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/memory.c
|
|
|
|
*
|
|
|
|
* Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* demand-loading started 01.12.91 - seems it is high on the list of
|
|
|
|
* things wanted, and it should be easy to implement. - Linus
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Ok, demand-loading was easy, shared pages a little bit tricker. Shared
|
|
|
|
* pages started 02.12.91, seems to work. - Linus.
|
|
|
|
*
|
|
|
|
* Tested sharing by executing about 30 /bin/sh: under the old kernel it
|
|
|
|
* would have taken more than the 6M I have free, but it worked well as
|
|
|
|
* far as I could see.
|
|
|
|
*
|
|
|
|
* Also corrected some "invalidate()"s - I wasn't doing enough of them.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Real VM (paging to/from disk) started 18.12.91. Much more work and
|
|
|
|
* thought has to go into this. Oh, well..
|
|
|
|
* 19.12.91 - works, somewhat. Sometimes I get faults, don't know why.
|
|
|
|
* Found it. Everything seems to work now.
|
|
|
|
* 20.12.91 - Ok, making the swap-device changeable like the root.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* 05.04.94 - Multi-page memory management added for v1.1.
|
2017-02-24 22:59:01 +00:00
|
|
|
* Idea by Alex Bligh (alex@cconcepts.co.uk)
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
|
|
|
* 16.07.99 - Support of BIGMEM added by Gerhard Wichert, Siemens AG
|
|
|
|
* (Gerhard.Wichert@pdb.siemens.de)
|
|
|
|
*
|
|
|
|
* Aug/Sep 2004 Changed to four level page tables (Andi Kleen)
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/kernel_stat.h>
|
|
|
|
#include <linux/mm.h>
|
2022-01-14 22:06:10 +00:00
|
|
|
#include <linux/mm_inline.h>
|
2017-02-08 17:51:29 +00:00
|
|
|
#include <linux/sched/mm.h>
|
2017-02-08 17:51:30 +00:00
|
|
|
#include <linux/sched/coredump.h>
|
2017-02-08 17:51:31 +00:00
|
|
|
#include <linux/sched/numa_balancing.h>
|
2017-02-08 17:51:36 +00:00
|
|
|
#include <linux/sched/task.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/hugetlb.h>
|
|
|
|
#include <linux/mman.h>
|
|
|
|
#include <linux/swap.h>
|
|
|
|
#include <linux/highmem.h>
|
|
|
|
#include <linux/pagemap.h>
|
2017-09-08 23:11:43 +00:00
|
|
|
#include <linux/memremap.h>
|
2022-09-15 15:03:48 +00:00
|
|
|
#include <linux/kmsan.h>
|
2009-09-22 00:02:01 +00:00
|
|
|
#include <linux/ksm.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/rmap.h>
|
2011-10-16 06:01:52 +00:00
|
|
|
#include <linux/export.h>
|
2006-07-14 07:24:37 +00:00
|
|
|
#include <linux/delayacct.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/init.h>
|
2016-01-16 00:56:40 +00:00
|
|
|
#include <linux/pfn_t.h>
|
2006-09-26 06:30:58 +00:00
|
|
|
#include <linux/writeback.h>
|
2008-02-07 08:13:53 +00:00
|
|
|
#include <linux/memcontrol.h>
|
mmu-notifiers: core
With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages.
There are secondary MMUs (with secondary sptes and secondary tlbs) too.
sptes in the kvm case are shadow pagetables, but when I say spte in
mmu-notifier context, I mean "secondary pte". In GRU case there's no
actual secondary pte and there's only a secondary tlb because the GRU
secondary MMU has no knowledge about sptes and every secondary tlb miss
event in the MMU always generates a page fault that has to be resolved by
the CPU (this is not the case of KVM where the a secondary tlb miss will
walk sptes in hardware and it will refill the secondary tlb transparently
to software if the corresponding spte is present). The same way
zap_page_range has to invalidate the pte before freeing the page, the spte
(and secondary tlb) must also be invalidated before any page is freed and
reused.
Currently we take a page_count pin on every page mapped by sptes, but that
means the pages can't be swapped whenever they're mapped by any spte
because they're part of the guest working set. Furthermore a spte unmap
event can immediately lead to a page to be freed when the pin is released
(so requiring the same complex and relatively slow tlb_gather smp safe
logic we have in zap_page_range and that can be avoided completely if the
spte unmap event doesn't require an unpin of the page previously mapped in
the secondary MMU).
