
|
node_set_state(nid, N_MEMORY);
|
|
|
|
check_for_memory(pgdat, nid);
|
[PATCH] Introduce mechanism for registering active regions of memory
At a basic level, architectures define structures to record where active
ranges of page frames are located. Once located, the code to calculate zone
sizes and holes in each architecture is very similar. Some of this zone and
hole sizing code is difficult to read for no good reason. This set of patches
eliminates the similar-looking architecture-specific code.
The patches introduce a mechanism where architectures register where the
active ranges of page frames are with add_active_range(). When all areas have
been discovered, free_area_init_nodes() is called to initialise the pgdat and
zones. The zone sizes and holes are then calculated in an architecture
independent manner.
Patch 1 introduces the mechanism for registering and initialising PFN ranges
Patch 2 changes ppc to use the mechanism - 139 arch-specific LOC removed
Patch 3 changes x86 to use the mechanism - 136 arch-specific LOC removed
Patch 4 changes x86_64 to use the mechanism - 74 arch-specific LOC removed
Patch 5 changes ia64 to use the mechanism - 52 arch-specific LOC removed
Patch 6 accounts for mem_map as a memory hole as the pages are not reclaimable.
It adjusts the watermarks slightly
Tony Luck has successfully tested for ia64 on Itanium with tiger_defconfig,
gensparse_defconfig and defconfig. Bob Picco has also tested and debugged on
IA64. Jack Steiner successfully boot tested on a mammoth SGI IA64-based
machine. These were on patches against 2.6.17-rc1 and release 3 of these
patches but there have been no ia64-changes since release 3.
There are differences in the zone sizes for x86_64 as the arch-specific code
for x86_64 accounts the kernel image and the starting mem_maps as memory holes
but the architecture-independent code accounts the memory as present.
The big benefit of this set of patches is a sizable reduction of
architecture-specific code, some of which is very hairy. There should be a
greater reduction when other architectures use the same mechanisms for zone
and hole sizing but I lack the hardware to test on.
Additional credit;
Dave Hansen for the initial suggestion and comments on early patches
Andy Whitcroft for reviewing early versions and catching numerous
errors
Tony Luck for testing and debugging on IA64
Bob Picco for fixing bugs related to pfn registration, reviewing a
number of patch revisions, providing a number of suggestions
on future direction and testing heavily
Jack Steiner and Robin Holt for testing on IA64 and clarifying
issues related to memory holes
Yasunori for testing on IA64
Andi Kleen for reviewing and feeding back about x86_64
Christian Kujau for providing valuable information related to ACPI
problems on x86_64 and testing potential fixes
This patch:
Define the structure to represent an active range of page frames within a node
in an architecture independent manner. Architectures are expected to register
active ranges of PFNs using add_active_range(nid, start_pfn, end_pfn) and call
free_area_init_nodes() passing the PFNs of the end of each zone.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "Keith Mannthey" <kmannth@gmail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-27 08:49:43 +00:00
|
|
|
}
|
|
|
|
}
|
2007-07-17 11:03:12 +00:00
|
|
|
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
static int __init cmdline_parse_core(char *p, unsigned long *core,
|
|
|
|
unsigned long *percent)
|
2007-07-17 11:03:12 +00:00
|
|
|
{
|
|
|
|
unsigned long long coremem;
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
char *endptr;
|
|
|
|
|
2007-07-17 11:03:12 +00:00
|
|
|
if (!p)
|
|
|
|
return -EINVAL;
|
|
|
|
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
/* Value may be a percentage of total memory, otherwise bytes */
|
|
|
|
coremem = simple_strtoull(p, &endptr, 0);
|
|
|
|
if (*endptr == '%') {
|
|
|
|
/* Paranoid check for percent values greater than 100 */
|
|
|
|
WARN_ON(coremem > 100);
|
2007-07-17 11:03:12 +00:00
|
|
|
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
*percent = coremem;
|
|
|
|
} else {
|
|
|
|
coremem = memparse(p, &p);
|
|
|
|
/* Paranoid check that UL is enough for the coremem value */
|
|
|
|
WARN_ON((coremem >> PAGE_SHIFT) > ULONG_MAX);
|
2007-07-17 11:03:12 +00:00
|
|
|
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
*core = coremem >> PAGE_SHIFT;
|
|
|
|
*percent = 0UL;
|
|
|
|
}
|
2007-07-17 11:03:12 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2007-07-17 11:03:14 +00:00
|
|
|
|
2007-07-17 11:03:15 +00:00
|
|
|
/*
|
|
|
|
* kernelcore=size sets the amount of memory for use for allocations that
|
|
|
|
* cannot be reclaimed or migrated.
|
|
|
|
*/
|
|
|
|
static int __init cmdline_parse_kernelcore(char *p)
|
|
|
|
{
|
2016-03-15 21:55:22 +00:00
|
|
|
/* parse kernelcore=mirror */
|
|
|
|
if (parse_option_str(p, "mirror")) {
|
|
|
|
mirrored_kernelcore = true;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
return cmdline_parse_core(p, &required_kernelcore,
|
|
|
|
&required_kernelcore_percent);
|
2007-07-17 11:03:15 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* movablecore=size sets the amount of memory for use for allocations that
|
|
|
|
* can be reclaimed or migrated.
|
|
|
|
*/
|
|
|
|
static int __init cmdline_parse_movablecore(char *p)
|
|
|
|
{
|
mm, page_alloc: extend kernelcore and movablecore for percent
Both kernelcore= and movablecore= can be used to define the amount of
ZONE_NORMAL and ZONE_MOVABLE on a system, respectively. This requires
the system memory capacity to be known when specifying the command line,
however.
This introduces the ability to define both kernelcore= and movablecore=
as a percentage of total system memory. This is convenient for systems
software that wants to define the amount of ZONE_MOVABLE, for example,
as a proportion of a system's memory rather than a hardcoded byte value.
To define the percentage, the final character of the parameter should be
a '%'.
mhocko: "why is anyone using these options nowadays?"
rientjes:
:
: Fragmentation of non-__GFP_MOVABLE pages due to low on memory
: situations can pollute most pageblocks on the system, as much as 1GB of
: slab being fragmented over 128GB of memory, for example. When the
: amount of kernel memory is well bounded for certain systems, it is
: better to aggressively reclaim from existing MIGRATE_UNMOVABLE
: pageblocks rather than eagerly fallback to others.
:
: We have additional patches that help with this fragmentation if you're
: interested, specifically kcompactd compaction of MIGRATE_UNMOVABLE
: pageblocks triggered by fallback of non-__GFP_MOVABLE allocations and
: draining of pcp lists back to the zone free area to prevent stranding.
[rientjes@google.com: updates]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802131700160.71590@chino.kir.corp.google.com
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1802121622470.179479@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:23:09 +00:00
|
|
|
return cmdline_parse_core(p, &required_movablecore,
|
|
|
|
&required_movablecore_percent);
|
2007-07-17 11:03:15 +00:00
|
|
|
}
|
|
|
|
|
2007-07-17 11:03:14 +00:00
|
|
|
early_param("kernelcore", cmdline_parse_kernelcore);
|
2007-07-17 11:03:15 +00:00
|
|
|
early_param("movablecore", cmdline_parse_movablecore);
|
2007-07-17 11:03:14 +00:00
|
|
|
|
2013-07-03 22:03:14 +00:00
|
|
|
void adjust_managed_page_count(struct page *page, long count)
|
|
|
|
{
|
2018-12-28 08:34:24 +00:00
|
|
|
atomic_long_add(count, &page_zone(page)->managed_pages);
|
2018-12-28 08:34:29 +00:00
|
|
|
totalram_pages_add(count);
|
2013-07-03 22:03:21 +00:00
|
|
|
#ifdef CONFIG_HIGHMEM
|
|
|
|
if (PageHighMem(page))
|
2018-12-28 08:34:29 +00:00
|
|
|
totalhigh_pages_add(count);
|
2013-07-03 22:03:21 +00:00
|
|
|
#endif
|
2013-07-03 22:03:14 +00:00
|
|
|
}
|
2013-07-03 22:03:21 +00:00
|
|
|
EXPORT_SYMBOL(adjust_managed_page_count);
|
2013-07-03 22:03:14 +00:00
|
|
|
|
2018-12-28 08:36:03 +00:00
|
|
|
unsigned long free_reserved_area(void *start, void *end, int poison, const char *s)
|
2013-04-29 22:06:21 +00:00
|
|
|
{
|
mm: change signature of free_reserved_area() to fix building warnings
Change signature of free_reserved_area() according to Russell King's
suggestion to fix following build warnings:
arch/arm/mm/init.c: In function 'mem_init':
arch/arm/mm/init.c:603:2: warning: passing argument 1 of 'free_reserved_area' makes integer from pointer without a cast [enabled by default]
free_reserved_area(__va(PHYS_PFN_OFFSET), swapper_pg_dir, 0, NULL);
^
In file included from include/linux/mman.h:4:0,
from arch/arm/mm/init.c:15:
include/linux/mm.h:1301:22: note: expected 'long unsigned int' but argument is of type 'void *'
extern unsigned long free_reserved_area(unsigned long start, unsigned long end,
mm/page_alloc.c: In function 'free_reserved_area':
>> mm/page_alloc.c:5134:3: warning: passing argument 1 of 'virt_to_phys' makes pointer from integer without a cast [enabled by default]
In file included from arch/mips/include/asm/page.h:49:0,
from include/linux/mmzone.h:20,
from include/linux/gfp.h:4,
from include/linux/mm.h:8,
from mm/page_alloc.c:18:
arch/mips/include/asm/io.h:119:29: note: expected 'const volatile void *' but argument is of type 'long unsigned int'
mm/page_alloc.c: In function 'free_area_init_nodes':
mm/page_alloc.c:5030:34: warning: array subscript is below array bounds [-Warray-bounds]
Also address some minor code review comments.
Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: <sworddragon2@aol.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Jianguo Wu <wujianguo@huawei.com>
Cc: Joonsoo Kim <js1304@gmail.com>
Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michel Lespinasse <walken@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-07-03 22:02:48 +00:00
|
|
|
void *pos;
|
|
|
|
unsigned long pages = 0;
|
2013-04-29 22:06:21 +00:00
|
|
|
|
mm: change signature of free_reserved_area() to fix building warnings
Change signature of free_reserved_area() according to Russell King's
suggestion to fix following build warnings:
arch/arm/mm/init.c: In function 'mem_init':
arch/arm/mm/init.c:603:2: warning: passing argument 1 of 'free_reserved_area' makes integer from pointer without a cast [enabled by default]
free_reserved_area(__va(PHYS_PFN_OFFSET), swapper_pg_dir, 0, NULL);
^
In file included from include/linux/mman.h:4:0,
from arch/arm/mm/init.c:15:
include/linux/mm.h:1301:22: note: expected 'long unsigned int' but argument is of type 'void *'
extern unsigned long free_reserved_area(unsigned long start, unsigned long end,
mm/page_alloc.c: In function 'free_reserved_area':
>> mm/page_alloc.c:5134:3: warning: passing argument 1 of 'virt_to_phys' makes pointer from integer without a cast [enabled by default]
In file included from arch/mips/include/asm/page.h:49:0,
from include/linux/mmzone.h:20,
from include/linux/gfp.h:4,
from include/linux/mm.h:8,
from mm/page_alloc.c:18:
arch/mips/include/asm/io.h:119:29: note: expected 'const volatile void *' but argument is of type 'long unsigned int'
mm/page_alloc.c: In function 'free_area_init_nodes':
mm/page_alloc.c:5030:34: warning: array subscript is below array bounds [-Warray-bounds]
Also address some minor code review comments.
Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: <sworddragon2@aol.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Jianguo Wu <wujianguo@huawei.com>
Cc: Joonsoo Kim <js1304@gmail.com>
Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michel Lespinasse <walken@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-07-03 22:02:48 +00:00
|
|
|
start = (void *)PAGE_ALIGN((unsigned long)start);
|
|
|
|
end = (void *)((unsigned long)end & PAGE_MASK);
|
|
|
|
for (pos = start; pos < end; pos += PAGE_SIZE, pages++) {
|
2018-08-02 22:58:26 +00:00
|
|
|
struct page *page = virt_to_page(pos);
|
|
|
|
void *direct_map_addr;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* 'direct_map_addr' might be different from 'pos'
|
|
|
|
* because some architectures' virt_to_page()
|
|
|
|
* work with aliases. Getting the direct map
|
|
|
|
* address ensures that we get a _writeable_
|
|
|
|
* alias for the memset().
|
|
|
|
*/
|
|
|
|
direct_map_addr = page_address(page);
|
2020-12-22 20:01:49 +00:00
|
|
|
/*
|
|
|
|
* Perform a kasan-unchecked memset() since this memory
|
|
|
|
* has not been initialized.
|
|
|
|
*/
|
|
|
|
direct_map_addr = kasan_reset_tag(direct_map_addr);
|
2013-07-03 22:02:51 +00:00
|
|
|
if ((unsigned int)poison <= 0xFF)
|
2018-08-02 22:58:26 +00:00
|
|
|
memset(direct_map_addr, poison, PAGE_SIZE);
|
|
|
|
|
|
|
|
free_reserved_page(page);
|
2013-04-29 22:06:21 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (pages && s)
|
2016-10-25 14:51:14 +00:00
|
|
|
pr_info("Freeing %s memory: %ldK\n",
|
|
|
|
s, pages << (PAGE_SHIFT - 10));
|
2013-04-29 22:06:21 +00:00
|
|
|
|
|
|
|
return pages;
|
|
|
|
}
|
|
|
|
|
2013-07-03 22:03:41 +00:00
|
|
|
void __init mem_init_print_info(const char *str)
|
|
|
|
{
|
|
|
|
unsigned long physpages, codesize, datasize, rosize, bss_size;
|
|
|
|
unsigned long init_code_size, init_data_size;
|
|
|
|
|
|
|
|
physpages = get_num_physpages();
|
|
|
|
codesize = _etext - _stext;
|
|
|
|
datasize = _edata - _sdata;
|
|
|
|
rosize = __end_rodata - __start_rodata;
|
|
|
|
bss_size = __bss_stop - __bss_start;
|
|
|
|
init_data_size = __init_end - __init_begin;
|
|
|
|
init_code_size = _einittext - _sinittext;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Detect special cases and adjust section sizes accordingly:
|
|
|
|
* 1) .init.* may be embedded into .data sections
|
|
|
|
* 2) .init.text.* may be out of [__init_begin, __init_end],
|
|
|
|
* please refer to arch/tile/kernel/vmlinux.lds.S.
|
|
|
|
* 3) .rodata.* may be embedded into .text or .data sections.
|
|
|
|
*/
|
|
|
|
#define adj_init_size(start, end, size, pos, adj) \
|
2013-09-11 21:20:34 +00:00
|
|
|
do { \
|
|
|
|
if (start <= pos && pos < end && size > adj) \
|
|
|
|
size -= adj; \
|
|
|
|
} while (0)
|
2013-07-03 22:03:41 +00:00
|
|
|
|
|
|
|
adj_init_size(__init_begin, __init_end, init_data_size,
|
|
|
|
_sinittext, init_code_size);
|
|
|
|
adj_init_size(_stext, _etext, codesize, _sinittext, init_code_size);
|
|
|
|
adj_init_size(_sdata, _edata, datasize, __init_begin, init_data_size);
|
|
|
|
adj_init_size(_stext, _etext, codesize, __start_rodata, rosize);
|
|
|
|
adj_init_size(_sdata, _edata, datasize, __start_rodata, rosize);
|
|
|
|
|
|
|
|
#undef adj_init_size
|
|
|
|
|
2016-03-17 21:19:47 +00:00
|
|
|
pr_info("Memory: %luK/%luK available (%luK kernel code, %luK rwdata, %luK rodata, %luK init, %luK bss, %luK reserved, %luK cma-reserved"
|
2013-07-03 22:03:41 +00:00
|
|
|
#ifdef CONFIG_HIGHMEM
|
2016-03-17 21:19:47 +00:00
|
|
|
", %luK highmem"
|
2013-07-03 22:03:41 +00:00
|
|
|
#endif
|
2016-03-17 21:19:47 +00:00
|
|
|
"%s%s)\n",
|
|
|
|
nr_free_pages() << (PAGE_SHIFT - 10),
|
|
|
|
physpages << (PAGE_SHIFT - 10),
|
|
|
|
codesize >> 10, datasize >> 10, rosize >> 10,
|
|
|
|
(init_data_size + init_code_size) >> 10, bss_size >> 10,
|
2018-12-28 08:34:29 +00:00
|
|
|
(physpages - totalram_pages() - totalcma_pages) << (PAGE_SHIFT - 10),
|
2016-03-17 21:19:47 +00:00
|
|
|
totalcma_pages << (PAGE_SHIFT - 10),
|
2013-07-03 22:03:41 +00:00
|
|
|
#ifdef CONFIG_HIGHMEM
|
2018-12-28 08:34:29 +00:00
|
|
|
totalhigh_pages() << (PAGE_SHIFT - 10),
|
2013-07-03 22:03:41 +00:00
|
|
|
#endif
|
2016-03-17 21:19:47 +00:00
|
|
|
str ? ", " : "", str ? str : "");
|
2013-07-03 22:03:41 +00:00
|
|
|
}
|
|
|
|
|
2006-09-27 08:49:56 +00:00
|
|
|
/**
|
2006-10-04 09:15:25 +00:00
|
|
|
* set_dma_reserve - set the specified number of pages reserved in the first zone
|
|
|
|
* @new_dma_reserve: The number of pages to mark reserved
|
2006-09-27 08:49:56 +00:00
|
|
|
*
|
2015-09-08 22:04:10 +00:00
|
|
|
* The per-cpu batchsize and zone watermarks are determined by managed_pages.
|
2006-09-27 08:49:56 +00:00
|
|
|
* In the DMA zone, a significant percentage may be consumed by kernel image
|
|
|
|
* and other unfreeable allocations which can skew the watermarks badly. This
|
2006-10-04 09:15:25 +00:00
|
|
|
* function may optionally be used to account for unfreeable pages in the
|
|
|
|
* first zone (e.g., ZONE_DMA). The effect will be lower watermarks and
|
|
|
|
* smaller per-cpu batchsize.