The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know
when the VM is swapping or freeing or doing anything on the primary MMU so
that the secondary MMU code can drop sptes before the pages are freed,
avoiding all page pinning and allowing 100% reliable swapping of guest
physical address space. Furthermore it avoids the code that teardown the
mappings of the secondary MMU, to implement a logic like tlb_gather in
zap_page_range that would require many IPI to flush other cpu tlbs, for
each fixed number of spte unmapped.
To make an example: if what happens on the primary MMU is a protection
downgrade (from writeable to wrprotect) the secondary MMU mappings will be
invalidated, and the next secondary-mmu-page-fault will call
get_user_pages and trigger a do_wp_page through get_user_pages if it
called get_user_pages with write=1, and it'll re-establishing an updated
spte or secondary-tlb-mapping on the copied page. Or it will setup a
readonly spte or readonly tlb mapping if it's a guest-read, if it calls
get_user_pages with write=0. This is just an example.
This allows to map any page pointed by any pte (and in turn visible in the
primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an
full MMU with both sptes and secondary-tlb like the shadow-pagetable layer
with kvm), or a remote DMA in software like XPMEM (hence needing of
schedule in XPMEM code to send the invalidate to the remote node, while no
need to schedule in kvm/gru as it's an immediate event like invalidating
primary-mmu pte).
At least for KVM without this patch it's impossible to swap guests
reliably. And having this feature and removing the page pin allows
several other optimizations that simplify life considerably.
Dependencies:
1) mm_take_all_locks() to register the mmu notifier when the whole VM
isn't doing anything with "mm". This allows mmu notifier users to keep
track if the VM is in the middle of the invalidate_range_begin/end
critical section with an atomic counter incraese in range_begin and
decreased in range_end. No secondary MMU page fault is allowed to map
any spte or secondary tlb reference, while the VM is in the middle of
range_begin/end as any page returned by get_user_pages in that critical
section could later immediately be freed without any further
->invalidate_page notification (invalidate_range_begin/end works on
ranges and ->invalidate_page isn't called immediately before freeing
the page). To stop all page freeing and pagetable overwrites the
mmap_sem must be taken in write mode and all other anon_vma/i_mmap
locks must be taken too.
2) It'd be a waste to add branches in the VM if nobody could possibly
run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if
CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of
mmu notifiers, but this already allows to compile a KVM external module
against a kernel with mmu notifiers enabled and from the next pull from
kvm.git we'll start using them. And GRU/XPMEM will also be able to
continue the development by enabling KVM=m in their config, until they
submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can
also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n).
This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM
are all =n.
The mmu_notifier_register call can fail because mm_take_all_locks may be
interrupted by a signal and return -EINTR. Because mmu_notifier_reigster
is used when a driver startup, a failure can be gracefully handled. Here
an example of the change applied to kvm to register the mmu notifiers.
Usually when a driver startups other allocations are required anyway and
-ENOMEM failure paths exists already.
struct kvm *kvm_arch_create_vm(void)
{
struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ int err;
if (!kvm)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
+ kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops;
+ err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm);
+ if (err) {
+ kfree(kvm);
+ return ERR_PTR(err);
+ }
+
return kvm;
}
mmu_notifier_unregister returns void and it's reliable.
The patch also adds a few needed but missing includes that would prevent
kernel to compile after these changes on non-x86 archs (x86 didn't need
them by luck).
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix mm/filemap_xip.c build]
[akpm@linux-foundation.org: fix mm/mmu_notifier.c build]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Jack Steiner <steiner@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Kanoj Sarcar <kanojsarcar@yahoo.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Avi Kivity <avi@qumranet.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Chris Wright <chrisw@redhat.com>
Cc: Marcelo Tosatti <marcelo@kvack.org>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: Izik Eidus <izike@qumranet.com>
Cc: Anthony Liguori <aliguori@us.ibm.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>
2008-07-28 22:46:29 +00:00
|
|
|
#include <linux/mmu_notifier.h>
|
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
|
|
|
#include <linux/swapops.h>
|
|
|
|
#include <linux/elf.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
|
|
|
#include <linux/gfp.h>
|
2012-11-02 11:33:45 +00:00
|
|
|
#include <linux/migrate.h>
|
2012-12-18 00:01:23 +00:00
|
|
|
#include <linux/string.h>
|
2022-08-18 13:10:41 +00:00
|
|
|
#include <linux/memory-tiers.h>
|
2014-04-07 22:37:22 +00:00
|
|
|
#include <linux/debugfs.h>
|
2015-09-04 22:46:20 +00:00
|
|
|
#include <linux/userfaultfd_k.h>
|
2016-05-12 16:29:19 +00:00
|
|
|
#include <linux/dax.h>
|
2017-08-18 22:16:15 +00:00
|
|
|
#include <linux/oom.h>
|
2019-03-05 23:42:58 +00:00
|
|
|
#include <linux/numa.h>
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
#include <linux/perf_event.h>
|
|
|
|
#include <linux/ptrace.h>
|
2020-09-04 23:35:43 +00:00
|
|
|
#include <linux/vmalloc.h>
|
memory tiering: hot page selection with hint page fault latency
Patch series "memory tiering: hot page selection", v4.
To optimize page placement in a memory tiering system with NUMA balancing,
the hot pages in the slow memory nodes need to be identified.
Essentially, the original NUMA balancing implementation selects the mostly
recently accessed (MRU) pages to promote. But this isn't a perfect
algorithm to identify the hot pages. Because the pages with quite low
access frequency may be accessed eventually given the NUMA balancing page
table scanning period could be quite long (e.g. 60 seconds). So in this
patchset, we implement a new hot page identification algorithm based on
the latency between NUMA balancing page table scanning and hint page
fault. Which is a kind of mostly frequently accessed (MFU) algorithm.
In NUMA balancing memory tiering mode, if there are hot pages in slow
memory node and cold pages in fast memory node, we need to promote/demote
hot/cold pages between the fast and cold memory nodes.
A choice is to promote/demote as fast as possible. But the CPU cycles and
memory bandwidth consumed by the high promoting/demoting throughput will
hurt the latency of some workload because of accessing inflating and slow
memory bandwidth contention.
A way to resolve this issue is to restrict the max promoting/demoting
throughput. It will take longer to finish the promoting/demoting. But
the workload latency will be better. This is implemented in this patchset
as the page promotion rate limit mechanism.
The promotion hot threshold is workload and system configuration
dependent. So in this patchset, a method to adjust the hot threshold
automatically is implemented. The basic idea is to control the number of
the candidate promotion pages to match the promotion rate limit.
We used the pmbench memory accessing benchmark tested the patchset on a
2-socket server system with DRAM and PMEM installed. The test results are
as follows,
pmbench score promote rate
(accesses/s) MB/s
------------- ------------
base 146887704.1 725.6
hot selection 165695601.2 544.0
rate limit 162814569.8 165.2
auto adjustment 170495294.0 136.9
From the results above,
With hot page selection patch [1/3], the pmbench score increases about
12.8%, and promote rate (overhead) decreases about 25.0%, compared with
base kernel.
With rate limit patch [2/3], pmbench score decreases about 1.7%, and
promote rate decreases about 69.6%, compared with hot page selection
patch.
With threshold auto adjustment patch [3/3], pmbench score increases about
4.7%, and promote rate decrease about 17.1%, compared with rate limit
patch.
Baolin helped to test the patchset with MySQL on a machine which contains
1 DRAM node (30G) and 1 PMEM node (126G).
sysbench /usr/share/sysbench/oltp_read_write.lua \
......
--tables=200 \
--table-size=1000000 \
--report-interval=10 \
--threads=16 \
--time=120
The tps can be improved about 5%.
This patch (of 3):
To optimize page placement in a memory tiering system with NUMA balancing,
the hot pages in the slow memory node need to be identified. Essentially,
the original NUMA balancing implementation selects the mostly recently
accessed (MRU) pages to promote. But this isn't a perfect algorithm to
identify the hot pages. Because the pages with quite low access frequency
may be accessed eventually given the NUMA balancing page table scanning
period could be quite long (e.g. 60 seconds). The most frequently
accessed (MFU) algorithm is better.
So, in this patch we implemented a better hot page selection algorithm.
Which is based on NUMA balancing page table scanning and hint page fault
as follows,
- When the page tables of the processes are scanned to change PTE/PMD
to be PROT_NONE, the current time is recorded in struct page as scan
time.
- When the page is accessed, hint page fault will occur. The scan
time is gotten from the struct page. And The hint page fault
latency is defined as
hint page fault time - scan time
The shorter the hint page fault latency of a page is, the higher the
probability of their access frequency to be higher. So the hint page
fault latency is a better estimation of the page hot/cold.
It's hard to find some extra space in struct page to hold the scan time.
Fortunately, we can reuse some bits used by the original NUMA balancing.