|
2006-09-27 08:49:56 +00:00
|
|
|
*/
|
|
|
|
void __init set_dma_reserve(unsigned long new_dma_reserve)
|
|
|
|
{
|
|
|
|
dma_reserve = new_dma_reserve;
|
|
|
|
}
|
|
|
|
|
2016-11-03 14:50:02 +00:00
|
|
|
static int page_alloc_cpu_dead(unsigned int cpu)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
|
2016-11-03 14:50:02 +00:00
|
|
|
lru_add_drain_cpu(cpu);
|
|
|
|
drain_pages(cpu);
|
2008-02-05 06:29:11 +00:00
|
|
|
|
2016-11-03 14:50:02 +00:00
|
|
|
/*
|
|
|
|
* Spill the event counters of the dead processor
|
|
|
|
* into the current processors event counters.
|
|
|
|
* This artificially elevates the count of the current
|
|
|
|
* processor.
|
|
|
|
*/
|
|
|
|
vm_events_fold_cpu(cpu);
|
2008-02-05 06:29:11 +00:00
|
|
|
|
2016-11-03 14:50:02 +00:00
|
|
|
/*
|
|
|
|
* Zero the differential counters of the dead processor
|
|
|
|
* so that the vm statistics are consistent.
|
|
|
|
*
|
|
|
|
* This is only okay since the processor is dead and cannot
|
|
|
|
* race with what we are doing.
|
|
|
|
*/
|
|
|
|
cpu_vm_stats_fold(cpu);
|
|
|
|
return 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2019-07-12 03:59:12 +00:00
|
|
|
#ifdef CONFIG_NUMA
|
|
|
|
int hashdist = HASHDIST_DEFAULT;
|
|
|
|
|
|
|
|
static int __init set_hashdist(char *str)
|
|
|
|
{
|
|
|
|
if (!str)
|
|
|
|
return 0;
|
|
|
|
hashdist = simple_strtoul(str, &str, 0);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("hashdist=", set_hashdist);
|
|
|
|
#endif
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
void __init page_alloc_init(void)
|
|
|
|
{
|
2016-11-03 14:50:02 +00:00
|
|
|
int ret;
|
|
|
|
|
2019-07-12 03:59:12 +00:00
|
|
|
#ifdef CONFIG_NUMA
|
|
|
|
if (num_node_state(N_MEMORY) == 1)
|
|
|
|
hashdist = 0;
|
|
|
|
#endif
|
|
|
|
|
2016-11-03 14:50:02 +00:00
|
|
|
ret = cpuhp_setup_state_nocalls(CPUHP_PAGE_ALLOC_DEAD,
|
|
|
|
"mm/page_alloc:dead", NULL,
|
|
|
|
page_alloc_cpu_dead);
|
|
|
|
WARN_ON(ret < 0);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2006-04-11 05:52:59 +00:00
|
|
|
/*
|
2015-09-08 22:04:13 +00:00
|
|
|
* calculate_totalreserve_pages - called when sysctl_lowmem_reserve_ratio
|
2006-04-11 05:52:59 +00:00
|
|
|
* or min_free_kbytes changes.
|
|
|
|
*/
|
|
|
|
static void calculate_totalreserve_pages(void)
|
|
|
|
{
|
|
|
|
struct pglist_data *pgdat;
|
|
|
|
unsigned long reserve_pages = 0;
|
2006-09-26 06:31:18 +00:00
|
|
|
enum zone_type i, j;
|
2006-04-11 05:52:59 +00:00
|
|
|
|
|
|
|
for_each_online_pgdat(pgdat) {
|
2016-07-28 22:46:11 +00:00
|
|
|
|
|
|
|
pgdat->totalreserve_pages = 0;
|
|
|
|
|
2006-04-11 05:52:59 +00:00
|
|
|
for (i = 0; i < MAX_NR_ZONES; i++) {
|
|
|
|
struct zone *zone = pgdat->node_zones + i;
|
2014-08-06 23:07:14 +00:00
|
|
|
long max = 0;
|
2018-12-28 08:34:24 +00:00
|
|
|
unsigned long managed_pages = zone_managed_pages(zone);
|
2006-04-11 05:52:59 +00:00
|
|
|
|
|
|
|
/* Find valid and maximum lowmem_reserve in the zone */
|
|
|
|
for (j = i; j < MAX_NR_ZONES; j++) {
|
|
|
|
if (zone->lowmem_reserve[j] > max)
|
|
|
|
max = zone->lowmem_reserve[j];
|
|
|
|
}
|
|
|
|
|
2009-06-16 22:32:12 +00:00
|
|
|
/* we treat the high watermark as reserved pages. */
|
|
|
|
max += high_wmark_pages(zone);
|
2006-04-11 05:52:59 +00:00
|
|
|
|
2018-12-28 08:34:20 +00:00
|
|
|
if (max > managed_pages)
|
|
|
|
max = managed_pages;
|
2016-01-14 23:20:15 +00:00
|
|
|
|
2016-07-28 22:46:11 +00:00
|
|
|
pgdat->totalreserve_pages += max;
|
2016-01-14 23:20:15 +00:00
|
|
|
|
2006-04-11 05:52:59 +00:00
|
|
|
reserve_pages += max;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
totalreserve_pages = reserve_pages;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* setup_per_zone_lowmem_reserve - called whenever
|
2015-09-08 22:04:13 +00:00
|
|
|
* sysctl_lowmem_reserve_ratio changes. Ensures that each zone
|
2005-04-16 22:20:36 +00:00
|
|
|
* has a correct pages reserved value, so an adequate number of
|
|
|
|
* pages are left in the zone after a successful __alloc_pages().
|
|
|
|
*/
|
|
|
|
static void setup_per_zone_lowmem_reserve(void)
|
|
|
|
{
|
|
|
|
struct pglist_data *pgdat;
|
2020-12-15 03:11:22 +00:00
|
|
|
enum zone_type i, j;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-03-27 09:15:59 +00:00
|
|
|
for_each_online_pgdat(pgdat) {
|
2020-12-15 03:11:22 +00:00
|
|
|
for (i = 0; i < MAX_NR_ZONES - 1; i++) {
|
|
|
|
struct zone *zone = &pgdat->node_zones[i];
|
|
|
|
int ratio = sysctl_lowmem_reserve_ratio[i];
|
|
|
|
bool clear = !ratio || !zone_managed_pages(zone);
|
|
|
|
unsigned long managed_pages = 0;
|
|
|
|
|
|
|
|
for (j = i + 1; j < MAX_NR_ZONES; j++) {
|
|
|
|
if (clear) {
|
|
|
|
zone->lowmem_reserve[j] = 0;
|
2018-04-10 23:30:11 +00:00
|
|
|
} else {
|
2020-12-15 03:11:22 +00:00
|
|
|
struct zone *upper_zone = &pgdat->node_zones[j];
|
|
|
|
|
|
|
|
managed_pages += zone_managed_pages(upper_zone);
|
|
|
|
zone->lowmem_reserve[j] = managed_pages / ratio;
|
2018-04-10 23:30:11 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2006-04-11 05:52:59 +00:00
|
|
|
|
|
|
|
/* update totalreserve_pages */
|
|
|
|
calculate_totalreserve_pages();
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2011-04-25 21:36:42 +00:00
|
|
|
static void __setup_per_zone_wmarks(void)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
unsigned long pages_min = min_free_kbytes >> (PAGE_SHIFT - 10);
|
|
|
|
unsigned long lowmem_pages = 0;
|
|
|
|
struct zone *zone;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
/* Calculate total number of !ZONE_HIGHMEM pages */
|
|
|
|
for_each_zone(zone) {
|
|
|
|
if (!is_highmem(zone))
|
2018-12-28 08:34:24 +00:00
|
|
|
lowmem_pages += zone_managed_pages(zone);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
for_each_zone(zone) {
|
2006-05-15 16:43:59 +00:00
|
|
|
u64 tmp;
|
|
|
|
|
2008-10-19 03:27:11 +00:00
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
2018-12-28 08:34:24 +00:00
|
|
|
tmp = (u64)pages_min * zone_managed_pages(zone);
|
2006-05-15 16:43:59 +00:00
|
|
|
do_div(tmp, lowmem_pages);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (is_highmem(zone)) {
|
|
|
|
/*
|
2005-11-14 00:06:45 +00:00
|
|
|
* __GFP_HIGH and PF_MEMALLOC allocations usually don't
|
|
|
|
* need highmem pages, so cap pages_min to a small
|
|
|
|
* value here.
|
|
|
|
*
|
2009-06-16 22:32:12 +00:00
|
|
|
* The WMARK_HIGH-WMARK_LOW and (WMARK_LOW-WMARK_MIN)
|
2019-03-05 23:46:22 +00:00
|
|
|
* deltas control async page reclaim, and so should
|
2005-11-14 00:06:45 +00:00
|
|
|
* not be capped for highmem.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2013-02-23 00:32:22 +00:00
|
|
|
unsigned long min_pages;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2018-12-28 08:34:24 +00:00
|
|
|
min_pages = zone_managed_pages(zone) / 1024;
|
2013-02-23 00:32:22 +00:00
|
|
|
min_pages = clamp(min_pages, SWAP_CLUSTER_MAX, 128UL);
|
2018-12-28 08:35:44 +00:00
|
|
|
zone->_watermark[WMARK_MIN] = min_pages;
|
2005-04-16 22:20:36 +00:00
|
|
|
} else {
|
2005-11-14 00:06:45 +00:00
|
|
|
/*
|
|
|
|
* If it's a lowmem zone, reserve a number of pages
|
2005-04-16 22:20:36 +00:00
|
|
|
* proportionate to the zone's size.
|
|
|
|
*/
|
2018-12-28 08:35:44 +00:00
|
|
|
zone->_watermark[WMARK_MIN] = tmp;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2016-03-17 21:19:14 +00:00
|
|
|
/*
|
|
|
|
* Set the kswapd watermarks distance according to the
|
|
|
|
* scale factor in proportion to available memory, but
|
|
|
|
* ensure a minimum size on small systems.
|
|
|
|
*/
|
|
|
|
tmp = max_t(u64, tmp >> 2,
|
2018-12-28 08:34:24 +00:00
|
|
|
mult_frac(zone_managed_pages(zone),
|
2016-03-17 21:19:14 +00:00
|
|
|
watermark_scale_factor, 10000));
|
|
|
|
|
mm, page_alloc: reset the zone->watermark_boost early
Updating the zone watermarks by any means, like min_free_kbytes,
water_mark_scale_factor etc, when ->watermark_boost is set will result in
higher low and high watermarks than the user asked.
Below are the steps to reproduce the problem on system setup of Android
kernel running on Snapdragon hardware.
1) Default settings of the system are as below:
#cat /proc/sys/vm/min_free_kbytes = 5162
#cat /proc/zoneinfo | grep -e boost -e low -e "high " -e min -e Node
Node 0, zone Normal
min 797
low 8340
high 8539
2) Monitor the zone->watermark_boost(by adding a debug print in the
kernel) and whenever it is greater than zero value, write the same
value of min_free_kbytes obtained from step 1.
#echo 5162 > /proc/sys/vm/min_free_kbytes
3) Then read the zone watermarks in the system while the
->watermark_boost is zero. This should show the same values of
watermarks as step 1 but shown a higher values than asked.
#cat /proc/zoneinfo | grep -e boost -e low -e "high " -e min -e Node
Node 0, zone Normal
min 797
low 21148
high 21347
These higher values are because of updating the zone watermarks using the
macro min_wmark_pages(zone) which also adds the zone->watermark_boost.
#define min_wmark_pages(z) (z->_watermark[WMARK_MIN] +
z->watermark_boost)
So the steps that lead to the issue are:
1) On the extfrag event, watermarks are boosted by storing the required
value in ->watermark_boost.
2) User tries to update the zone watermarks level in the system through
min_free_kbytes or watermark_scale_factor.
3) Later, when kswapd woke up, it resets the zone->watermark_boost to
zero.
In step 2), we use the min_wmark_pages() macro to store the watermarks
in the zone structure thus the values are always offsetted by
->watermark_boost value. This can be avoided by resetting the
->watermark_boost to zero before it is used.
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Link: http://lkml.kernel.org/r/1589457511-4255-1-git-send-email-charante@codeaurora.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-03 22:59:14 +00:00
|
|
|
zone->watermark_boost = 0;
|
2018-12-28 08:35:44 +00:00
|
|
|
zone->_watermark[WMARK_LOW] = min_wmark_pages(zone) + tmp;
|
|
|
|
zone->_watermark[WMARK_HIGH] = min_wmark_pages(zone) + tmp * 2;
|
2012-01-25 11:49:24 +00:00
|
|
|
|
2008-10-19 03:27:11 +00:00
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2006-04-11 05:52:59 +00:00
|
|
|
|
|
|
|
/* update totalreserve_pages */
|
|
|
|
calculate_totalreserve_pages();
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2011-04-25 21:36:42 +00:00
|
|
|
/**
|
|
|
|
* setup_per_zone_wmarks - called when min_free_kbytes changes
|
|
|
|
* or when memory is hot-{added|removed}
|
|
|
|
*
|
|
|
|
* Ensures that the watermark[min,low,high] values for each zone are set
|
|
|
|
* correctly with respect to min_free_kbytes.
|
|
|
|
*/
|
|
|
|
void setup_per_zone_wmarks(void)
|
|
|
|
{
|
2017-09-06 23:20:37 +00:00
|
|
|
static DEFINE_SPINLOCK(lock);
|
|
|
|
|
|
|
|
spin_lock(&lock);
|
2011-04-25 21:36:42 +00:00
|
|
|
__setup_per_zone_wmarks();
|
2017-09-06 23:20:37 +00:00
|
|
|
spin_unlock(&lock);
|
2011-04-25 21:36:42 +00:00
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Initialise min_free_kbytes.
|
|
|
|
*
|
|
|
|
* For small machines we want it small (128k min). For large machines
|
2020-07-03 22:15:30 +00:00
|
|
|
* we want it large (256MB max). But it is not linear, because network
|
2005-04-16 22:20:36 +00:00
|
|
|
* bandwidth does not increase linearly with machine size. We use
|
|
|
|
*
|
2013-09-11 21:20:34 +00:00
|
|
|
* min_free_kbytes = 4 * sqrt(lowmem_kbytes), for better accuracy:
|
2005-04-16 22:20:36 +00:00
|
|
|
* min_free_kbytes = sqrt(lowmem_kbytes * 16)
|
|
|
|
*
|
|
|
|
* which yields
|
|
|
|
*
|
|
|
|
* 16MB: 512k
|
|
|
|
* 32MB: 724k
|
|
|
|
* 64MB: 1024k
|
|
|
|
* 128MB: 1448k
|
|
|
|
* 256MB: 2048k
|
|
|
|
* 512MB: 2896k
|
|
|
|
* 1024MB: 4096k
|
|
|
|
* 2048MB: 5792k
|
|
|
|
* 4096MB: 8192k
|
|
|
|
* 8192MB: 11584k
|
|
|
|
* 16384MB: 16384k
|
|
|
|
*/
|
2011-05-25 00:11:32 +00:00
|
|
|
int __meminit init_per_zone_wmark_min(void)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
unsigned long lowmem_kbytes;
|
2013-07-08 23:00:40 +00:00
|
|
|
int new_min_free_kbytes;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
lowmem_kbytes = nr_free_buffer_pages() * (PAGE_SIZE >> 10);
|
2013-07-08 23:00:40 +00:00
|
|
|
new_min_free_kbytes = int_sqrt(lowmem_kbytes * 16);
|
|
|
|
|
|
|
|
if (new_min_free_kbytes > user_min_free_kbytes) {
|
|
|
|
min_free_kbytes = new_min_free_kbytes;
|
|
|
|
if (min_free_kbytes < 128)
|
|
|
|
min_free_kbytes = 128;
|
2020-04-02 04:09:44 +00:00
|
|
|
if (min_free_kbytes > 262144)
|
|
|
|
min_free_kbytes = 262144;
|
2013-07-08 23:00:40 +00:00
|
|
|
} else {
|
|
|
|
pr_warn("min_free_kbytes is not updated to %d because user defined value %d is preferred\n",
|
|
|
|
new_min_free_kbytes, user_min_free_kbytes);
|
|
|
|
}
|
2009-06-16 22:32:48 +00:00
|
|
|
setup_per_zone_wmarks();
|
2011-05-25 00:11:33 +00:00
|
|
|
refresh_zone_stat_thresholds();
|
2005-04-16 22:20:36 +00:00
|
|
|
setup_per_zone_lowmem_reserve();
|
2016-08-10 23:27:49 +00:00
|
|
|
|
|
|
|
#ifdef CONFIG_NUMA
|
|
|
|
setup_min_unmapped_ratio();
|
|
|
|
setup_min_slab_ratio();
|
|
|
|
#endif
|
|
|
|
|
2020-10-11 06:16:40 +00:00
|
|
|
khugepaged_min_free_kbytes_update();
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2020-08-21 00:42:24 +00:00
|
|
|
postcore_initcall(init_per_zone_wmark_min)
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/*
|
2013-09-11 21:20:34 +00:00
|
|
|
* min_free_kbytes_sysctl_handler - just a wrapper around proc_dointvec() so
|
2005-04-16 22:20:36 +00:00
|
|
|
* that we can call two helper functions whenever min_free_kbytes
|
|
|
|
* changes.