NUMA balancing uses some bits in struct page to store the page accessing
CPU and PID (referring to page_cpupid_xchg_last()). Which is used by the
multi-stage node selection algorithm to avoid to migrate pages shared
accessed by the NUMA nodes back and forth. But for pages in the slow
memory node, even if they are shared accessed by multiple NUMA nodes, as
long as the pages are hot, they need to be promoted to the fast memory
node. So the accessing CPU and PID information are unnecessary for the
slow memory pages. We can reuse these bits in struct page to record the
scan time. For the fast memory pages, these bits are used as before.
For the hot threshold, the default value is 1 second, which works well in
our performance test. All pages with hint page fault latency < hot
threshold will be considered hot.
It's hard for users to determine the hot threshold. So we don't provide a
kernel ABI to set it, just provide a debugfs interface for advanced users
to experiment. We will continue to work on a hot threshold automatic
adjustment mechanism.
The downside of the above method is that the response time to the workload
hot spot changing may be much longer. For example,
- A previous cold memory area becomes hot
- The hint page fault will be triggered. But the hint page fault
latency isn't shorter than the hot threshold. So the pages will
not be promoted.
- When the memory area is scanned again, maybe after a scan period,
the hint page fault latency measured will be shorter than the hot
threshold and the pages will be promoted.
To mitigate this, if there are enough free space in the fast memory node,
the hot threshold will not be used, all pages will be promoted upon the
hint page fault for fast response.
Thanks Zhong Jiang reported and tested the fix for a bug when disabling
memory tiering mode dynamically.
Link: https://lkml.kernel.org/r/20220713083954.34196-1-ying.huang@intel.com
Link: https://lkml.kernel.org/r/20220713083954.34196-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Tested-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Wei Xu <weixugc@google.com>
Cc: osalvador <osalvador@suse.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Zhong Jiang <zhongjiang-ali@linux.alibaba.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-07-13 08:39:51 +00:00
|
|
|
#include <linux/sched/sysctl.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
|
mm: emit tracepoint when RSS changes
Useful to track how RSS is changing per TGID to detect spikes in RSS and
memory hogs. Several Android teams have been using this patch in
various kernel trees for half a year now. Many reported to me it is
really useful so I'm posting it upstream.
Initial patch developed by Tim Murray. Changes I made from original
patch: o Prevent any additional space consumed by mm_struct.
Regarding the fact that the RSS may change too often thus flooding the
traces - note that, there is some "hysterisis" with this already. That
is - We update the counter only if we receive 64 page faults due to
SPLIT_RSS_ACCOUNTING. However, during zapping or copying of pte range,
the RSS is updated immediately which can become noisy/flooding. In a
previous discussion, we agreed that BPF or ftrace can be used to rate
limit the signal if this becomes an issue.
Also note that I added wrappers to trace_rss_stat to prevent compiler
errors where linux/mm.h is included from tracing code, causing errors
such as:
CC kernel/trace/power-traces.o
In file included from ./include/trace/define_trace.h:102,
from ./include/trace/events/kmem.h:342,
from ./include/linux/mm.h:31,
from ./include/linux/ring_buffer.h:5,
from ./include/linux/trace_events.h:6,
from ./include/trace/events/power.h:12,
from kernel/trace/power-traces.c:15:
./include/trace/trace_events.h:113:22: error: field `ent' has incomplete type
struct trace_entry ent; \
Link: http://lore.kernel.org/r/20190903200905.198642-1-joel@joelfernandes.org
Link: http://lkml.kernel.org/r/20191001172817.234886-1-joel@joelfernandes.org
Co-developed-by: Tim Murray <timmurray@google.com>
Signed-off-by: Tim Murray <timmurray@google.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Carmen Jackson <carmenjackson@google.com>
Cc: Mayank Gupta <mayankgupta@google.com>
Cc: Daniel Colascione <dancol@google.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-12-01 01:50:30 +00:00
|
|
|
#include <trace/events/kmem.h>
|
|
|
|
|
2009-09-18 19:55:55 +00:00
|
|
|
#include <asm/io.h>
|
mm/gup, x86/mm/pkeys: Check VMAs and PTEs for protection keys
Today, for normal faults and page table walks, we check the VMA
and/or PTE to ensure that it is compatible with the action. For
instance, if we get a write fault on a non-writeable VMA, we
SIGSEGV.
We try to do the same thing for protection keys. Basically, we
try to make sure that if a user does this:
mprotect(ptr, size, PROT_NONE);
*ptr = foo;
they see the same effects with protection keys when they do this:
mprotect(ptr, size, PROT_READ|PROT_WRITE);
set_pkey(ptr, size, 4);
wrpkru(0xffffff3f); // access disable pkey 4
*ptr = foo;
The state to do that checking is in the VMA, but we also
sometimes have to do it on the page tables only, like when doing
a get_user_pages_fast() where we have no VMA.
We add two functions and expose them to generic code:
arch_pte_access_permitted(pte_flags, write)
arch_vma_access_permitted(vma, write)
These are, of course, backed up in x86 arch code with checks
against the PTE or VMA's protection key.
But, there are also cases where we do not want to respect
protection keys. When we ptrace(), for instance, we do not want
to apply the tracer's PKRU permissions to the PTEs from the
process being traced.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexey Kardashevskiy <aik@ozlabs.ru>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boaz Harrosh <boaz@plexistor.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave@sr71.net>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: David Hildenbrand <dahi@linux.vnet.ibm.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Dominik Dingel <dingel@linux.vnet.ibm.com>
Cc: Dominik Vogt <vogt@linux.vnet.ibm.com>
Cc: Guan Xuetao <gxt@mprc.pku.edu.cn>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Jerome Marchand <jmarchan@redhat.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Matthew Wilcox <willy@linux.intel.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Sasha Levin <sasha.levin@oracle.com>
Cc: Shachar Raindel <raindel@mellanox.com>
Cc: Stephen Smalley <sds@tycho.nsa.gov>
Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: linux-arch@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: linux-mm@kvack.org
Cc: linux-s390@vger.kernel.org
Cc: linuxppc-dev@lists.ozlabs.org
Link: http://lkml.kernel.org/r/20160212210219.14D5D715@viggo.jf.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-12 21:02:19 +00:00
|
|
|
#include <asm/mmu_context.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <asm/pgalloc.h>
|
2016-12-24 19:46:01 +00:00
|
|
|
#include <linux/uaccess.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <asm/tlb.h>
|
|
|
|
#include <asm/tlbflush.h>
|
|
|
|
|
2020-09-04 23:35:43 +00:00
|
|
|
#include "pgalloc-track.h"
|
2008-07-24 04:27:10 +00:00
|
|
|
#include "internal.h"
|
2022-05-10 01:20:47 +00:00
|
|
|
#include "swap.h"
|
2008-07-24 04:27:10 +00:00
|
|
|
|
2018-02-16 15:25:53 +00:00
|
|
|
#if defined(LAST_CPUPID_NOT_IN_PAGE_FLAGS) && !defined(CONFIG_COMPILE_TEST)
|
2013-10-07 10:29:20 +00:00
|
|
|
#warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_cpupid.
|
2013-02-23 00:34:32 +00:00
|
|
|
#endif
|
|
|
|
|
2021-06-29 02:43:01 +00:00
|
|
|
#ifndef CONFIG_NUMA
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long max_mapnr;
|
|
|
|
EXPORT_SYMBOL(max_mapnr);
|
2017-02-24 22:59:01 +00:00
|
|
|
|
|
|
|
struct page *mem_map;
|
2005-04-16 22:20:36 +00:00
|
|
|
EXPORT_SYMBOL(mem_map);
|
|
|
|
#endif
|
|
|
|
|
2022-05-13 03:22:52 +00:00
|
|
|
static vm_fault_t do_fault(struct vm_fault *vmf);
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
static vm_fault_t do_anonymous_page(struct vm_fault *vmf);
|
|
|
|
static bool vmf_pte_changed(struct vm_fault *vmf);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return true if the original pte was a uffd-wp pte marker (so the pte was
|
|
|
|
* wr-protected).
|
|
|
|
*/
|
|
|
|
static bool vmf_orig_pte_uffd_wp(struct vm_fault *vmf)
|
|
|
|
{
|
|
|
|
if (!(vmf->flags & FAULT_FLAG_ORIG_PTE_VALID))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return pte_marker_uffd_wp(vmf->orig_pte);
|
|
|
|
}
|
2022-05-13 03:22:52 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* A number of key systems in x86 including ioremap() rely on the assumption
|
|
|
|
* that high_memory defines the upper bound on direct map memory, then end
|
|
|
|
* of ZONE_NORMAL. Under CONFIG_DISCONTIG this means that max_low_pfn and
|
|
|
|
* highstart_pfn must be the same; there must be no gap between ZONE_NORMAL
|
|
|
|
* and ZONE_HIGHMEM.