|
|
|
|
*/
|
2014-06-06 21:38:09 +00:00
|
|
|
int min_free_kbytes_sysctl_handler(struct ctl_table *table, int write,
|
2020-04-24 06:43:38 +00:00
|
|
|
void *buffer, size_t *length, loff_t *ppos)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2014-01-23 23:53:17 +00:00
|
|
|
int rc;
|
|
|
|
|
|
|
|
rc = proc_dointvec_minmax(table, write, buffer, length, ppos);
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
|
2013-07-08 23:00:40 +00:00
|
|
|
if (write) {
|
|
|
|
user_min_free_kbytes = min_free_kbytes;
|
2009-06-16 22:32:48 +00:00
|
|
|
setup_per_zone_wmarks();
|
2013-07-08 23:00:40 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2016-03-17 21:19:14 +00:00
|
|
|
int watermark_scale_factor_sysctl_handler(struct ctl_table *table, int write,
|
2020-04-24 06:43:38 +00:00
|
|
|
void *buffer, size_t *length, loff_t *ppos)
|
2016-03-17 21:19:14 +00:00
|
|
|
{
|
|
|
|
int rc;
|
|
|
|
|
|
|
|
rc = proc_dointvec_minmax(table, write, buffer, length, ppos);
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
|
|
|
|
if (write)
|
|
|
|
setup_per_zone_wmarks();
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-07-03 07:24:13 +00:00
|
|
|
#ifdef CONFIG_NUMA
|
2016-08-10 23:27:49 +00:00
|
|
|
static void setup_min_unmapped_ratio(void)
|
2006-07-03 07:24:13 +00:00
|
|
|
{
|
2016-08-10 23:27:49 +00:00
|
|
|
pg_data_t *pgdat;
|
2006-07-03 07:24:13 +00:00
|
|
|
struct zone *zone;
|
|
|
|
|
2016-07-28 22:46:32 +00:00
|
|
|
for_each_online_pgdat(pgdat)
|
2016-08-10 23:27:46 +00:00
|
|
|
pgdat->min_unmapped_pages = 0;
|
2016-07-28 22:46:32 +00:00
|
|
|
|
2006-07-03 07:24:13 +00:00
|
|
|
for_each_zone(zone)
|
2018-12-28 08:34:24 +00:00
|
|
|
zone->zone_pgdat->min_unmapped_pages += (zone_managed_pages(zone) *
|
|
|
|
sysctl_min_unmapped_ratio) / 100;
|
2006-07-03 07:24:13 +00:00
|
|
|
}
|
2006-09-26 06:31:52 +00:00
|
|
|
|
2016-08-10 23:27:49 +00:00
|
|
|
|
|
|
|
int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *table, int write,
|
2020-04-24 06:43:38 +00:00
|
|
|
void *buffer, size_t *length, loff_t *ppos)
|
2006-09-26 06:31:52 +00:00
|
|
|
{
|
|
|
|
int rc;
|
|
|
|
|
2009-09-23 22:57:19 +00:00
|
|
|
rc = proc_dointvec_minmax(table, write, buffer, length, ppos);
|
2006-09-26 06:31:52 +00:00
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
|
2016-08-10 23:27:49 +00:00
|
|
|
setup_min_unmapped_ratio();
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void setup_min_slab_ratio(void)
|
|
|
|
{
|
|
|
|
pg_data_t *pgdat;
|
|
|
|
struct zone *zone;
|
|
|
|
|
2016-07-28 22:46:32 +00:00
|
|
|
for_each_online_pgdat(pgdat)
|
|
|
|
pgdat->min_slab_pages = 0;
|
|
|
|
|
2006-09-26 06:31:52 +00:00
|
|
|
for_each_zone(zone)
|
2018-12-28 08:34:24 +00:00
|
|
|
zone->zone_pgdat->min_slab_pages += (zone_managed_pages(zone) *
|
|
|
|
sysctl_min_slab_ratio) / 100;
|
2016-08-10 23:27:49 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *table, int write,
|
2020-04-24 06:43:38 +00:00
|
|
|
void *buffer, size_t *length, loff_t *ppos)
|
2016-08-10 23:27:49 +00:00
|
|
|
{
|
|
|
|
int rc;
|
|
|
|
|
|
|
|
rc = proc_dointvec_minmax(table, write, buffer, length, ppos);
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
|
|
|
|
setup_min_slab_ratio();
|
|
|
|
|
2006-09-26 06:31:52 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2006-07-03 07:24:13 +00:00
|
|
|
#endif
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* lowmem_reserve_ratio_sysctl_handler - just a wrapper around
|
|
|
|
* proc_dointvec() so that we can call setup_per_zone_lowmem_reserve()
|
|
|
|
* whenever sysctl_lowmem_reserve_ratio changes.
|
|
|
|
*
|
|
|
|
* The reserve ratio obviously has absolutely no relation with the
|
2009-06-16 22:32:12 +00:00
|
|
|
* minimum watermarks. The lowmem reserve ratio can only make sense
|
2005-04-16 22:20:36 +00:00
|
|
|
* if in function of the boot time zone sizes.
|
|
|
|
*/
|
2014-06-06 21:38:09 +00:00
|
|
|
int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *table, int write,
|
2020-04-24 06:43:38 +00:00
|
|
|
void *buffer, size_t *length, loff_t *ppos)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2020-06-03 22:58:48 +00:00
|
|
|
int i;
|
|
|
|
|
2009-09-23 22:57:19 +00:00
|
|
|
proc_dointvec_minmax(table, write, buffer, length, ppos);
|
2020-06-03 22:58:48 +00:00
|
|
|
|
|
|
|
for (i = 0; i < MAX_NR_ZONES; i++) {
|
|
|
|
if (sysctl_lowmem_reserve_ratio[i] < 1)
|
|
|
|
sysctl_lowmem_reserve_ratio[i] = 0;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
setup_per_zone_lowmem_reserve();
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-01-08 09:00:40 +00:00
|
|
|
/*
|
|
|
|
* percpu_pagelist_fraction - changes the pcp->high for each zone on each
|
2013-09-11 21:20:34 +00:00
|
|
|
* cpu. It is the fraction of total pages in each zone that a hot per cpu
|
|
|
|
* pagelist can have before it gets flushed back to buddy allocator.
|
2006-01-08 09:00:40 +00:00
|
|
|
*/
|
2014-06-06 21:38:09 +00:00
|
|
|
int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *table, int write,
|
2020-04-24 06:43:38 +00:00
|
|
|
void *buffer, size_t *length, loff_t *ppos)
|
2006-01-08 09:00:40 +00:00
|
|
|
{
|
|
|
|
struct zone *zone;
|
2014-06-23 20:22:04 +00:00
|
|
|
int old_percpu_pagelist_fraction;
|
2006-01-08 09:00:40 +00:00
|
|
|
int ret;
|
|
|
|
|
2014-06-23 20:22:04 +00:00
|
|
|
mutex_lock(&pcp_batch_high_lock);
|
|
|
|
old_percpu_pagelist_fraction = percpu_pagelist_fraction;
|
|
|
|
|
2009-09-23 22:57:19 +00:00
|
|
|
ret = proc_dointvec_minmax(table, write, buffer, length, ppos);
|
2014-06-23 20:22:04 +00:00
|
|
|
if (!write || ret < 0)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/* Sanity checking to avoid pcp imbalance */
|
|
|
|
if (percpu_pagelist_fraction &&
|
|
|
|
percpu_pagelist_fraction < MIN_PERCPU_PAGELIST_FRACTION) {
|
|
|
|
percpu_pagelist_fraction = old_percpu_pagelist_fraction;
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* No change? */
|
|
|
|
if (percpu_pagelist_fraction == old_percpu_pagelist_fraction)
|
|
|
|
goto out;
|
2013-07-03 22:01:29 +00:00
|
|
|
|
2019-12-01 01:55:11 +00:00
|
|
|
for_each_populated_zone(zone)
|
2020-12-15 03:10:43 +00:00
|
|
|
zone_set_pageset_high_and_batch(zone);
|
2014-06-23 20:22:04 +00:00
|
|
|
out:
|
2013-07-03 22:01:29 +00:00
|
|
|
mutex_unlock(&pcp_batch_high_lock);
|
2014-06-23 20:22:04 +00:00
|
|
|
return ret;
|
2006-01-08 09:00:40 +00:00
|
|
|
}
|
|
|
|
|
2016-10-07 23:59:15 +00:00
|
|
|
#ifndef __HAVE_ARCH_RESERVED_KERNEL_PAGES
|
|
|
|
/*
|
|
|
|
* Returns the number of pages that arch has reserved but
|
|
|
|
* is not known to alloc_large_system_hash().
|
|
|
|
*/
|
|
|
|
static unsigned long __init arch_reserved_kernel_pages(void)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2017-07-06 22:39:14 +00:00
|
|
|
/*
|
|
|
|
* Adaptive scale is meant to reduce sizes of hash tables on large memory
|
|
|
|
* machines. As memory size is increased the scale is also increased but at
|
|
|
|
* slower pace. Starting from ADAPT_SCALE_BASE (64G), every time memory
|
|
|
|
* quadruples the scale is increased by one, which means the size of hash table
|
|
|
|
* only doubles, instead of quadrupling as well.
|
|
|
|
* Because 32-bit systems cannot have large physical memory, where this scaling
|
|
|
|
* makes sense, it is disabled on such platforms.
|
|
|
|
*/
|
|
|
|
#if __BITS_PER_LONG > 32
|
|
|
|
#define ADAPT_SCALE_BASE (64ul << 30)
|
|
|
|
#define ADAPT_SCALE_SHIFT 2
|
|
|
|
#define ADAPT_SCALE_NPAGES (ADAPT_SCALE_BASE >> PAGE_SHIFT)
|
|
|
|
#endif
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* allocate a large system hash table from bootmem
|
|
|
|
* - it is assumed that the hash table must contain an exact power-of-2
|
|
|
|
* quantity of entries
|
|
|
|
* - limit is the number of hash buckets, not the total allocation size
|
|
|
|
*/
|
|
|
|
void *__init alloc_large_system_hash(const char *tablename,
|
|
|
|
unsigned long bucketsize,
|
|
|
|
unsigned long numentries,
|
|
|
|
int scale,
|
|
|
|
int flags,
|
|
|
|
unsigned int *_hash_shift,
|
|
|
|
unsigned int *_hash_mask,
|
2012-05-23 13:33:35 +00:00
|
|
|
unsigned long low_limit,
|
|
|
|
unsigned long high_limit)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2012-05-23 13:33:35 +00:00
|
|
|
unsigned long long max = high_limit;
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long log2qty, size;
|
|
|
|
void *table = NULL;
|
2017-07-06 22:39:08 +00:00
|
|
|
gfp_t gfp_flags;
|
2019-07-12 03:59:09 +00:00
|
|
|
bool virt;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* allow the kernel cmdline to have a say */
|
|
|
|
if (!numentries) {
|
|
|
|
/* round applicable memory size up to nearest megabyte */
|
2006-12-07 04:37:33 +00:00
|
|
|
numentries = nr_kernel_pages;
|
2016-10-07 23:59:15 +00:00
|
|
|
numentries -= arch_reserved_kernel_pages();
|
2013-09-11 21:20:26 +00:00
|
|
|
|
|
|
|
/* It isn't necessary when PAGE_SIZE >= 1MB */
|
|
|
|
if (PAGE_SHIFT < 20)
|
|
|
|
numentries = round_up(numentries, (1<<20)/PAGE_SIZE);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2017-07-06 22:39:14 +00:00
|
|
|
#if __BITS_PER_LONG > 32
|
|
|
|
if (!high_limit) {
|
|
|
|
unsigned long adapt;
|
|
|
|
|
|
|
|
for (adapt = ADAPT_SCALE_NPAGES; adapt < numentries;
|
|
|
|
adapt <<= ADAPT_SCALE_SHIFT)
|
|
|
|
scale++;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* limit to 1 bucket per 2^scale bytes of low memory */
|
|
|
|
if (scale > PAGE_SHIFT)
|
|
|
|
numentries >>= (scale - PAGE_SHIFT);
|
|
|
|
else
|
|
|
|
numentries <<= (PAGE_SHIFT - scale);
|
2007-01-06 00:36:30 +00:00
|
|
|
|
|
|
|
/* Make sure we've got at least a 0-order allocation.. */
|
2009-09-22 00:03:07 +00:00
|
|
|
if (unlikely(flags & HASH_SMALL)) {
|
|
|
|
/* Makes no sense without HASH_EARLY */
|
|
|
|
WARN_ON(!(flags & HASH_EARLY));
|
|
|
|
if (!(numentries >> *_hash_shift)) {
|
|
|
|
numentries = 1UL << *_hash_shift;
|
|
|
|
BUG_ON(!numentries);
|
|
|
|
}
|
|
|
|
} else if (unlikely((numentries * bucketsize) < PAGE_SIZE))
|
2007-01-06 00:36:30 +00:00
|
|
|
numentries = PAGE_SIZE / bucketsize;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2006-03-25 11:08:02 +00:00
|
|
|
numentries = roundup_pow_of_two(numentries);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* limit allocation size to 1/16 total memory by default */
|
|
|
|
if (max == 0) {
|
|
|
|
max = ((unsigned long long)nr_all_pages << PAGE_SHIFT) >> 4;
|
|
|
|
do_div(max, bucketsize);
|
|
|
|
}
|
2012-02-08 20:39:07 +00:00
|
|
|
max = min(max, 0x80000000ULL);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2012-05-23 13:33:35 +00:00
|
|
|
if (numentries < low_limit)
|
|
|
|
numentries = low_limit;
|
2005-04-16 22:20:36 +00:00
|
|
|
if (numentries > max)
|
|
|
|
numentries = max;
|
|
|
|
|
2006-12-08 10:37:49 +00:00
|
|
|
log2qty = ilog2(numentries);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2017-07-06 22:39:08 +00:00
|
|
|
gfp_flags = (flags & HASH_ZERO) ? GFP_ATOMIC | __GFP_ZERO : GFP_ATOMIC;
|
2005-04-16 22:20:36 +00:00
|
|
|
do {
|
2019-07-12 03:59:09 +00:00
|
|
|
virt = false;
|
2005-04-16 22:20:36 +00:00
|
|
|
size = bucketsize << log2qty;
|
2017-11-16 01:36:27 +00:00
|
|
|
if (flags & HASH_EARLY) {
|
|
|
|
if (flags & HASH_ZERO)
|
2019-03-12 06:30:42 +00:00
|
|
|
table = memblock_alloc(size, SMP_CACHE_BYTES);
|
2017-11-16 01:36:27 +00:00
|
|
|
else
|
memblock: stop using implicit alignment to SMP_CACHE_BYTES
When a memblock allocation APIs are called with align = 0, the alignment
is implicitly set to SMP_CACHE_BYTES.
Implicit alignment is done deep in the memblock allocator and it can
come as a surprise. Not that such an alignment would be wrong even
when used incorrectly but it is better to be explicit for the sake of
clarity and the prinicple of the least surprise.
Replace all such uses of memblock APIs with the 'align' parameter
explicitly set to SMP_CACHE_BYTES and stop implicit alignment assignment
in the memblock internal allocation functions.
For the case when memblock APIs are used via helper functions, e.g. like
iommu_arena_new_node() in Alpha, the helper functions were detected with
Coccinelle's help and then manually examined and updated where
appropriate.
The direct memblock APIs users were updated using the semantic patch below:
@@
expression size, min_addr, max_addr, nid;
@@
(
|
- memblock_alloc_try_nid_raw(size, 0, min_addr, max_addr, nid)
+ memblock_alloc_try_nid_raw(size, SMP_CACHE_BYTES, min_addr, max_addr,
nid)
|
- memblock_alloc_try_nid_nopanic(size, 0, min_addr, max_addr, nid)
+ memblock_alloc_try_nid_nopanic(size, SMP_CACHE_BYTES, min_addr, max_addr,
nid)
|
- memblock_alloc_try_nid(size, 0, min_addr, max_addr, nid)
+ memblock_alloc_try_nid(size, SMP_CACHE_BYTES, min_addr, max_addr, nid)
|
- memblock_alloc(size, 0)
+ memblock_alloc(size, SMP_CACHE_BYTES)
|
- memblock_alloc_raw(size, 0)
+ memblock_alloc_raw(size, SMP_CACHE_BYTES)
|
- memblock_alloc_from(size, 0, min_addr)
+ memblock_alloc_from(size, SMP_CACHE_BYTES, min_addr)
|
- memblock_alloc_nopanic(size, 0)
+ memblock_alloc_nopanic(size, SMP_CACHE_BYTES)
|
- memblock_alloc_low(size, 0)
+ memblock_alloc_low(size, SMP_CACHE_BYTES)
|
- memblock_alloc_low_nopanic(size, 0)
+ memblock_alloc_low_nopanic(size, SMP_CACHE_BYTES)
|
- memblock_alloc_from_nopanic(size, 0, min_addr)
+ memblock_alloc_from_nopanic(size, SMP_CACHE_BYTES, min_addr)
|
- memblock_alloc_node(size, 0, nid)
+ memblock_alloc_node(size, SMP_CACHE_BYTES, nid)
)
[mhocko@suse.com: changelog update]
[akpm@linux-foundation.org: coding-style fixes]
[rppt@linux.ibm.com: fix missed uses of implicit alignment]
Link: http://lkml.kernel.org/r/20181016133656.GA10925@rapoport-lnx
Link: http://lkml.kernel.org/r/1538687224-17535-1-git-send-email-rppt@linux.vnet.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Acked-by: Paul Burton <paul.burton@mips.com> [MIPS]
Acked-by: Michael Ellerman <mpe@ellerman.id.au> [powerpc]
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Guan Xuetao <gxt@pku.edu.cn>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tony Luck <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-30 22:09:57 +00:00
|
|
|
table = memblock_alloc_raw(size,
|
|
|
|
SMP_CACHE_BYTES);
|
2019-07-12 03:59:09 +00:00
|
|
|
} else if (get_order(size) >= MAX_ORDER || hashdist) {
|
2020-06-02 04:51:40 +00:00
|
|
|
table = __vmalloc(size, gfp_flags);
|
2019-07-12 03:59:09 +00:00
|
|
|
virt = true;
|
2017-11-16 01:36:27 +00:00
|
|
|
} else {
|
2007-07-16 06:38:05 +00:00
|
|
|
/*
|
|
|
|
* If bucketsize is not a power-of-two, we may free
|
2009-06-16 22:32:19 +00:00
|
|
|
* some pages at the end of hash table which
|
|
|
|
* alloc_pages_exact() automatically does
|
2007-07-16 06:38:05 +00:00
|
|
|
*/
|
2019-07-12 03:59:09 +00:00
|
|
|
table = alloc_pages_exact(size, gfp_flags);
|
|
|
|
kmemleak_alloc(table, size, 1, gfp_flags);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
} while (!table && size > PAGE_SIZE && --log2qty);
|
|
|
|
|
|
|
|
if (!table)
|
|
|
|
panic("Failed to allocate %s hash table\n", tablename);
|
|
|
|
|
2019-07-12 03:59:09 +00:00
|
|
|
pr_info("%s hash table entries: %ld (order: %d, %lu bytes, %s)\n",
|
|
|
|
tablename, 1UL << log2qty, ilog2(size) - PAGE_SHIFT, size,
|
|
|
|
virt ? "vmalloc" : "linear");
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
if (_hash_shift)
|
|
|
|
*_hash_shift = log2qty;
|
|
|
|
if (_hash_mask)
|
|
|
|
*_hash_mask = (1 << log2qty) - 1;
|
|
|
|
|
|
|
|
return table;
|
|
|
|
}
|
2006-03-27 09:15:25 +00:00
|
|
|
|
2007-10-16 08:26:11 +00:00
|
|
|
/*
|
2012-07-31 23:43:01 +00:00
|
|
|
* This function checks whether pageblock includes unmovable pages or not.
|
|
|
|
*
|
2013-09-11 21:20:34 +00:00
|
|
|
* PageLRU check without isolation or lru_lock could race so that
|
2017-02-24 22:57:39 +00:00
|
|
|
* MIGRATE_MOVABLE block might include unmovable pages. And __PageMovable
|
|
|
|
* check without lock_page also may miss some movable non-lru pages at
|
|
|
|
* race condition. So you can't expect this function should be exact.
|
mm/hotplug: silence a lockdep splat with printk()
It is not that hard to trigger lockdep splats by calling printk from
under zone->lock. Most of them are false positives caused by lock
chains introduced early in the boot process and they do not cause any
real problems (although most of the early boot lock dependencies could
happen after boot as well). There are some console drivers which do
allocate from the printk context as well and those should be fixed. In
any case, false positives are not that trivial to workaround and it is
far from optimal to lose lockdep functionality for something that is a
non-issue.
So change has_unmovable_pages() so that it no longer calls dump_page()
itself - instead it returns a "struct page *" of the unmovable page back
to the caller so that in the case of a has_unmovable_pages() failure,
the caller can call dump_page() after releasing zone->lock. Also, make
dump_page() is able to report a CMA page as well, so the reason string
from has_unmovable_pages() can be removed.
Even though has_unmovable_pages doesn't hold any reference to the
returned page this should be reasonably safe for the purpose of
reporting the page (dump_page) because it cannot be hotremoved in the
context of memory unplug. The state of the page might change but that
is the case even with the existing code as zone->lock only plays role
for free pages.
While at it, remove a similar but unnecessary debug-only printk() as
well. A sample of one of those lockdep splats is,
WARNING: possible circular locking dependency detected
------------------------------------------------------
test.sh/8653 is trying to acquire lock:
ffffffff865a4460 (console_owner){-.-.}, at:
console_unlock+0x207/0x750
but task is already holding lock:
ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #3 (&(&zone->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock+0x2f/0x40
rmqueue_bulk.constprop.21+0xb6/0x1160
get_page_from_freelist+0x898/0x22c0
__alloc_pages_nodemask+0x2f3/0x1cd0
alloc_pages_current+0x9c/0x110
allocate_slab+0x4c6/0x19c0
new_slab+0x46/0x70
___slab_alloc+0x58b/0x960
__slab_alloc+0x43/0x70
__kmalloc+0x3ad/0x4b0
__tty_buffer_request_room+0x100/0x250
tty_insert_flip_string_fixed_flag+0x67/0x110
pty_write+0xa2/0xf0
n_tty_write+0x36b/0x7b0
tty_write+0x284/0x4c0
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
redirected_tty_write+0x6a/0xc0
do_iter_write+0x248/0x2a0
vfs_writev+0x106/0x1e0
do_writev+0xd4/0x180
__x64_sys_writev+0x45/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #2 (&(&port->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
tty_port_tty_get+0x20/0x60
tty_port_default_wakeup+0xf/0x30
tty_port_tty_wakeup+0x39/0x40
uart_write_wakeup+0x2a/0x40
serial8250_tx_chars+0x22e/0x440
serial8250_handle_irq.part.8+0x14a/0x170
serial8250_default_handle_irq+0x5c/0x90
serial8250_interrupt+0xa6/0x130
__handle_irq_event_percpu+0x78/0x4f0
handle_irq_event_percpu+0x70/0x100
handle_irq_event+0x5a/0x8b
handle_edge_irq+0x117/0x370
do_IRQ+0x9e/0x1e0
ret_from_intr+0x0/0x2a
cpuidle_enter_state+0x156/0x8e0
cpuidle_enter+0x41/0x70
call_cpuidle+0x5e/0x90
do_idle+0x333/0x370
cpu_startup_entry+0x1d/0x1f
start_secondary+0x290/0x330
secondary_startup_64+0xb6/0xc0
-> #1 (&port_lock_key){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
serial8250_console_write+0x3e4/0x450
univ8250_console_write+0x4b/0x60
console_unlock+0x501/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
-> #0 (console_owner){-.-.}:
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
other info that might help us debug this:
Chain exists of:
console_owner --> &(&port->lock)->rlock --> &(&zone->lock)->rlock
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&(&zone->lock)->rlock);
lock(&(&port->lock)->rlock);
lock(&(&zone->lock)->rlock);
lock(console_owner);
*** DEADLOCK ***
9 locks held by test.sh/8653:
#0: ffff88839ba7d408 (sb_writers#4){.+.+}, at:
vfs_write+0x25f/0x290
#1: ffff888277618880 (&of->mutex){+.+.}, at:
kernfs_fop_write+0x128/0x240
#2: ffff8898131fc218 (kn->count#115){.+.+}, at:
kernfs_fop_write+0x138/0x240
#3: ffffffff86962a80 (device_hotplug_lock){+.+.}, at:
lock_device_hotplug_sysfs+0x16/0x50
#4: ffff8884374f4990 (&dev->mutex){....}, at:
device_offline+0x70/0x110
#5: ffffffff86515250 (cpu_hotplug_lock.rw_sem){++++}, at:
__offline_pages+0xbf/0xa10
#6: ffffffff867405f0 (mem_hotplug_lock.rw_sem){++++}, at:
percpu_down_write+0x87/0x2f0
#7: ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
#8: ffffffff865a4920 (console_lock){+.+.}, at:
vprintk_emit+0x100/0x340
stack backtrace:
Hardware name: HPE ProLiant DL560 Gen10/ProLiant DL560 Gen10,
BIOS U34 05/21/2019
Call Trace:
dump_stack+0x86/0xca
print_circular_bug.cold.31+0x243/0x26e
check_noncircular+0x29e/0x2e0
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Link: http://lkml.kernel.org/r/20200117181200.20299-1-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-31 06:14:57 +00:00
|
|
|
*
|
|
|
|
* Returns a page without holding a reference. If the caller wants to
|
2020-08-12 01:33:14 +00:00
|
|
|
* dereference that page (e.g., dumping), it has to make sure that it
|
mm/hotplug: silence a lockdep splat with printk()
It is not that hard to trigger lockdep splats by calling printk from
under zone->lock. Most of them are false positives caused by lock
chains introduced early in the boot process and they do not cause any
real problems (although most of the early boot lock dependencies could
happen after boot as well). There are some console drivers which do
allocate from the printk context as well and those should be fixed. In
any case, false positives are not that trivial to workaround and it is
far from optimal to lose lockdep functionality for something that is a
non-issue.
So change has_unmovable_pages() so that it no longer calls dump_page()
itself - instead it returns a "struct page *" of the unmovable page back
to the caller so that in the case of a has_unmovable_pages() failure,
the caller can call dump_page() after releasing zone->lock. Also, make
dump_page() is able to report a CMA page as well, so the reason string
from has_unmovable_pages() can be removed.
Even though has_unmovable_pages doesn't hold any reference to the
returned page this should be reasonably safe for the purpose of
reporting the page (dump_page) because it cannot be hotremoved in the
context of memory unplug. The state of the page might change but that
is the case even with the existing code as zone->lock only plays role
for free pages.
While at it, remove a similar but unnecessary debug-only printk() as
well. A sample of one of those lockdep splats is,
WARNING: possible circular locking dependency detected
------------------------------------------------------
test.sh/8653 is trying to acquire lock:
ffffffff865a4460 (console_owner){-.-.}, at:
console_unlock+0x207/0x750
but task is already holding lock:
ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #3 (&(&zone->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock+0x2f/0x40
rmqueue_bulk.constprop.21+0xb6/0x1160
get_page_from_freelist+0x898/0x22c0
__alloc_pages_nodemask+0x2f3/0x1cd0
alloc_pages_current+0x9c/0x110
allocate_slab+0x4c6/0x19c0
new_slab+0x46/0x70
___slab_alloc+0x58b/0x960
__slab_alloc+0x43/0x70
__kmalloc+0x3ad/0x4b0
__tty_buffer_request_room+0x100/0x250
tty_insert_flip_string_fixed_flag+0x67/0x110
pty_write+0xa2/0xf0
n_tty_write+0x36b/0x7b0
tty_write+0x284/0x4c0
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
redirected_tty_write+0x6a/0xc0
do_iter_write+0x248/0x2a0
vfs_writev+0x106/0x1e0
do_writev+0xd4/0x180
__x64_sys_writev+0x45/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #2 (&(&port->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
tty_port_tty_get+0x20/0x60
tty_port_default_wakeup+0xf/0x30
tty_port_tty_wakeup+0x39/0x40
uart_write_wakeup+0x2a/0x40
serial8250_tx_chars+0x22e/0x440
serial8250_handle_irq.part.8+0x14a/0x170
serial8250_default_handle_irq+0x5c/0x90
serial8250_interrupt+0xa6/0x130
__handle_irq_event_percpu+0x78/0x4f0
handle_irq_event_percpu+0x70/0x100
handle_irq_event+0x5a/0x8b
handle_edge_irq+0x117/0x370
do_IRQ+0x9e/0x1e0
ret_from_intr+0x0/0x2a
cpuidle_enter_state+0x156/0x8e0
cpuidle_enter+0x41/0x70
call_cpuidle+0x5e/0x90
do_idle+0x333/0x370
cpu_startup_entry+0x1d/0x1f
start_secondary+0x290/0x330
secondary_startup_64+0xb6/0xc0
-> #1 (&port_lock_key){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
serial8250_console_write+0x3e4/0x450
univ8250_console_write+0x4b/0x60
console_unlock+0x501/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
-> #0 (console_owner){-.-.}:
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
other info that might help us debug this:
Chain exists of:
console_owner --> &(&port->lock)->rlock --> &(&zone->lock)->rlock
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&(&zone->lock)->rlock);
lock(&(&port->lock)->rlock);
lock(&(&zone->lock)->rlock);
lock(console_owner);
*** DEADLOCK ***
9 locks held by test.sh/8653:
#0: ffff88839ba7d408 (sb_writers#4){.+.+}, at:
vfs_write+0x25f/0x290
#1: ffff888277618880 (&of->mutex){+.+.}, at:
kernfs_fop_write+0x128/0x240
#2: ffff8898131fc218 (kn->count#115){.+.+}, at:
kernfs_fop_write+0x138/0x240
#3: ffffffff86962a80 (device_hotplug_lock){+.+.}, at:
lock_device_hotplug_sysfs+0x16/0x50
#4: ffff8884374f4990 (&dev->mutex){....}, at:
device_offline+0x70/0x110
#5: ffffffff86515250 (cpu_hotplug_lock.rw_sem){++++}, at:
__offline_pages+0xbf/0xa10
#6: ffffffff867405f0 (mem_hotplug_lock.rw_sem){++++}, at:
percpu_down_write+0x87/0x2f0
#7: ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
#8: ffffffff865a4920 (console_lock){+.+.}, at:
vprintk_emit+0x100/0x340
stack backtrace:
Hardware name: HPE ProLiant DL560 Gen10/ProLiant DL560 Gen10,
BIOS U34 05/21/2019
Call Trace:
dump_stack+0x86/0xca
print_circular_bug.cold.31+0x243/0x26e
check_noncircular+0x29e/0x2e0
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Link: http://lkml.kernel.org/r/20200117181200.20299-1-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-31 06:14:57 +00:00
|
|
|
* cannot get removed (e.g., via memory unplug) concurrently.
|
|
|
|
*
|
2007-10-16 08:26:11 +00:00
|
|
|
*/
|
mm/hotplug: silence a lockdep splat with printk()
It is not that hard to trigger lockdep splats by calling printk from
under zone->lock. Most of them are false positives caused by lock
chains introduced early in the boot process and they do not cause any
real problems (although most of the early boot lock dependencies could
happen after boot as well). There are some console drivers which do
allocate from the printk context as well and those should be fixed. In
any case, false positives are not that trivial to workaround and it is
far from optimal to lose lockdep functionality for something that is a
non-issue.
So change has_unmovable_pages() so that it no longer calls dump_page()
itself - instead it returns a "struct page *" of the unmovable page back
to the caller so that in the case of a has_unmovable_pages() failure,
the caller can call dump_page() after releasing zone->lock. Also, make
dump_page() is able to report a CMA page as well, so the reason string
from has_unmovable_pages() can be removed.
Even though has_unmovable_pages doesn't hold any reference to the
returned page this should be reasonably safe for the purpose of
reporting the page (dump_page) because it cannot be hotremoved in the
context of memory unplug. The state of the page might change but that
is the case even with the existing code as zone->lock only plays role
for free pages.
While at it, remove a similar but unnecessary debug-only printk() as
well. A sample of one of those lockdep splats is,
WARNING: possible circular locking dependency detected
------------------------------------------------------
test.sh/8653 is trying to acquire lock:
ffffffff865a4460 (console_owner){-.-.}, at:
console_unlock+0x207/0x750
but task is already holding lock:
ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #3 (&(&zone->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock+0x2f/0x40
rmqueue_bulk.constprop.21+0xb6/0x1160
get_page_from_freelist+0x898/0x22c0
__alloc_pages_nodemask+0x2f3/0x1cd0
alloc_pages_current+0x9c/0x110
allocate_slab+0x4c6/0x19c0
new_slab+0x46/0x70
___slab_alloc+0x58b/0x960
__slab_alloc+0x43/0x70
__kmalloc+0x3ad/0x4b0
__tty_buffer_request_room+0x100/0x250
tty_insert_flip_string_fixed_flag+0x67/0x110
pty_write+0xa2/0xf0
n_tty_write+0x36b/0x7b0
tty_write+0x284/0x4c0
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
redirected_tty_write+0x6a/0xc0
do_iter_write+0x248/0x2a0
vfs_writev+0x106/0x1e0
do_writev+0xd4/0x180
__x64_sys_writev+0x45/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #2 (&(&port->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
tty_port_tty_get+0x20/0x60
tty_port_default_wakeup+0xf/0x30
tty_port_tty_wakeup+0x39/0x40
uart_write_wakeup+0x2a/0x40
serial8250_tx_chars+0x22e/0x440
serial8250_handle_irq.part.8+0x14a/0x170
serial8250_default_handle_irq+0x5c/0x90
serial8250_interrupt+0xa6/0x130
__handle_irq_event_percpu+0x78/0x4f0
handle_irq_event_percpu+0x70/0x100
handle_irq_event+0x5a/0x8b
handle_edge_irq+0x117/0x370
do_IRQ+0x9e/0x1e0
ret_from_intr+0x0/0x2a
cpuidle_enter_state+0x156/0x8e0
cpuidle_enter+0x41/0x70
call_cpuidle+0x5e/0x90
do_idle+0x333/0x370
cpu_startup_entry+0x1d/0x1f
start_secondary+0x290/0x330
secondary_startup_64+0xb6/0xc0
-> #1 (&port_lock_key){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
serial8250_console_write+0x3e4/0x450
univ8250_console_write+0x4b/0x60
console_unlock+0x501/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
-> #0 (console_owner){-.-.}:
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
other info that might help us debug this:
Chain exists of:
console_owner --> &(&port->lock)->rlock --> &(&zone->lock)->rlock
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&(&zone->lock)->rlock);
lock(&(&port->lock)->rlock);
lock(&(&zone->lock)->rlock);
lock(console_owner);
*** DEADLOCK ***
9 locks held by test.sh/8653:
#0: ffff88839ba7d408 (sb_writers#4){.+.+}, at:
vfs_write+0x25f/0x290
#1: ffff888277618880 (&of->mutex){+.+.}, at:
kernfs_fop_write+0x128/0x240
#2: ffff8898131fc218 (kn->count#115){.+.+}, at:
kernfs_fop_write+0x138/0x240
#3: ffffffff86962a80 (device_hotplug_lock){+.+.}, at:
lock_device_hotplug_sysfs+0x16/0x50
#4: ffff8884374f4990 (&dev->mutex){....}, at:
device_offline+0x70/0x110
#5: ffffffff86515250 (cpu_hotplug_lock.rw_sem){++++}, at:
__offline_pages+0xbf/0xa10
#6: ffffffff867405f0 (mem_hotplug_lock.rw_sem){++++}, at:
percpu_down_write+0x87/0x2f0
#7: ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
#8: ffffffff865a4920 (console_lock){+.+.}, at:
vprintk_emit+0x100/0x340
stack backtrace:
Hardware name: HPE ProLiant DL560 Gen10/ProLiant DL560 Gen10,
BIOS U34 05/21/2019
Call Trace:
dump_stack+0x86/0xca
print_circular_bug.cold.31+0x243/0x26e
check_noncircular+0x29e/0x2e0
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Link: http://lkml.kernel.org/r/20200117181200.20299-1-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-31 06:14:57 +00:00
|
|
|
struct page *has_unmovable_pages(struct zone *zone, struct page *page,
|
|
|
|
int migratetype, int flags)
|
2010-10-26 21:21:30 +00:00
|
|
|
{
|
2019-04-19 00:50:30 +00:00
|
|
|
unsigned long iter = 0;
|
|
|
|
unsigned long pfn = page_to_pfn(page);
|
2020-10-13 23:55:39 +00:00
|
|
|
unsigned long offset = pfn % pageblock_nr_pages;
|
2011-12-29 12:09:50 +00:00
|
|
|
|
2019-04-19 00:50:30 +00:00
|
|
|
if (is_migrate_cma_page(page)) {
|
|
|
|
/*
|
|
|
|
* CMA allocations (alloc_contig_range) really need to mark
|
|
|
|
* isolate CMA pageblocks even when they are not movable in fact
|
|
|
|
* so consider them movable here.
|
|
|
|
*/
|
|
|
|
if (is_migrate_cma(migratetype))
|
mm/hotplug: silence a lockdep splat with printk()
It is not that hard to trigger lockdep splats by calling printk from
under zone->lock. Most of them are false positives caused by lock
chains introduced early in the boot process and they do not cause any
real problems (although most of the early boot lock dependencies could
happen after boot as well). There are some console drivers which do
allocate from the printk context as well and those should be fixed. In
any case, false positives are not that trivial to workaround and it is
far from optimal to lose lockdep functionality for something that is a
non-issue.