|
|
|
|
*/
|
2017-02-24 22:59:01 +00:00
|
|
|
void *high_memory;
|
2005-04-16 22:20:36 +00:00
|
|
|
EXPORT_SYMBOL(high_memory);
|
|
|
|
|
2008-02-06 21:39:44 +00:00
|
|
|
/*
|
|
|
|
* Randomize the address space (stacks, mmaps, brk, etc.).
|
|
|
|
*
|
|
|
|
* ( When CONFIG_COMPAT_BRK=y we exclude brk from randomization,
|
|
|
|
* as ancient (libc5 based) binaries can segfault. )
|
|
|
|
*/
|
|
|
|
int randomize_va_space __read_mostly =
|
|
|
|
#ifdef CONFIG_COMPAT_BRK
|
|
|
|
1;
|
|
|
|
#else
|
|
|
|
2;
|
|
|
|
#endif
|
2006-02-16 22:41:58 +00:00
|
|
|
|
2020-11-24 18:48:26 +00:00
|
|
|
#ifndef arch_wants_old_prefaulted_pte
|
|
|
|
static inline bool arch_wants_old_prefaulted_pte(void)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Transitioning a PTE from 'old' to 'young' can be expensive on
|
|
|
|
* some architectures, even if it's performed in hardware. By
|
|
|
|
* default, "false" means prefaulted entries will be 'young'.
|
|
|
|
*/
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2006-02-16 22:41:58 +00:00
|
|
|
static int __init disable_randmaps(char *s)
|
|
|
|
{
|
|
|
|
randomize_va_space = 0;
|
2006-03-31 10:30:33 +00:00
|
|
|
return 1;
|
2006-02-16 22:41:58 +00:00
|
|
|
}
|
|
|
|
__setup("norandmaps", disable_randmaps);
|
|
|
|
|
2009-09-22 00:03:34 +00:00
|
|
|
unsigned long zero_pfn __read_mostly;
|
2014-09-12 20:17:23 +00:00
|
|
|
EXPORT_SYMBOL(zero_pfn);
|
|
|
|
|
2017-02-24 22:59:01 +00:00
|
|
|
unsigned long highest_memmap_pfn __read_mostly;
|
|
|
|
|
2009-09-22 00:03:30 +00:00
|
|
|
/*
|
|
|
|
* CONFIG_MMU architectures set up ZERO_PAGE in their paging_init()
|
|
|
|
*/
|
|
|
|
static int __init init_zero_pfn(void)
|
|
|
|
{
|
|
|
|
zero_pfn = page_to_pfn(ZERO_PAGE(0));
|
|
|
|
return 0;
|
|
|
|
}
|
mm: fix race by making init_zero_pfn() early_initcall
There are code paths that rely on zero_pfn to be fully initialized
before core_initcall. For example, wq_sysfs_init() is a core_initcall
function that eventually results in a call to kernel_execve, which
causes a page fault with a subsequent mmput. If zero_pfn is not
initialized by then it may not get cleaned up properly and result in an
error:
BUG: Bad rss-counter state mm:(ptrval) type:MM_ANONPAGES val:1
Here is an analysis of the race as seen on a MIPS device. On this
particular MT7621 device (Ubiquiti ER-X), zero_pfn is PFN 0 until
initialized, at which point it becomes PFN 5120:
1. wq_sysfs_init calls into kobject_uevent_env at core_initcall:
kobject_uevent_env+0x7e4/0x7ec
kset_register+0x68/0x88
bus_register+0xdc/0x34c
subsys_virtual_register+0x34/0x78
wq_sysfs_init+0x1c/0x4c
do_one_initcall+0x50/0x1a8
kernel_init_freeable+0x230/0x2c8
kernel_init+0x10/0x100
ret_from_kernel_thread+0x14/0x1c
2. kobject_uevent_env() calls call_usermodehelper_exec() which executes
kernel_execve asynchronously.
3. Memory allocations in kernel_execve cause a page fault, bumping the
MM reference counter:
add_mm_counter_fast+0xb4/0xc0
handle_mm_fault+0x6e4/0xea0
__get_user_pages.part.78+0x190/0x37c
__get_user_pages_remote+0x128/0x360
get_arg_page+0x34/0xa0
copy_string_kernel+0x194/0x2a4
kernel_execve+0x11c/0x298
call_usermodehelper_exec_async+0x114/0x194
4. In case zero_pfn has not been initialized yet, zap_pte_range does
not decrement the MM_ANONPAGES RSS counter and the BUG message is
triggered shortly afterwards when __mmdrop checks the ref counters:
__mmdrop+0x98/0x1d0
free_bprm+0x44/0x118
kernel_execve+0x160/0x1d8
call_usermodehelper_exec_async+0x114/0x194
ret_from_kernel_thread+0x14/0x1c
To avoid races such as described above, initialize init_zero_pfn at
early_initcall level. Depending on the architecture, ZERO_PAGE is
either constant or gets initialized even earlier, at paging_init, so
there is no issue with initializing zero_pfn earlier.