So change has_unmovable_pages() so that it no longer calls dump_page()
itself - instead it returns a "struct page *" of the unmovable page back
to the caller so that in the case of a has_unmovable_pages() failure,
the caller can call dump_page() after releasing zone->lock. Also, make
dump_page() is able to report a CMA page as well, so the reason string
from has_unmovable_pages() can be removed.
Even though has_unmovable_pages doesn't hold any reference to the
returned page this should be reasonably safe for the purpose of
reporting the page (dump_page) because it cannot be hotremoved in the
context of memory unplug. The state of the page might change but that
is the case even with the existing code as zone->lock only plays role
for free pages.
While at it, remove a similar but unnecessary debug-only printk() as
well. A sample of one of those lockdep splats is,
WARNING: possible circular locking dependency detected
------------------------------------------------------
test.sh/8653 is trying to acquire lock:
ffffffff865a4460 (console_owner){-.-.}, at:
console_unlock+0x207/0x750
but task is already holding lock:
ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #3 (&(&zone->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock+0x2f/0x40
rmqueue_bulk.constprop.21+0xb6/0x1160
get_page_from_freelist+0x898/0x22c0
__alloc_pages_nodemask+0x2f3/0x1cd0
alloc_pages_current+0x9c/0x110
allocate_slab+0x4c6/0x19c0
new_slab+0x46/0x70
___slab_alloc+0x58b/0x960
__slab_alloc+0x43/0x70
__kmalloc+0x3ad/0x4b0
__tty_buffer_request_room+0x100/0x250
tty_insert_flip_string_fixed_flag+0x67/0x110
pty_write+0xa2/0xf0
n_tty_write+0x36b/0x7b0
tty_write+0x284/0x4c0
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
redirected_tty_write+0x6a/0xc0
do_iter_write+0x248/0x2a0
vfs_writev+0x106/0x1e0
do_writev+0xd4/0x180
__x64_sys_writev+0x45/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #2 (&(&port->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
tty_port_tty_get+0x20/0x60
tty_port_default_wakeup+0xf/0x30
tty_port_tty_wakeup+0x39/0x40
uart_write_wakeup+0x2a/0x40
serial8250_tx_chars+0x22e/0x440
serial8250_handle_irq.part.8+0x14a/0x170
serial8250_default_handle_irq+0x5c/0x90
serial8250_interrupt+0xa6/0x130
__handle_irq_event_percpu+0x78/0x4f0
handle_irq_event_percpu+0x70/0x100
handle_irq_event+0x5a/0x8b
handle_edge_irq+0x117/0x370
do_IRQ+0x9e/0x1e0
ret_from_intr+0x0/0x2a
cpuidle_enter_state+0x156/0x8e0
cpuidle_enter+0x41/0x70
call_cpuidle+0x5e/0x90
do_idle+0x333/0x370
cpu_startup_entry+0x1d/0x1f
start_secondary+0x290/0x330
secondary_startup_64+0xb6/0xc0
-> #1 (&port_lock_key){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
serial8250_console_write+0x3e4/0x450
univ8250_console_write+0x4b/0x60
console_unlock+0x501/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
-> #0 (console_owner){-.-.}:
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
other info that might help us debug this:
Chain exists of:
console_owner --> &(&port->lock)->rlock --> &(&zone->lock)->rlock
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&(&zone->lock)->rlock);
lock(&(&port->lock)->rlock);
lock(&(&zone->lock)->rlock);
lock(console_owner);
*** DEADLOCK ***
9 locks held by test.sh/8653:
#0: ffff88839ba7d408 (sb_writers#4){.+.+}, at:
vfs_write+0x25f/0x290
#1: ffff888277618880 (&of->mutex){+.+.}, at:
kernfs_fop_write+0x128/0x240
#2: ffff8898131fc218 (kn->count#115){.+.+}, at:
kernfs_fop_write+0x138/0x240
#3: ffffffff86962a80 (device_hotplug_lock){+.+.}, at:
lock_device_hotplug_sysfs+0x16/0x50
#4: ffff8884374f4990 (&dev->mutex){....}, at:
device_offline+0x70/0x110
#5: ffffffff86515250 (cpu_hotplug_lock.rw_sem){++++}, at:
__offline_pages+0xbf/0xa10
#6: ffffffff867405f0 (mem_hotplug_lock.rw_sem){++++}, at:
percpu_down_write+0x87/0x2f0
#7: ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
#8: ffffffff865a4920 (console_lock){+.+.}, at:
vprintk_emit+0x100/0x340
stack backtrace:
Hardware name: HPE ProLiant DL560 Gen10/ProLiant DL560 Gen10,
BIOS U34 05/21/2019
Call Trace:
dump_stack+0x86/0xca
print_circular_bug.cold.31+0x243/0x26e
check_noncircular+0x29e/0x2e0
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Link: http://lkml.kernel.org/r/20200117181200.20299-1-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-31 06:14:57 +00:00
|
|
|
return NULL;
|
2019-04-19 00:50:30 +00:00
|
|
|
|
2020-01-31 06:15:01 +00:00
|
|
|
return page;
|
2019-04-19 00:50:30 +00:00
|
|
|
}
|
2017-11-16 01:33:26 +00:00
|
|
|
|
2020-10-13 23:55:39 +00:00
|
|
|
for (; iter < pageblock_nr_pages - offset; iter++) {
|
2020-01-31 06:14:04 +00:00
|
|
|
if (!pfn_valid_within(pfn + iter))
|
2010-10-26 21:21:30 +00:00
|
|
|
continue;
|
2011-02-25 22:44:25 +00:00
|
|
|
|
2020-01-31 06:14:04 +00:00
|
|
|
page = pfn_to_page(pfn + iter);
|
2013-09-11 21:22:09 +00:00
|
|
|
|
mm/page_alloc: tweak comments in has_unmovable_pages()
Patch series "mm / virtio-mem: support ZONE_MOVABLE", v5.
When introducing virtio-mem, the semantics of ZONE_MOVABLE were rather
unclear, which is why we special-cased ZONE_MOVABLE such that partially
plugged blocks would never end up in ZONE_MOVABLE.
Now that the semantics are much clearer (and are documented in patch #6),
let's support partially plugged memory blocks in ZONE_MOVABLE, allowing
partially plugged memory blocks to be online to ZONE_MOVABLE and also
unplugging from such memory blocks. This avoids surprises when onlining
of memory blocks suddenly fails, just because they are not completely
populated by virtio-mem (yet).
This is especially helpful for testing, but also paves the way for
virtio-mem optimizations, allowing more memory to get reliably unplugged.
Cleanup has_unmovable_pages() and set_migratetype_isolate(), providing
better documentation of how ZONE_MOVABLE interacts with different kind of
unmovable pages (memory offlining vs. alloc_contig_range()).
This patch (of 6):
Let's move the split comment regarding bootmem allocations and memory
holes, especially in the context of ZONE_MOVABLE, to the PageReserved()
check.
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Baoquan He <bhe@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Qian Cai <cai@lca.pw>
Link: http://lkml.kernel.org/r/20200816125333.7434-1-david@redhat.com
Link: http://lkml.kernel.org/r/20200816125333.7434-2-david@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-13 23:55:17 +00:00
|
|
|
/*
|
|
|
|
* Both, bootmem allocations and memory holes are marked
|
|
|
|
* PG_reserved and are unmovable. We can even have unmovable
|
|
|
|
* allocations inside ZONE_MOVABLE, for example when
|
|
|
|
* specifying "movablecore".
|
|
|
|
*/
|
2017-11-16 01:33:30 +00:00
|
|
|
if (PageReserved(page))
|
2020-01-31 06:15:01 +00:00
|
|
|
return page;
|
2017-11-16 01:33:30 +00:00
|
|
|
|
2018-11-16 23:08:15 +00:00
|
|
|
/*
|
|
|
|
* If the zone is movable and we have ruled out all reserved
|
|
|
|
* pages then it should be reasonably safe to assume the rest
|
|
|
|
* is movable.
|
|
|
|
*/
|
|
|
|
if (zone_idx(zone) == ZONE_MOVABLE)
|
|
|
|
continue;
|
|
|
|
|
2013-09-11 21:22:09 +00:00
|
|
|
/*
|
|
|
|
* Hugepages are not in LRU lists, but they're movable.
|
2020-04-02 04:10:31 +00:00
|
|
|
* THPs are on the LRU, but need to be counted as #small pages.
|
2019-03-05 23:46:22 +00:00
|
|
|
* We need not scan over tail pages because we don't
|
2013-09-11 21:22:09 +00:00
|
|
|
* handle each tail page individually in migration.
|
|
|
|
*/
|
2020-04-02 04:10:31 +00:00
|
|
|
if (PageHuge(page) || PageTransCompound(page)) {
|
mm, page_alloc: fix has_unmovable_pages for HugePages
While playing with gigantic hugepages and memory_hotplug, I triggered
the following #PF when "cat memoryX/removable":
BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
#PF error: [normal kernel read fault]
PGD 0 P4D 0
Oops: 0000 [#1] SMP PTI
CPU: 1 PID: 1481 Comm: cat Tainted: G E 4.20.0-rc6-mm1-1-default+ #18
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014
RIP: 0010:has_unmovable_pages+0x154/0x210
Call Trace:
is_mem_section_removable+0x7d/0x100
removable_show+0x90/0xb0
dev_attr_show+0x1c/0x50
sysfs_kf_seq_show+0xca/0x1b0
seq_read+0x133/0x380
__vfs_read+0x26/0x180
vfs_read+0x89/0x140
ksys_read+0x42/0x90
do_syscall_64+0x5b/0x180
entry_SYSCALL_64_after_hwframe+0x44/0xa9
The reason is we do not pass the Head to page_hstate(), and so, the call
to compound_order() in page_hstate() returns 0, so we end up checking
all hstates's size to match PAGE_SIZE.
Obviously, we do not find any hstate matching that size, and we return
NULL. Then, we dereference that NULL pointer in
hugepage_migration_supported() and we got the #PF from above.
Fix that by getting the head page before calling page_hstate().
Also, since gigantic pages span several pageblocks, re-adjust the logic
for skipping pages. While are it, we can also get rid of the
round_up().
[osalvador@suse.de: remove round_up(), adjust skip pages logic per Michal]
Link: http://lkml.kernel.org/r/20181221062809.31771-1-osalvador@suse.de
Link: http://lkml.kernel.org/r/20181217225113.17864-1-osalvador@suse.de
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-21 22:31:00 +00:00
|
|
|
struct page *head = compound_head(page);
|
|
|
|
unsigned int skip_pages;
|
2018-09-04 22:45:59 +00:00
|
|
|
|
2020-04-02 04:10:31 +00:00
|
|
|
if (PageHuge(page)) {
|
|
|
|
if (!hugepage_migration_supported(page_hstate(head)))
|
|
|
|
return page;
|
|
|
|
} else if (!PageLRU(head) && !__PageMovable(head)) {
|
2020-01-31 06:15:01 +00:00
|
|
|
return page;
|
2020-04-02 04:10:31 +00:00
|
|
|
}
|
2018-09-04 22:45:59 +00:00
|
|
|
|
2019-09-23 22:34:30 +00:00
|
|
|
skip_pages = compound_nr(head) - (page - head);
|
mm, page_alloc: fix has_unmovable_pages for HugePages
While playing with gigantic hugepages and memory_hotplug, I triggered
the following #PF when "cat memoryX/removable":
BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
#PF error: [normal kernel read fault]
PGD 0 P4D 0
Oops: 0000 [#1] SMP PTI
CPU: 1 PID: 1481 Comm: cat Tainted: G E 4.20.0-rc6-mm1-1-default+ #18
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014
RIP: 0010:has_unmovable_pages+0x154/0x210
Call Trace:
is_mem_section_removable+0x7d/0x100
removable_show+0x90/0xb0
dev_attr_show+0x1c/0x50
sysfs_kf_seq_show+0xca/0x1b0
seq_read+0x133/0x380
__vfs_read+0x26/0x180
vfs_read+0x89/0x140
ksys_read+0x42/0x90
do_syscall_64+0x5b/0x180
entry_SYSCALL_64_after_hwframe+0x44/0xa9
The reason is we do not pass the Head to page_hstate(), and so, the call
to compound_order() in page_hstate() returns 0, so we end up checking
all hstates's size to match PAGE_SIZE.
Obviously, we do not find any hstate matching that size, and we return
NULL. Then, we dereference that NULL pointer in
hugepage_migration_supported() and we got the #PF from above.
Fix that by getting the head page before calling page_hstate().
Also, since gigantic pages span several pageblocks, re-adjust the logic
for skipping pages. While are it, we can also get rid of the
round_up().
[osalvador@suse.de: remove round_up(), adjust skip pages logic per Michal]
Link: http://lkml.kernel.org/r/20181221062809.31771-1-osalvador@suse.de
Link: http://lkml.kernel.org/r/20181217225113.17864-1-osalvador@suse.de
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-21 22:31:00 +00:00
|
|
|
iter += skip_pages - 1;
|
2013-09-11 21:22:09 +00:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2012-07-31 23:42:59 +00:00
|
|
|
/*
|
|
|
|
* We can't use page_count without pin a page
|
|
|
|
* because another CPU can free compound page.
|
|
|
|
* This check already skips compound tails of THP
|
2016-05-20 00:10:49 +00:00
|
|
|
* because their page->_refcount is zero at all time.
|
2012-07-31 23:42:59 +00:00
|
|
|
*/
|
2016-03-17 21:19:26 +00:00
|
|
|
if (!page_ref_count(page)) {
|
2010-10-26 21:21:30 +00:00
|
|
|
if (PageBuddy(page))
|
2020-10-16 03:10:15 +00:00
|
|
|
iter += (1 << buddy_order(page)) - 1;
|
2010-10-26 21:21:30 +00:00
|
|
|
continue;
|
|
|
|
}
|
2012-07-31 23:42:59 +00:00
|
|
|
|
2012-12-12 00:00:45 +00:00
|
|
|
/*
|
|
|
|
* The HWPoisoned page may be not in buddy system, and
|
|
|
|
* page_count() is not 0.
|
|
|
|
*/
|
2019-12-01 01:54:07 +00:00
|
|
|
if ((flags & MEMORY_OFFLINE) && PageHWPoison(page))
|
2012-12-12 00:00:45 +00:00
|
|
|
continue;
|
|
|
|
|
mm: Allow to offline unmovable PageOffline() pages via MEM_GOING_OFFLINE
virtio-mem wants to allow to offline memory blocks of which some parts
were unplugged (allocated via alloc_contig_range()), especially, to later
offline and remove completely unplugged memory blocks. The important part
is that PageOffline() has to remain set until the section is offline, so
these pages will never get accessed (e.g., when dumping). The pages should
not be handed back to the buddy (which would require clearing PageOffline()
and result in issues if offlining fails and the pages are suddenly in the
buddy).
Let's allow to do that by allowing to isolate any PageOffline() page
when offlining. This way, we can reach the memory hotplug notifier
MEM_GOING_OFFLINE, where the driver can signal that he is fine with
offlining this page by dropping its reference count. PageOffline() pages
with a reference count of 0 can then be skipped when offlining the
pages (like if they were free, however they are not in the buddy).
Anybody who uses PageOffline() pages and does not agree to offline them
(e.g., Hyper-V balloon, XEN balloon, VMWare balloon for 2MB pages) will not
decrement the reference count and make offlining fail when trying to
migrate such an unmovable page. So there should be no observable change.
Same applies to balloon compaction users (movable PageOffline() pages), the
pages will simply be migrated.
Note 1: If offlining fails, a driver has to increment the reference
count again in MEM_CANCEL_OFFLINE.
Note 2: A driver that makes use of this has to be aware that re-onlining
the memory block has to be handled by hooking into onlining code
(online_page_callback_t), resetting the page PageOffline() and
not giving them to the buddy.
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Tested-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Juergen Gross <jgross@suse.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Anthony Yznaga <anthony.yznaga@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Pingfan Liu <kernelfans@gmail.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/20200507140139.17083-7-david@redhat.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
2020-05-07 14:01:30 +00:00
|
|
|
/*
|
|
|
|
* We treat all PageOffline() pages as movable when offlining
|
|
|
|
* to give drivers a chance to decrement their reference count
|
|
|
|
* in MEM_GOING_OFFLINE in order to indicate that these pages
|
|
|
|
* can be offlined as there are no direct references anymore.
|
|
|
|
* For actually unmovable PageOffline() where the driver does
|
|
|
|
* not support this, we will fail later when trying to actually
|
|
|
|
* move these pages that still have a reference count > 0.
|
|
|
|
* (false negatives in this function only)
|
|
|
|
*/
|
|
|
|
if ((flags & MEMORY_OFFLINE) && PageOffline(page))
|
|
|
|
continue;
|
|
|
|
|
2020-01-31 06:14:04 +00:00
|
|
|
if (__PageMovable(page) || PageLRU(page))
|
2017-02-24 22:57:39 +00:00
|
|
|
continue;
|
|
|
|
|
2010-10-26 21:21:30 +00:00
|
|
|
/*
|
mm: vmscan: invoke slab shrinkers from shrink_zone()
The slab shrinkers are currently invoked from the zonelist walkers in
kswapd, direct reclaim, and zone reclaim, all of which roughly gauge the
eligible LRU pages and assemble a nodemask to pass to NUMA-aware
shrinkers, which then again have to walk over the nodemask. This is
redundant code, extra runtime work, and fairly inaccurate when it comes to
the estimation of actually scannable LRU pages. The code duplication will
only get worse when making the shrinkers cgroup-aware and requiring them
to have out-of-band cgroup hierarchy walks as well.
Instead, invoke the shrinkers from shrink_zone(), which is where all
reclaimers end up, to avoid this duplication.
Take the count for eligible LRU pages out of get_scan_count(), which
considers many more factors than just the availability of swap space, like
zone_reclaimable_pages() currently does. Accumulate the number over all
visited lruvecs to get the per-zone value.
Some nodes have multiple zones due to memory addressing restrictions. To
avoid putting too much pressure on the shrinkers, only invoke them once
for each such node, using the class zone of the allocation as the pivot
zone.
For now, this integrates the slab shrinking better into the reclaim logic
and gets rid of duplicative invocations from kswapd, direct reclaim, and
zone reclaim. It also prepares for cgroup-awareness, allowing
memcg-capable shrinkers to be added at the lruvec level without much
duplication of both code and runtime work.
This changes kswapd behavior, which used to invoke the shrinkers for each
zone, but with scan ratios gathered from the entire node, resulting in
meaningless pressure quantities on multi-zone nodes.
Zone reclaim behavior also changes. It used to shrink slabs until the
same amount of pages were shrunk as were reclaimed from the LRUs. Now it
merely invokes the shrinkers once with the zone's scan ratio, which makes
the shrinkers go easier on caches that implement aging and would prefer
feeding back pressure from recently used slab objects to unused LRU pages.
[vdavydov@parallels.com: assure class zone is populated]
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dave Chinner <david@fromorbit.com>
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:13 +00:00
|
|
|
* If there are RECLAIMABLE pages, we need to check
|
|
|
|
* it. But now, memory offline itself doesn't call
|
|
|
|
* shrink_node_slabs() and it still to be fixed.
|
2010-10-26 21:21:30 +00:00
|
|
|
*/
|
2020-01-31 06:15:01 +00:00
|
|
|
return page;
|
2010-10-26 21:21:30 +00:00
|
|
|
}
|
mm/hotplug: silence a lockdep splat with printk()
It is not that hard to trigger lockdep splats by calling printk from
under zone->lock. Most of them are false positives caused by lock
chains introduced early in the boot process and they do not cause any
real problems (although most of the early boot lock dependencies could
happen after boot as well). There are some console drivers which do
allocate from the printk context as well and those should be fixed. In
any case, false positives are not that trivial to workaround and it is
far from optimal to lose lockdep functionality for something that is a
non-issue.
So change has_unmovable_pages() so that it no longer calls dump_page()
itself - instead it returns a "struct page *" of the unmovable page back
to the caller so that in the case of a has_unmovable_pages() failure,
the caller can call dump_page() after releasing zone->lock. Also, make
dump_page() is able to report a CMA page as well, so the reason string
from has_unmovable_pages() can be removed.
Even though has_unmovable_pages doesn't hold any reference to the
returned page this should be reasonably safe for the purpose of
reporting the page (dump_page) because it cannot be hotremoved in the
context of memory unplug. The state of the page might change but that
is the case even with the existing code as zone->lock only plays role
for free pages.
While at it, remove a similar but unnecessary debug-only printk() as
well. A sample of one of those lockdep splats is,
WARNING: possible circular locking dependency detected
------------------------------------------------------
test.sh/8653 is trying to acquire lock:
ffffffff865a4460 (console_owner){-.-.}, at:
console_unlock+0x207/0x750
but task is already holding lock:
ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #3 (&(&zone->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock+0x2f/0x40
rmqueue_bulk.constprop.21+0xb6/0x1160
get_page_from_freelist+0x898/0x22c0
__alloc_pages_nodemask+0x2f3/0x1cd0
alloc_pages_current+0x9c/0x110
allocate_slab+0x4c6/0x19c0
new_slab+0x46/0x70
___slab_alloc+0x58b/0x960
__slab_alloc+0x43/0x70
__kmalloc+0x3ad/0x4b0
__tty_buffer_request_room+0x100/0x250
tty_insert_flip_string_fixed_flag+0x67/0x110
pty_write+0xa2/0xf0
n_tty_write+0x36b/0x7b0
tty_write+0x284/0x4c0
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
redirected_tty_write+0x6a/0xc0
do_iter_write+0x248/0x2a0
vfs_writev+0x106/0x1e0
do_writev+0xd4/0x180
__x64_sys_writev+0x45/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #2 (&(&port->lock)->rlock){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
tty_port_tty_get+0x20/0x60
tty_port_default_wakeup+0xf/0x30
tty_port_tty_wakeup+0x39/0x40
uart_write_wakeup+0x2a/0x40
serial8250_tx_chars+0x22e/0x440
serial8250_handle_irq.part.8+0x14a/0x170
serial8250_default_handle_irq+0x5c/0x90
serial8250_interrupt+0xa6/0x130
__handle_irq_event_percpu+0x78/0x4f0
handle_irq_event_percpu+0x70/0x100
handle_irq_event+0x5a/0x8b
handle_edge_irq+0x117/0x370
do_IRQ+0x9e/0x1e0
ret_from_intr+0x0/0x2a
cpuidle_enter_state+0x156/0x8e0
cpuidle_enter+0x41/0x70
call_cpuidle+0x5e/0x90
do_idle+0x333/0x370
cpu_startup_entry+0x1d/0x1f
start_secondary+0x290/0x330
secondary_startup_64+0xb6/0xc0
-> #1 (&port_lock_key){-.-.}:
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
_raw_spin_lock_irqsave+0x3a/0x50
serial8250_console_write+0x3e4/0x450
univ8250_console_write+0x4b/0x60
console_unlock+0x501/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
-> #0 (console_owner){-.-.}:
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
other info that might help us debug this:
Chain exists of:
console_owner --> &(&port->lock)->rlock --> &(&zone->lock)->rlock
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&(&zone->lock)->rlock);
lock(&(&port->lock)->rlock);
lock(&(&zone->lock)->rlock);
lock(console_owner);
*** DEADLOCK ***
9 locks held by test.sh/8653:
#0: ffff88839ba7d408 (sb_writers#4){.+.+}, at:
vfs_write+0x25f/0x290
#1: ffff888277618880 (&of->mutex){+.+.}, at:
kernfs_fop_write+0x128/0x240
#2: ffff8898131fc218 (kn->count#115){.+.+}, at:
kernfs_fop_write+0x138/0x240
#3: ffffffff86962a80 (device_hotplug_lock){+.+.}, at:
lock_device_hotplug_sysfs+0x16/0x50
#4: ffff8884374f4990 (&dev->mutex){....}, at:
device_offline+0x70/0x110
#5: ffffffff86515250 (cpu_hotplug_lock.rw_sem){++++}, at:
__offline_pages+0xbf/0xa10
#6: ffffffff867405f0 (mem_hotplug_lock.rw_sem){++++}, at:
percpu_down_write+0x87/0x2f0
#7: ffff88883fff3c58 (&(&zone->lock)->rlock){-.-.}, at:
__offline_isolated_pages+0x179/0x3e0
#8: ffffffff865a4920 (console_lock){+.+.}, at:
vprintk_emit+0x100/0x340
stack backtrace:
Hardware name: HPE ProLiant DL560 Gen10/ProLiant DL560 Gen10,
BIOS U34 05/21/2019
Call Trace:
dump_stack+0x86/0xca
print_circular_bug.cold.31+0x243/0x26e
check_noncircular+0x29e/0x2e0
check_prev_add+0x107/0xea0
validate_chain+0x8fc/0x1200
__lock_acquire+0x5b3/0xb40
lock_acquire+0x126/0x280
console_unlock+0x269/0x750
vprintk_emit+0x10d/0x340
vprintk_default+0x1f/0x30
vprintk_func+0x44/0xd4
printk+0x9f/0xc5
__offline_isolated_pages.cold.52+0x2f/0x30a
offline_isolated_pages_cb+0x17/0x30
walk_system_ram_range+0xda/0x160
__offline_pages+0x79c/0xa10
offline_pages+0x11/0x20
memory_subsys_offline+0x7e/0xc0
device_offline+0xd5/0x110
state_store+0xc6/0xe0
dev_attr_store+0x3f/0x60
sysfs_kf_write+0x89/0xb0
kernfs_fop_write+0x188/0x240
__vfs_write+0x50/0xa0
vfs_write+0x105/0x290
ksys_write+0xc6/0x160
__x64_sys_write+0x43/0x50
do_syscall_64+0xcc/0x76c
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Link: http://lkml.kernel.org/r/20200117181200.20299-1-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-01-31 06:14:57 +00:00
|
|
|
return NULL;
|
2010-10-26 21:21:30 +00:00
|
|
|
}
|
|
|
|
|
2019-05-14 00:19:00 +00:00
|
|
|
#ifdef CONFIG_CONTIG_ALLOC
|
2011-12-29 12:09:50 +00:00
|
|
|
static unsigned long pfn_max_align_down(unsigned long pfn)
|
|
|
|
{
|
|
|
|
return pfn & ~(max_t(unsigned long, MAX_ORDER_NR_PAGES,
|
|
|
|
pageblock_nr_pages) - 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
static unsigned long pfn_max_align_up(unsigned long pfn)
|
|
|
|
{
|
|
|
|
return ALIGN(pfn, max_t(unsigned long, MAX_ORDER_NR_PAGES,
|
|
|
|
pageblock_nr_pages));
|
|
|
|
}
|
|
|
|
|
|
|
|
/* [start, end) must belong to a single zone. */
|
2012-10-08 23:32:41 +00:00
|
|
|
static int __alloc_contig_migrate_range(struct compact_control *cc,
|
|
|
|
unsigned long start, unsigned long end)
|
2011-12-29 12:09:50 +00:00
|
|
|
{
|
|
|
|
/* This function is based on compact_zone() from compaction.c. */
|
2020-06-03 23:01:18 +00:00
|
|
|
unsigned int nr_reclaimed;
|
2011-12-29 12:09:50 +00:00
|
|
|
unsigned long pfn = start;
|
|
|
|
unsigned int tries = 0;
|
|
|
|
int ret = 0;
|
2020-08-12 01:37:31 +00:00
|
|
|
struct migration_target_control mtc = {
|
|
|
|
.nid = zone_to_nid(cc->zone),
|
|
|
|
.gfp_mask = GFP_USER | __GFP_MOVABLE | __GFP_RETRY_MAYFAIL,
|
|
|
|
};
|
2011-12-29 12:09:50 +00:00
|
|
|
|
2012-12-12 21:51:19 +00:00
|
|
|
migrate_prep();
|
2011-12-29 12:09:50 +00:00
|
|
|
|
2012-10-08 23:32:41 +00:00
|
|
|
while (pfn < end || !list_empty(&cc->migratepages)) {
|
2011-12-29 12:09:50 +00:00
|
|
|
if (fatal_signal_pending(current)) {
|
|
|
|
ret = -EINTR;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2012-10-08 23:32:41 +00:00
|
|
|
if (list_empty(&cc->migratepages)) {
|
|
|
|
cc->nr_migratepages = 0;
|
mm, compaction: move pageblock checks up from isolate_migratepages_range()
isolate_migratepages_range() is the main function of the compaction
scanner, called either on a single pageblock by isolate_migratepages()
during regular compaction, or on an arbitrary range by CMA's
__alloc_contig_migrate_range(). It currently perfoms two pageblock-wide
compaction suitability checks, and because of the CMA callpath, it tracks
if it crossed a pageblock boundary in order to repeat those checks.
However, closer inspection shows that those checks are always true for CMA:
- isolation_suitable() is true because CMA sets cc->ignore_skip_hint to true
- migrate_async_suitable() check is skipped because CMA uses sync compaction
We can therefore move the compaction-specific checks to
isolate_migratepages() and simplify isolate_migratepages_range().
Furthermore, we can mimic the freepage scanner family of functions, which
has isolate_freepages_block() function called both by compaction from
isolate_freepages() and by CMA from isolate_freepages_range(), where each
use-case adds own specific glue code. This allows further code
simplification.
Thus, we rename isolate_migratepages_range() to
isolate_migratepages_block() and limit its functionality to a single
pageblock (or its subset). For CMA, a new different
isolate_migratepages_range() is created as a CMA-specific wrapper for the
_block() function. The checks specific to compaction are moved to
isolate_migratepages(). As part of the unification of these two families
of functions, we remove the redundant zone parameter where applicable,
since zone pointer is already passed in cc->zone.
Furthermore, going back to compact_zone() and compact_finished() when
pageblock is found unsuitable (now by isolate_migratepages()) is wasteful
- the checks are meant to skip pageblocks quickly. The patch therefore
also introduces a simple loop into isolate_migratepages() so that it does
not return immediately on failed pageblock checks, but keeps going until
isolate_migratepages_range() gets called once. Similarily to
isolate_freepages(), the function periodically checks if it needs to
reschedule or abort async compaction.
[iamjoonsoo.kim@lge.com: fix isolated page counting bug in compaction]
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Minchan Kim <minchan@kernel.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-09 22:27:09 +00:00
|
|
|
pfn = isolate_migratepages_range(cc, pfn, end);
|
2011-12-29 12:09:50 +00:00
|
|
|
if (!pfn) {
|
|
|
|
ret = -EINTR;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
tries = 0;
|
|
|
|
} else if (++tries == 5) {
|
|
|
|
ret = ret < 0 ? ret : -EBUSY;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2012-10-08 23:33:51 +00:00
|
|
|
nr_reclaimed = reclaim_clean_pages_from_list(cc->zone,
|
|
|
|
&cc->migratepages);
|
|
|
|
cc->nr_migratepages -= nr_reclaimed;
|
2012-10-08 23:31:55 +00:00
|
|
|
|
2020-08-12 01:37:31 +00:00
|
|
|
ret = migrate_pages(&cc->migratepages, alloc_migration_target,
|
|
|
|
NULL, (unsigned long)&mtc, cc->mode, MR_CONTIG_RANGE);
|
2011-12-29 12:09:50 +00:00
|
|
|
}
|
2013-02-23 00:32:09 +00:00
|
|
|
if (ret < 0) {
|
|
|
|
putback_movable_pages(&cc->migratepages);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
return 0;
|
2011-12-29 12:09:50 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* alloc_contig_range() -- tries to allocate given range of pages
|
|
|
|
* @start: start PFN to allocate
|
|
|
|
* @end: one-past-the-last PFN to allocate
|
2012-04-03 13:06:15 +00:00
|
|
|
* @migratetype: migratetype of the underlaying pageblocks (either
|
|
|
|
* #MIGRATE_MOVABLE or #MIGRATE_CMA). All pageblocks
|
|
|
|
* in range must have the same migratetype and it must
|
|
|
|
* be either of the two.
|
2017-02-24 22:58:37 +00:00
|
|
|
* @gfp_mask: GFP mask to use during compaction
|
2011-12-29 12:09:50 +00:00
|
|
|
*
|
|
|
|
* The PFN range does not have to be pageblock or MAX_ORDER_NR_PAGES
|
2018-04-05 23:25:26 +00:00
|
|
|
* aligned. The PFN range must belong to a single zone.
|
2011-12-29 12:09:50 +00:00
|
|
|
*
|
2018-04-05 23:25:26 +00:00
|
|
|
* The first thing this routine does is attempt to MIGRATE_ISOLATE all
|
|
|
|
* pageblocks in the range. Once isolated, the pageblocks should not
|
|
|
|
* be modified by others.
|
2011-12-29 12:09:50 +00:00
|
|
|
*
|
2019-03-05 23:48:42 +00:00
|
|
|
* Return: zero on success or negative error code. On success all
|
2011-12-29 12:09:50 +00:00
|
|
|
* pages which PFN is in [start, end) are allocated for the caller and
|
|
|
|
* need to be freed with free_contig_range().
|
|
|
|
*/
|
2012-04-03 13:06:15 +00:00
|
|
|
int alloc_contig_range(unsigned long start, unsigned long end,
|
2017-02-24 22:58:37 +00:00
|
|
|
unsigned migratetype, gfp_t gfp_mask)
|
2011-12-29 12:09:50 +00:00
|
|
|
{
|
|
|
|
unsigned long outer_start, outer_end;
|
2015-11-07 00:29:57 +00:00
|
|
|
unsigned int order;
|
|
|
|
int ret = 0;
|
2011-12-29 12:09:50 +00:00
|
|
|
|
2012-10-08 23:32:41 +00:00
|
|
|
struct compact_control cc = {
|
|
|
|
.nr_migratepages = 0,
|
|
|
|
.order = -1,
|
|
|
|
.zone = page_zone(pfn_to_page(start)),
|
2014-06-04 23:08:28 +00:00
|
|
|
.mode = MIGRATE_SYNC,
|
2012-10-08 23:32:41 +00:00
|
|
|
.ignore_skip_hint = true,
|
2017-11-17 23:26:38 +00:00
|
|
|
.no_set_skip_hint = true,
|
mm: introduce memalloc_nofs_{save,restore} API
GFP_NOFS context is used for the following 5 reasons currently:
- to prevent from deadlocks when the lock held by the allocation
context would be needed during the memory reclaim
- to prevent from stack overflows during the reclaim because the
allocation is performed from a deep context already
- to prevent lockups when the allocation context depends on other
reclaimers to make a forward progress indirectly
- just in case because this would be safe from the fs POV
- silence lockdep false positives
Unfortunately overuse of this allocation context brings some problems to
the MM. Memory reclaim is much weaker (especially during heavy FS
metadata workloads), OOM killer cannot be invoked because the MM layer
doesn't have enough information about how much memory is freeable by the
FS layer.
In many cases it is far from clear why the weaker context is even used
and so it might be used unnecessarily. We would like to get rid of
those as much as possible. One way to do that is to use the flag in
scopes rather than isolated cases. Such a scope is declared when really
necessary, tracked per task and all the allocation requests from within
the context will simply inherit the GFP_NOFS semantic.
Not only this is easier to understand and maintain because there are
much less problematic contexts than specific allocation requests, this
also helps code paths where FS layer interacts with other layers (e.g.
crypto, security modules, MM etc...) and there is no easy way to convey
the allocation context between the layers.
Introduce memalloc_nofs_{save,restore} API to control the scope of
GFP_NOFS allocation context. This is basically copying
memalloc_noio_{save,restore} API we have for other restricted allocation
context GFP_NOIO. The PF_MEMALLOC_NOFS flag already exists and it is
just an alias for PF_FSTRANS which has been xfs specific until recently.
There are no more PF_FSTRANS users anymore so let's just drop it.
PF_MEMALLOC_NOFS is now checked in the MM layer and drops __GFP_FS
implicitly same as PF_MEMALLOC_NOIO drops __GFP_IO. memalloc_noio_flags
is renamed to current_gfp_context because it now cares about both
PF_MEMALLOC_NOFS and PF_MEMALLOC_NOIO contexts. Xfs code paths preserve
their semantic. kmem_flags_convert() doesn't need to evaluate the flag
anymore.
This patch shouldn't introduce any functional changes.