Link: https://lkml.kernel.org/r/CALCv0x2YqOXEAy2Q=hafjhHCtTHVodChv1qpM=niAXOpqEbt7w@mail.gmail.com
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: stable@vger.kernel.org
Tested-by: 周琰杰 (Zhou Yanjie) <zhouyanjie@wanyeetech.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-03-30 04:42:08 +00:00
|
|
|
early_initcall(init_zero_pfn);
|
2006-02-16 22:41:58 +00:00
|
|
|
|
2022-10-24 05:28:41 +00:00
|
|
|
void mm_trace_rss_stat(struct mm_struct *mm, int member)
|
mm: emit tracepoint when RSS changes
Useful to track how RSS is changing per TGID to detect spikes in RSS and
memory hogs. Several Android teams have been using this patch in
various kernel trees for half a year now. Many reported to me it is
really useful so I'm posting it upstream.
Initial patch developed by Tim Murray. Changes I made from original
patch: o Prevent any additional space consumed by mm_struct.
Regarding the fact that the RSS may change too often thus flooding the
traces - note that, there is some "hysterisis" with this already. That
is - We update the counter only if we receive 64 page faults due to
SPLIT_RSS_ACCOUNTING. However, during zapping or copying of pte range,
the RSS is updated immediately which can become noisy/flooding. In a
previous discussion, we agreed that BPF or ftrace can be used to rate
limit the signal if this becomes an issue.
Also note that I added wrappers to trace_rss_stat to prevent compiler
errors where linux/mm.h is included from tracing code, causing errors
such as:
CC kernel/trace/power-traces.o
In file included from ./include/trace/define_trace.h:102,
from ./include/trace/events/kmem.h:342,
from ./include/linux/mm.h:31,
from ./include/linux/ring_buffer.h:5,
from ./include/linux/trace_events.h:6,
from ./include/trace/events/power.h:12,
from kernel/trace/power-traces.c:15:
./include/trace/trace_events.h:113:22: error: field `ent' has incomplete type
struct trace_entry ent; \
Link: http://lore.kernel.org/r/20190903200905.198642-1-joel@joelfernandes.org
Link: http://lkml.kernel.org/r/20191001172817.234886-1-joel@joelfernandes.org
Co-developed-by: Tim Murray <timmurray@google.com>
Signed-off-by: Tim Murray <timmurray@google.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Carmen Jackson <carmenjackson@google.com>
Cc: Mayank Gupta <mayankgupta@google.com>
Cc: Daniel Colascione <dancol@google.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-12-01 01:50:30 +00:00
|
|
|
{
|
2022-10-24 05:28:41 +00:00
|
|
|
trace_rss_stat(mm, member);
|
mm: emit tracepoint when RSS changes
Useful to track how RSS is changing per TGID to detect spikes in RSS and
memory hogs. Several Android teams have been using this patch in
various kernel trees for half a year now. Many reported to me it is
really useful so I'm posting it upstream.
Initial patch developed by Tim Murray. Changes I made from original
patch: o Prevent any additional space consumed by mm_struct.
Regarding the fact that the RSS may change too often thus flooding the
traces - note that, there is some "hysterisis" with this already. That
is - We update the counter only if we receive 64 page faults due to
SPLIT_RSS_ACCOUNTING. However, during zapping or copying of pte range,
the RSS is updated immediately which can become noisy/flooding. In a
previous discussion, we agreed that BPF or ftrace can be used to rate
limit the signal if this becomes an issue.