Let's hope that filesystems will drop direct GFP_NOFS (resp. ~__GFP_FS)
usage as much as possible and only use a properly documented
memalloc_nofs_{save,restore} checkpoints where they are appropriate.
[akpm@linux-foundation.org: fix comment typo, reflow comment]
Link: http://lkml.kernel.org/r/20170306131408.9828-5-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-05-03 21:53:15 +00:00
|
|
|
.gfp_mask = current_gfp_context(gfp_mask),
|
2020-04-02 04:10:28 +00:00
|
|
|
.alloc_contig = true,
|
2012-10-08 23:32:41 +00:00
|
|
|
};
|
|
|
|
INIT_LIST_HEAD(&cc.migratepages);
|
|
|
|
|
2011-12-29 12:09:50 +00:00
|
|
|
/*
|
|
|
|
* What we do here is we mark all pageblocks in range as
|
|
|
|
* MIGRATE_ISOLATE. Because pageblock and max order pages may
|
|
|
|
* have different sizes, and due to the way page allocator
|
|
|
|
* work, we align the range to biggest of the two pages so
|
|
|
|
* that page allocator won't try to merge buddies from
|
|
|
|
* different pageblocks and change MIGRATE_ISOLATE to some
|
|
|
|
* other migration type.
|
|
|
|
*
|
|
|
|
* Once the pageblocks are marked as MIGRATE_ISOLATE, we
|
|
|
|
* migrate the pages from an unaligned range (ie. pages that
|
|
|
|
* we are interested in). This will put all the pages in
|
|
|
|
* range back to page allocator as MIGRATE_ISOLATE.
|
|
|
|
*
|
|
|
|
* When this is done, we take the pages in range from page
|
|
|
|
* allocator removing them from the buddy system. This way
|
|
|
|
* page allocator will never consider using them.
|
|
|
|
*
|
|
|
|
* This lets us mark the pageblocks back as
|
|
|
|
* MIGRATE_CMA/MIGRATE_MOVABLE so that free pages in the
|
|
|
|
* aligned range but not in the unaligned, original range are
|
|
|
|
* put back to page allocator so that buddy can use them.
|
|
|
|
*/
|
|
|
|
|
|
|
|
ret = start_isolate_page_range(pfn_max_align_down(start),
|
2018-12-28 08:33:56 +00:00
|
|
|
pfn_max_align_up(end), migratetype, 0);
|
2020-10-16 03:08:07 +00:00
|
|
|
if (ret)
|
2012-10-25 20:37:56 +00:00
|
|
|
return ret;
|
2011-12-29 12:09:50 +00:00
|
|
|
|
2020-12-15 03:10:56 +00:00
|
|
|
drain_all_pages(cc.zone);
|
|
|
|
|
mm/cma: always check which page caused allocation failure
Now, we have tracepoint in test_pages_isolated() to notify pfn which
cannot be isolated. But, in alloc_contig_range(), some error path
doesn't call test_pages_isolated() so it's still hard to know exact pfn
that causes allocation failure.
This patch change this situation by calling test_pages_isolated() in
almost error path. In allocation failure case, some overhead is added
by this change, but, allocation failure is really rare event so it would
not matter.
In fatal signal pending case, we don't call test_pages_isolated()
because this failure is intentional one.
There was a bogus outer_start problem due to unchecked buddy order and
this patch also fix it. Before this patch, it didn't matter, because
end result is same thing. But, after this patch, tracepoint will report
failed pfn so it should be accurate.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-14 23:18:45 +00:00
|
|
|
/*
|
|
|
|
* In case of -EBUSY, we'd like to know which page causes problem.
|
2017-11-30 00:10:01 +00:00
|
|
|
* So, just fall through. test_pages_isolated() has a tracepoint
|
|
|
|
* which will report the busy page.
|
|
|
|
*
|
|
|
|
* It is possible that busy pages could become available before
|
|
|
|
* the call to test_pages_isolated, and the range will actually be
|
|
|
|
* allocated. So, if we fall through be sure to clear ret so that
|
|
|
|
* -EBUSY is not accidentally used or returned to caller.
|
mm/cma: always check which page caused allocation failure
Now, we have tracepoint in test_pages_isolated() to notify pfn which
cannot be isolated. But, in alloc_contig_range(), some error path
doesn't call test_pages_isolated() so it's still hard to know exact pfn
that causes allocation failure.
This patch change this situation by calling test_pages_isolated() in
almost error path. In allocation failure case, some overhead is added
by this change, but, allocation failure is really rare event so it would
not matter.
In fatal signal pending case, we don't call test_pages_isolated()
because this failure is intentional one.
There was a bogus outer_start problem due to unchecked buddy order and
this patch also fix it. Before this patch, it didn't matter, because
end result is same thing. But, after this patch, tracepoint will report
failed pfn so it should be accurate.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-14 23:18:45 +00:00
|
|
|
*/
|
2012-10-08 23:32:41 +00:00
|
|
|
ret = __alloc_contig_migrate_range(&cc, start, end);
|
mm/cma: always check which page caused allocation failure
Now, we have tracepoint in test_pages_isolated() to notify pfn which
cannot be isolated. But, in alloc_contig_range(), some error path
doesn't call test_pages_isolated() so it's still hard to know exact pfn
that causes allocation failure.
This patch change this situation by calling test_pages_isolated() in
almost error path. In allocation failure case, some overhead is added
by this change, but, allocation failure is really rare event so it would
not matter.
In fatal signal pending case, we don't call test_pages_isolated()
because this failure is intentional one.
There was a bogus outer_start problem due to unchecked buddy order and
this patch also fix it. Before this patch, it didn't matter, because
end result is same thing. But, after this patch, tracepoint will report
failed pfn so it should be accurate.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-14 23:18:45 +00:00
|
|
|
if (ret && ret != -EBUSY)
|
2011-12-29 12:09:50 +00:00
|
|
|
goto done;
|
2017-11-30 00:10:01 +00:00
|
|
|
ret =0;
|
2011-12-29 12:09:50 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Pages from [start, end) are within a MAX_ORDER_NR_PAGES
|
|
|
|
* aligned blocks that are marked as MIGRATE_ISOLATE. What's
|
|
|
|
* more, all pages in [start, end) are free in page allocator.
|
|
|
|
* What we are going to do is to allocate all pages from
|
|
|
|
* [start, end) (that is remove them from page allocator).
|
|
|
|
*
|
|
|
|
* The only problem is that pages at the beginning and at the
|
|
|
|
* end of interesting range may be not aligned with pages that
|
|
|
|
* page allocator holds, ie. they can be part of higher order
|
|
|
|
* pages. Because of this, we reserve the bigger range and
|
|
|
|
* once this is done free the pages we are not interested in.
|
|
|
|
*
|
|
|
|
* We don't have to hold zone->lock here because the pages are
|
|
|
|
* isolated thus they won't get removed from buddy.
|
|
|
|
*/
|
|
|
|
|
|
|
|
lru_add_drain_all();
|
|
|
|
|
|
|
|
order = 0;
|
|
|
|
outer_start = start;
|
|
|
|
while (!PageBuddy(pfn_to_page(outer_start))) {
|
|
|
|
if (++order >= MAX_ORDER) {
|
mm/cma: always check which page caused allocation failure
Now, we have tracepoint in test_pages_isolated() to notify pfn which
cannot be isolated. But, in alloc_contig_range(), some error path
doesn't call test_pages_isolated() so it's still hard to know exact pfn
that causes allocation failure.
This patch change this situation by calling test_pages_isolated() in
almost error path. In allocation failure case, some overhead is added
by this change, but, allocation failure is really rare event so it would
not matter.
In fatal signal pending case, we don't call test_pages_isolated()
because this failure is intentional one.
There was a bogus outer_start problem due to unchecked buddy order and
this patch also fix it. Before this patch, it didn't matter, because
end result is same thing. But, after this patch, tracepoint will report
failed pfn so it should be accurate.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-14 23:18:45 +00:00
|
|
|
outer_start = start;
|
|
|
|
break;
|
2011-12-29 12:09:50 +00:00
|
|
|
}
|
|
|
|
outer_start &= ~0UL << order;
|
|
|
|
}
|
|
|
|
|
mm/cma: always check which page caused allocation failure
Now, we have tracepoint in test_pages_isolated() to notify pfn which
cannot be isolated. But, in alloc_contig_range(), some error path
doesn't call test_pages_isolated() so it's still hard to know exact pfn
that causes allocation failure.
This patch change this situation by calling test_pages_isolated() in
almost error path. In allocation failure case, some overhead is added
by this change, but, allocation failure is really rare event so it would
not matter.
In fatal signal pending case, we don't call test_pages_isolated()
because this failure is intentional one.
There was a bogus outer_start problem due to unchecked buddy order and
this patch also fix it. Before this patch, it didn't matter, because
end result is same thing. But, after this patch, tracepoint will report
failed pfn so it should be accurate.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-14 23:18:45 +00:00
|
|
|
if (outer_start != start) {
|
2020-10-16 03:10:15 +00:00
|
|
|
order = buddy_order(pfn_to_page(outer_start));
|
mm/cma: always check which page caused allocation failure
Now, we have tracepoint in test_pages_isolated() to notify pfn which
cannot be isolated. But, in alloc_contig_range(), some error path
doesn't call test_pages_isolated() so it's still hard to know exact pfn
that causes allocation failure.
This patch change this situation by calling test_pages_isolated() in
almost error path. In allocation failure case, some overhead is added
by this change, but, allocation failure is really rare event so it would
not matter.
In fatal signal pending case, we don't call test_pages_isolated()
because this failure is intentional one.
There was a bogus outer_start problem due to unchecked buddy order and
this patch also fix it. Before this patch, it didn't matter, because
end result is same thing. But, after this patch, tracepoint will report
failed pfn so it should be accurate.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-14 23:18:45 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* outer_start page could be small order buddy page and
|
|
|
|
* it doesn't include start page. Adjust outer_start
|
|
|
|
* in this case to report failed page properly
|
|
|
|
* on tracepoint in test_pages_isolated()
|
|
|
|
*/
|
|
|
|
if (outer_start + (1UL << order) <= start)
|
|
|
|
outer_start = start;
|
|
|
|
}
|
|
|
|
|
2011-12-29 12:09:50 +00:00
|
|
|
/* Make sure the range is really isolated. */
|
2019-12-01 01:54:07 +00:00
|
|
|
if (test_pages_isolated(outer_start, end, 0)) {
|
mm: ratelimit PFNs busy info message
The RDMA subsystem can generate several thousand of these messages per
second eventually leading to a kernel crash. Ratelimit these messages
to prevent this crash.
Doug said:
"I've been carrying a version of this for several kernel versions. I
don't remember when they started, but we have one (and only one) class
of machines: Dell PE R730xd, that generate these errors. When it
happens, without a rate limit, we get rcu timeouts and kernel oopses.
With the rate limit, we just get a lot of annoying kernel messages but
the machine continues on, recovers, and eventually the memory
operations all succeed"
And:
"> Well... why are all these EBUSY's occurring? It sounds inefficient
> (at least) but if it is expected, normal and unavoidable then
> perhaps we should just remove that message altogether?
I don't have an answer to that question. To be honest, I haven't
looked real hard. We never had this at all, then it started out of the
blue, but only on our Dell 730xd machines (and it hits all of them),
but no other classes or brands of machines. And we have our 730xd
machines loaded up with different brands and models of cards (for
instance one dedicated to mlx4 hardware, one for qib, one for mlx5, an
ocrdma/cxgb4 combo, etc), so the fact that it hit all of the machines
meant it wasn't tied to any particular brand/model of RDMA hardware.
To me, it always smelled of a hardware oddity specific to maybe the
CPUs or mainboard chipsets in these machines, so given that I'm not an
mm expert anyway, I never chased it down.
A few other relevant details: it showed up somewhere around 4.8/4.9 or
thereabouts. It never happened before, but the prinkt has been there
since the 3.18 days, so possibly the test to trigger this message was
changed, or something else in the allocator changed such that the
situation started happening on these machines?
And, like I said, it is specific to our 730xd machines (but they are
all identical, so that could mean it's something like their specific
ram configuration is causing the allocator to hit this on these
machine but not on other machines in the cluster, I don't want to say
it's necessarily the model of chipset or CPU, there are other bits of
identicalness between these machines)"
Link: http://lkml.kernel.org/r/499c0f6cc10d6eb829a67f2a4d75b4228a9b356e.1501695897.git.jtoppins@redhat.com
Signed-off-by: Jonathan Toppins <jtoppins@redhat.com>
Reviewed-by: Doug Ledford <dledford@redhat.com>
Tested-by: Doug Ledford <dledford@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-08-10 22:23:35 +00:00
|
|
|
pr_info_ratelimited("%s: [%lx, %lx) PFNs busy\n",
|
2014-11-13 23:19:27 +00:00
|
|
|
__func__, outer_start, end);
|
2011-12-29 12:09:50 +00:00
|
|
|
ret = -EBUSY;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
2012-01-25 11:49:24 +00:00
|
|
|
/* Grab isolated pages from freelists. */
|
2012-10-08 23:32:41 +00:00
|
|
|
outer_end = isolate_freepages_range(&cc, outer_start, end);
|
2011-12-29 12:09:50 +00:00
|
|
|
if (!outer_end) {
|
|
|
|
ret = -EBUSY;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Free head and tail (if any) */
|
|
|
|
if (start != outer_start)
|
|
|
|
free_contig_range(outer_start, start - outer_start);
|
|
|
|
if (end != outer_end)
|
|
|
|
free_contig_range(end, outer_end - end);
|
|
|
|
|
|
|
|
done:
|
|
|
|
undo_isolate_page_range(pfn_max_align_down(start),
|
2012-04-03 13:06:15 +00:00
|
|
|
pfn_max_align_up(end), migratetype);
|
2011-12-29 12:09:50 +00:00
|
|
|
return ret;
|
|
|
|
}
|
virtio-mem: Paravirtualized memory hotunplug part 2
We also want to unplug online memory (contained in online memory blocks
and, therefore, managed by the buddy), and eventually replug it later.
When requested to unplug memory, we use alloc_contig_range() to allocate
subblocks in online memory blocks (so we are the owner) and send them to
our hypervisor. When requested to plug memory, we can replug such memory
using free_contig_range() after asking our hypervisor.
We also want to mark all allocated pages PG_offline, so nobody will
touch them. To differentiate pages that were never onlined when
onlining the memory block from pages allocated via alloc_contig_range(), we
use PageDirty(). Based on this flag, virtio_mem_fake_online() can either
online the pages for the first time or use free_contig_range().
It is worth noting that there are no guarantees on how much memory can
actually get unplugged again. All device memory might completely be
fragmented with unmovable data, such that no subblock can get unplugged.
We are not touching the ZONE_MOVABLE. If memory is onlined to the
ZONE_MOVABLE, it can only get unplugged after that memory was offlined
manually by user space. In normal operation, virtio-mem memory is
suggested to be onlined to ZONE_NORMAL. In the future, we will try to
make unplug more likely to succeed.
Add a module parameter to control if online memory shall be touched.
As we want to access alloc_contig_range()/free_contig_range() from
kernel module context, export the symbols.
Note: Whenever virtio-mem uses alloc_contig_range(), all affected pages
are on the same node, in the same zone, and contain no holes.
Acked-by: Michal Hocko <mhocko@suse.com> # to export contig range allocator API
Tested-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/20200507140139.17083-6-david@redhat.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
2020-05-07 14:01:29 +00:00
|
|
|
EXPORT_SYMBOL(alloc_contig_range);
|
2019-12-01 01:55:06 +00:00
|
|
|
|
|
|
|
static int __alloc_contig_pages(unsigned long start_pfn,
|
|
|
|
unsigned long nr_pages, gfp_t gfp_mask)
|
|
|
|
{
|
|
|
|
unsigned long end_pfn = start_pfn + nr_pages;
|
|
|
|
|
|
|
|
return alloc_contig_range(start_pfn, end_pfn, MIGRATE_MOVABLE,
|
|
|
|
gfp_mask);
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool pfn_range_valid_contig(struct zone *z, unsigned long start_pfn,
|
|
|
|
unsigned long nr_pages)
|
|
|
|
{
|
|
|
|
unsigned long i, end_pfn = start_pfn + nr_pages;
|
|
|
|
struct page *page;
|
|
|
|
|
|
|
|
for (i = start_pfn; i < end_pfn; i++) {
|
|
|
|
page = pfn_to_online_page(i);
|
|
|
|
if (!page)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (page_zone(page) != z)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (PageReserved(page))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (page_count(page) > 0)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (PageHuge(page))
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool zone_spans_last_pfn(const struct zone *zone,
|
|
|
|
unsigned long start_pfn, unsigned long nr_pages)
|
|
|
|
{
|
|
|
|
unsigned long last_pfn = start_pfn + nr_pages - 1;
|
|
|
|
|
|
|
|
return zone_spans_pfn(zone, last_pfn);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* alloc_contig_pages() -- tries to find and allocate contiguous range of pages
|
|
|
|
* @nr_pages: Number of contiguous pages to allocate
|
|
|
|
* @gfp_mask: GFP mask to limit search and used during compaction
|
|
|
|
* @nid: Target node
|
|
|
|
* @nodemask: Mask for other possible nodes
|
|
|
|
*
|
|
|
|
* This routine is a wrapper around alloc_contig_range(). It scans over zones
|
|
|
|
* on an applicable zonelist to find a contiguous pfn range which can then be
|
|
|
|
* tried for allocation with alloc_contig_range(). This routine is intended
|
|
|
|
* for allocation requests which can not be fulfilled with the buddy allocator.
|
|
|
|
*
|
|
|
|
* The allocated memory is always aligned to a page boundary. If nr_pages is a
|
|
|
|
* power of two then the alignment is guaranteed to be to the given nr_pages
|
|
|
|
* (e.g. 1GB request would be aligned to 1GB).
|
|
|
|
*
|
|
|
|
* Allocated pages can be freed with free_contig_range() or by manually calling
|
|
|
|
* __free_page() on each allocated page.
|
|
|
|
*
|
|
|
|
* Return: pointer to contiguous pages on success, or NULL if not successful.
|
|
|
|
*/
|
|
|
|
struct page *alloc_contig_pages(unsigned long nr_pages, gfp_t gfp_mask,
|
|
|
|
int nid, nodemask_t *nodemask)
|
|
|
|
{
|
|
|
|
unsigned long ret, pfn, flags;
|
|
|
|
struct zonelist *zonelist;
|
|
|
|
struct zone *zone;
|
|
|
|
struct zoneref *z;
|
|
|
|
|
|
|
|
zonelist = node_zonelist(nid, gfp_mask);
|
|
|
|
for_each_zone_zonelist_nodemask(zone, z, zonelist,
|
|
|
|
gfp_zone(gfp_mask), nodemask) {
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
|
|
|
|
|
|
|
pfn = ALIGN(zone->zone_start_pfn, nr_pages);
|
|
|
|
while (zone_spans_last_pfn(zone, pfn, nr_pages)) {
|
|
|
|
if (pfn_range_valid_contig(zone, pfn, nr_pages)) {
|
|
|
|
/*
|
|
|
|
* We release the zone lock here because
|
|
|
|
* alloc_contig_range() will also lock the zone
|
|
|
|
* at some point. If there's an allocation
|
|
|
|
* spinning on this lock, it may win the race
|
|
|
|
* and cause alloc_contig_range() to fail...
|
|
|
|
*/
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
|
|
|
ret = __alloc_contig_pages(pfn, nr_pages,
|
|
|
|
gfp_mask);
|
|
|
|
if (!ret)
|
|
|
|
return pfn_to_page(pfn);
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
|
|
|
}
|
|
|
|
pfn += nr_pages;
|
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
2019-05-14 00:19:04 +00:00
|
|
|
#endif /* CONFIG_CONTIG_ALLOC */
|
2011-12-29 12:09:50 +00:00
|
|
|
|
2019-05-14 00:19:04 +00:00
|
|
|
void free_contig_range(unsigned long pfn, unsigned int nr_pages)
|
2011-12-29 12:09:50 +00:00
|
|
|
{
|
2012-12-20 23:05:18 +00:00
|
|
|
unsigned int count = 0;
|
|
|
|
|
|
|
|
for (; nr_pages--; pfn++) {
|
|
|
|
struct page *page = pfn_to_page(pfn);
|
|
|
|
|
|
|
|
count += page_count(page) != 1;
|
|
|
|
__free_page(page);
|
|
|
|
}
|
|
|
|
WARN(count != 0, "%d pages are still in use!\n", count);
|
2011-12-29 12:09:50 +00:00
|
|
|
}
|
virtio-mem: Paravirtualized memory hotunplug part 2
We also want to unplug online memory (contained in online memory blocks
and, therefore, managed by the buddy), and eventually replug it later.
When requested to unplug memory, we use alloc_contig_range() to allocate
subblocks in online memory blocks (so we are the owner) and send them to
our hypervisor. When requested to plug memory, we can replug such memory
using free_contig_range() after asking our hypervisor.
We also want to mark all allocated pages PG_offline, so nobody will
touch them. To differentiate pages that were never onlined when
onlining the memory block from pages allocated via alloc_contig_range(), we
use PageDirty(). Based on this flag, virtio_mem_fake_online() can either
online the pages for the first time or use free_contig_range().
It is worth noting that there are no guarantees on how much memory can
actually get unplugged again. All device memory might completely be
fragmented with unmovable data, such that no subblock can get unplugged.
We are not touching the ZONE_MOVABLE. If memory is onlined to the
ZONE_MOVABLE, it can only get unplugged after that memory was offlined
manually by user space. In normal operation, virtio-mem memory is
suggested to be onlined to ZONE_NORMAL. In the future, we will try to
make unplug more likely to succeed.
Add a module parameter to control if online memory shall be touched.
As we want to access alloc_contig_range()/free_contig_range() from
kernel module context, export the symbols.
Note: Whenever virtio-mem uses alloc_contig_range(), all affected pages
are on the same node, in the same zone, and contain no holes.
Acked-by: Michal Hocko <mhocko@suse.com> # to export contig range allocator API
Tested-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/20200507140139.17083-6-david@redhat.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
2020-05-07 14:01:29 +00:00
|
|
|
EXPORT_SYMBOL(free_contig_range);
|
2011-12-29 12:09:50 +00:00
|
|
|
|
2013-07-03 22:01:33 +00:00
|
|
|
/*
|
|
|
|
* The zone indicated has a new number of managed_pages; batch sizes and percpu
|
|
|
|
* page high values need to be recalulated.
|
|
|
|
*/
|
2012-07-31 23:43:35 +00:00
|
|
|
void __meminit zone_pcp_update(struct zone *zone)
|
|
|
|
{
|
2013-07-03 22:01:29 +00:00
|
|
|
mutex_lock(&pcp_batch_high_lock);
|
2020-12-15 03:10:43 +00:00
|
|
|
zone_set_pageset_high_and_batch(zone);
|
2013-07-03 22:01:29 +00:00
|
|
|
mutex_unlock(&pcp_batch_high_lock);
|
2012-07-31 23:43:35 +00:00
|
|
|
}
|
|
|
|
|
mm, page_alloc: disable pcplists during memory offline
Memory offlining relies on page isolation to guarantee a forward progress
because pages cannot be reused while they are isolated. But the page
isolation itself doesn't prevent from races while freed pages are stored
on pcp lists and thus can be reused. This can be worked around by
repeated draining of pcplists, as done by commit 968318261221
("mm/memory_hotplug: drain per-cpu pages again during memory offline").
David and Michal would prefer that this race was closed in a way that
callers of page isolation who need stronger guarantees don't need to
repeatedly drain. David suggested disabling pcplists usage completely
during page isolation, instead of repeatedly draining them.
To achieve this without adding special cases in alloc/free fastpath, we
can use the same approach as boot pagesets - when pcp->high is 0, any
pcplist addition will be immediately flushed.
The race can thus be closed by setting pcp->high to 0 and draining
pcplists once, before calling start_isolate_page_range(). The draining
will serialize after processes that already disabled interrupts and read
the old value of pcp->high in free_unref_page_commit(), and processes that
have not yet disabled interrupts, will observe pcp->high == 0 when they
are rescheduled, and skip pcplists. This guarantees no stray pages on
pcplists in zones where isolation happens.
This patch thus adds zone_pcp_disable() and zone_pcp_enable() functions
that page isolation users can call before start_isolate_page_range() and
after unisolating (or offlining) the isolated pages.
Also, drain_all_pages() is optimized to only execute on cpus where
pcplists are not empty. The check can however race with a free to pcplist
that has not yet increased the pcp->count from 0 to 1. Thus make the
drain optionally skip the racy check and drain on all cpus, and use this
option in zone_pcp_disable().
As we have to avoid external updates to high and batch while pcplists are
disabled, we take pcp_batch_high_lock in zone_pcp_disable() and release it
in zone_pcp_enable(). This also synchronizes multiple users of
zone_pcp_disable()/enable().
Currently the only user of this functionality is offline_pages().
[vbabka@suse.cz: add comment, per David]
Link: https://lkml.kernel.org/r/527480ef-ed72-e1c1-52a0-1c5b0113df45@suse.cz
Link: https://lkml.kernel.org/r/20201111092812.11329-8-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Suggested-by: David Hildenbrand <david@redhat.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-12-15 03:10:59 +00:00
|
|
|
/*
|
|
|
|
* Effectively disable pcplists for the zone by setting the high limit to 0
|
|
|
|
* and draining all cpus. A concurrent page freeing on another CPU that's about
|
|
|
|
* to put the page on pcplist will either finish before the drain and the page
|
|
|
|
* will be drained, or observe the new high limit and skip the pcplist.
|
|
|
|
*
|
|
|
|
* Must be paired with a call to zone_pcp_enable().
|
|
|
|
*/
|
|
|
|
void zone_pcp_disable(struct zone *zone)
|
|
|
|
{
|
|
|
|
mutex_lock(&pcp_batch_high_lock);
|
|
|
|
__zone_set_pageset_high_and_batch(zone, 0, 1);
|
|
|
|
__drain_all_pages(zone, true);
|
|
|
|
}
|
|
|
|
|
|
|
|
void zone_pcp_enable(struct zone *zone)
|
|
|
|
{
|
|
|
|
__zone_set_pageset_high_and_batch(zone, zone->pageset_high, zone->pageset_batch);
|
|
|
|
mutex_unlock(&pcp_batch_high_lock);
|
|
|
|
}
|
|
|
|
|
2012-07-31 23:43:32 +00:00
|
|
|
void zone_pcp_reset(struct zone *zone)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
2012-10-08 23:33:39 +00:00
|
|
|
int cpu;
|
|
|
|
struct per_cpu_pageset *pset;
|
2012-07-31 23:43:32 +00:00
|
|
|
|
|
|
|
/* avoid races with drain_pages() */
|
|
|
|
local_irq_save(flags);
|
|
|
|
if (zone->pageset != &boot_pageset) {
|
2012-10-08 23:33:39 +00:00
|
|
|
for_each_online_cpu(cpu) {
|
|
|
|
pset = per_cpu_ptr(zone->pageset, cpu);
|
|
|
|
drain_zonestat(zone, pset);
|
|
|
|
}
|
2012-07-31 23:43:32 +00:00
|
|
|
free_percpu(zone->pageset);
|
|
|
|
zone->pageset = &boot_pageset;
|
|
|
|
}
|
|
|
|
local_irq_restore(flags);
|
|
|
|
}
|
|
|
|
|
2012-12-12 00:01:01 +00:00
|
|
|
#ifdef CONFIG_MEMORY_HOTREMOVE
|
2007-10-16 08:26:12 +00:00
|
|
|
/*
|
2020-10-16 03:07:59 +00:00
|
|
|
* All pages in the range must be in a single zone, must not contain holes,
|
|
|
|
* must span full sections, and must be isolated before calling this function.
|
2007-10-16 08:26:12 +00:00
|
|
|
*/
|
2020-10-16 03:07:59 +00:00
|
|
|
void __offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn)
|
2007-10-16 08:26:12 +00:00
|
|
|
{
|
2020-10-16 03:07:59 +00:00
|
|
|
unsigned long pfn = start_pfn;
|
2007-10-16 08:26:12 +00:00
|
|
|
struct page *page;
|
|
|
|
struct zone *zone;
|
2019-12-01 01:54:03 +00:00
|
|
|
unsigned int order;
|
2007-10-16 08:26:12 +00:00
|
|
|
unsigned long flags;
|
2019-05-14 00:21:24 +00:00
|
|
|
|
2017-07-06 22:37:56 +00:00
|
|
|
offline_mem_sections(pfn, end_pfn);
|
2007-10-16 08:26:12 +00:00
|
|
|
zone = page_zone(pfn_to_page(pfn));
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
|
|
|
while (pfn < end_pfn) {
|
|
|
|
page = pfn_to_page(pfn);
|
2012-12-12 00:00:45 +00:00
|
|
|
/*
|
|
|
|
* The HWPoisoned page may be not in buddy system, and
|
|
|
|
* page_count() is not 0.
|
|
|
|
*/
|
|
|
|
if (unlikely(!PageBuddy(page) && PageHWPoison(page))) {
|
|
|
|
pfn++;
|
|
|
|
continue;
|
|
|
|
}
|
mm: Allow to offline unmovable PageOffline() pages via MEM_GOING_OFFLINE
virtio-mem wants to allow to offline memory blocks of which some parts
were unplugged (allocated via alloc_contig_range()), especially, to later
offline and remove completely unplugged memory blocks. The important part
is that PageOffline() has to remain set until the section is offline, so
these pages will never get accessed (e.g., when dumping). The pages should
not be handed back to the buddy (which would require clearing PageOffline()
and result in issues if offlining fails and the pages are suddenly in the
buddy).
Let's allow to do that by allowing to isolate any PageOffline() page
when offlining. This way, we can reach the memory hotplug notifier
MEM_GOING_OFFLINE, where the driver can signal that he is fine with
offlining this page by dropping its reference count. PageOffline() pages
with a reference count of 0 can then be skipped when offlining the
pages (like if they were free, however they are not in the buddy).
Anybody who uses PageOffline() pages and does not agree to offline them
(e.g., Hyper-V balloon, XEN balloon, VMWare balloon for 2MB pages) will not
decrement the reference count and make offlining fail when trying to
migrate such an unmovable page. So there should be no observable change.
Same applies to balloon compaction users (movable PageOffline() pages), the
pages will simply be migrated.
Note 1: If offlining fails, a driver has to increment the reference
count again in MEM_CANCEL_OFFLINE.
Note 2: A driver that makes use of this has to be aware that re-onlining
the memory block has to be handled by hooking into onlining code
(online_page_callback_t), resetting the page PageOffline() and
not giving them to the buddy.
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Tested-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Juergen Gross <jgross@suse.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Anthony Yznaga <anthony.yznaga@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Pingfan Liu <kernelfans@gmail.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/20200507140139.17083-7-david@redhat.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
2020-05-07 14:01:30 +00:00
|
|
|
/*
|
|
|
|
* At this point all remaining PageOffline() pages have a
|
|
|
|
* reference count of 0 and can simply be skipped.
|
|
|
|
*/
|
|
|
|
if (PageOffline(page)) {
|
|
|
|
BUG_ON(page_count(page));
|
|
|
|
BUG_ON(PageBuddy(page));
|
|
|
|
pfn++;
|
|
|
|
continue;
|
|
|
|
}
|
2012-12-12 00:00:45 +00:00
|
|
|
|
2007-10-16 08:26:12 +00:00
|
|
|
BUG_ON(page_count(page));
|
|
|
|
BUG_ON(!PageBuddy(page));
|
2020-10-16 03:10:15 +00:00
|
|
|
order = buddy_order(page);
|
2020-04-07 03:04:49 +00:00
|
|
|
del_page_from_free_list(page, zone, order);
|
2007-10-16 08:26:12 +00:00
|
|
|
pfn += (1 << order);
|
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
|
|
|
}
|
|
|
|
#endif
|
2009-12-16 11:19:58 +00:00
|
|
|
|
|
|
|
bool is_free_buddy_page(struct page *page)
|
|
|
|
{
|
|
|
|
struct zone *zone = page_zone(page);
|
|
|
|
unsigned long pfn = page_to_pfn(page);
|
|
|
|
unsigned long flags;
|
2014-06-04 23:10:21 +00:00
|
|
|
unsigned int order;
|
2009-12-16 11:19:58 +00:00
|
|
|
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
|
|
|
for (order = 0; order < MAX_ORDER; order++) {
|
|
|
|
struct page *page_head = page - (pfn & ((1 << order) - 1));
|
|
|
|
|
2020-10-16 03:10:15 +00:00
|
|
|
if (PageBuddy(page_head) && buddy_order(page_head) >= order)
|
2009-12-16 11:19:58 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
|
|
|
|
|
|
|
return order < MAX_ORDER;
|
|
|
|
}
|
2018-08-24 00:00:42 +00:00
|
|
|
|
|
|
|
#ifdef CONFIG_MEMORY_FAILURE
|
|
|
|
/*
|
2020-10-16 03:07:05 +00:00
|
|
|
* Break down a higher-order page in sub-pages, and keep our target out of
|
|
|
|
* buddy allocator.
|
2018-08-24 00:00:42 +00:00
|
|
|
*/
|
2020-10-16 03:07:05 +00:00
|
|
|
static void break_down_buddy_pages(struct zone *zone, struct page *page,
|
|
|
|
struct page *target, int low, int high,
|
|
|
|
int migratetype)
|
|
|
|
{
|
|
|
|
unsigned long size = 1 << high;
|
|
|
|
struct page *current_buddy, *next_page;
|
|
|
|
|
|
|
|
while (high > low) {
|
|
|
|
high--;
|
|
|
|
size >>= 1;
|
|
|
|
|
|
|
|
if (target >= &page[size]) {
|
|
|
|
next_page = page + size;
|
|
|
|
current_buddy = page;
|
|
|
|
} else {
|
|
|
|
next_page = page;
|
|
|
|
current_buddy = page + size;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (set_page_guard(zone, current_buddy, high, migratetype))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (current_buddy != target) {
|
|
|
|
add_to_free_list(current_buddy, zone, high, migratetype);
|
2020-10-16 03:10:15 +00:00
|
|
|
set_buddy_order(current_buddy, high);
|
2020-10-16 03:07:05 +00:00
|
|
|
page = next_page;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Take a page that will be marked as poisoned off the buddy allocator.
|
|
|
|
*/
|
|
|
|
bool take_page_off_buddy(struct page *page)
|
2018-08-24 00:00:42 +00:00
|
|
|
{
|
|
|
|
struct zone *zone = page_zone(page);
|
|
|
|
unsigned long pfn = page_to_pfn(page);
|
|
|
|
unsigned long flags;
|
|
|
|
unsigned int order;
|
2020-10-16 03:07:05 +00:00
|
|
|
bool ret = false;
|
2018-08-24 00:00:42 +00:00
|
|
|
|
|
|
|
spin_lock_irqsave(&zone->lock, flags);
|
|
|
|
for (order = 0; order < MAX_ORDER; order++) {
|
|
|
|
struct page *page_head = page - (pfn & ((1 << order) - 1));
|
2020-10-16 03:10:15 +00:00
|
|
|
int page_order = buddy_order(page_head);
|
2018-08-24 00:00:42 +00:00
|
|
|
|
2020-10-16 03:10:15 +00:00
|
|
|
if (PageBuddy(page_head) && page_order >= order) {
|
2020-10-16 03:07:05 +00:00
|
|
|
unsigned long pfn_head = page_to_pfn(page_head);
|
|
|
|
int migratetype = get_pfnblock_migratetype(page_head,
|
|
|
|
pfn_head);
|
|
|
|
|
2020-10-16 03:10:15 +00:00
|
|
|
del_page_from_free_list(page_head, zone, page_order);
|
2020-10-16 03:07:05 +00:00
|
|
|
break_down_buddy_pages(zone, page_head, page, 0,
|
2020-10-16 03:10:15 +00:00
|
|
|
page_order, migratetype);
|
2020-10-16 03:07:05 +00:00
|
|
|
ret = true;
|
2018-08-24 00:00:42 +00:00
|
|
|
break;
|
|
|
|
}
|
2020-10-16 03:07:05 +00:00
|
|
|
if (page_count(page_head) > 0)
|
|
|
|
break;
|
2018-08-24 00:00:42 +00:00
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&zone->lock, flags);
|
2020-10-16 03:07:05 +00:00
|
|
|
return ret;
|
2018-08-24 00:00:42 +00:00
|
|
|
}
|
|
|
|
#endif
|