Also note that I added wrappers to trace_rss_stat to prevent compiler
errors where linux/mm.h is included from tracing code, causing errors
such as:
CC kernel/trace/power-traces.o
In file included from ./include/trace/define_trace.h:102,
from ./include/trace/events/kmem.h:342,
from ./include/linux/mm.h:31,
from ./include/linux/ring_buffer.h:5,
from ./include/linux/trace_events.h:6,
from ./include/trace/events/power.h:12,
from kernel/trace/power-traces.c:15:
./include/trace/trace_events.h:113:22: error: field `ent' has incomplete type
struct trace_entry ent; \
Link: http://lore.kernel.org/r/20190903200905.198642-1-joel@joelfernandes.org
Link: http://lkml.kernel.org/r/20191001172817.234886-1-joel@joelfernandes.org
Co-developed-by: Tim Murray <timmurray@google.com>
Signed-off-by: Tim Murray <timmurray@google.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Carmen Jackson <carmenjackson@google.com>
Cc: Mayank Gupta <mayankgupta@google.com>
Cc: Daniel Colascione <dancol@google.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-12-01 01:50:30 +00:00
|
|
|
}
|
2010-03-05 21:41:39 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Note: this doesn't free the actual pages themselves. That
|
|
|
|
* has been handled earlier when unmapping all the memory regions.
|
|
|
|
*/
|
mm: Pass virtual address to [__]p{te,ud,md}_free_tlb()
mm: Pass virtual address to [__]p{te,ud,md}_free_tlb()
Upcoming paches to support the new 64-bit "BookE" powerpc architecture
will need to have the virtual address corresponding to PTE page when
freeing it, due to the way the HW table walker works.
Basically, the TLB can be loaded with "large" pages that cover the whole
virtual space (well, sort-of, half of it actually) represented by a PTE
page, and which contain an "indirect" bit indicating that this TLB entry
RPN points to an array of PTEs from which the TLB can then create direct
entries. Thus, in order to invalidate those when PTE pages are deleted,
we need the virtual address to pass to tlbilx or tlbivax instructions.
The old trick of sticking it somewhere in the PTE page struct page sucks
too much, the address is almost readily available in all call sites and
almost everybody implemets these as macros, so we may as well add the
argument everywhere. I added it to the pmd and pud variants for consistency.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: David Howells <dhowells@redhat.com> [MN10300 & FRV]
Acked-by: Nick Piggin <npiggin@suse.de>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com> [s390]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-07-22 05:44:28 +00:00
|
|
|
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
|
|
|
|
unsigned long addr)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2008-02-08 12:22:04 +00:00
|
|
|
pgtable_t token = pmd_pgtable(*pmd);
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
pmd_clear(pmd);
|
mm: Pass virtual address to [__]p{te,ud,md}_free_tlb()
mm: Pass virtual address to [__]p{te,ud,md}_free_tlb()
Upcoming paches to support the new 64-bit "BookE" powerpc architecture
will need to have the virtual address corresponding to PTE page when
freeing it, due to the way the HW table walker works.
Basically, the TLB can be loaded with "large" pages that cover the whole
virtual space (well, sort-of, half of it actually) represented by a PTE
page, and which contain an "indirect" bit indicating that this TLB entry
RPN points to an array of PTEs from which the TLB can then create direct
entries. Thus, in order to invalidate those when PTE pages are deleted,
we need the virtual address to pass to tlbilx or tlbivax instructions.
The old trick of sticking it somewhere in the PTE page struct page sucks
too much, the address is almost readily available in all call sites and
almost everybody implemets these as macros, so we may as well add the
argument everywhere. I added it to the pmd and pud variants for consistency.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: David Howells <dhowells@redhat.com> [MN10300 & FRV]
Acked-by: Nick Piggin <npiggin@suse.de>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com> [s390]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-07-22 05:44:28 +00:00
|
|
|
pte_free_tlb(tlb, token, addr);
|
2017-11-16 01:35:37 +00:00
|
|
|
mm_dec_nr_ptes(tlb->mm);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
static inline void free_pmd_range(struct mmu_gather *tlb, pud_t *pud,
|
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
unsigned long floor, unsigned long ceiling)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pmd_t *pmd;
|
|
|
|
unsigned long next;
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
unsigned long start;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
start = addr;
|
2005-04-16 22:20:36 +00:00
|
|
|
pmd = pmd_offset(pud, addr);
|
|
|
|
do {
|
|
|
|
next = pmd_addr_end(addr, end);
|
|
|
|
if (pmd_none_or_clear_bad(pmd))
|
|
|
|
continue;
|
mm: Pass virtual address to [__]p{te,ud,md}_free_tlb()
mm: Pass virtual address to [__]p{te,ud,md}_free_tlb()
Upcoming paches to support the new 64-bit "BookE" powerpc architecture
will need to have the virtual address corresponding to PTE page when
freeing it, due to the way the HW table walker works.
Basically, the TLB can be loaded with "large" pages that cover the whole
virtual space (well, sort-of, half of it actually) represented by a PTE
page, and which contain an "indirect" bit indicating that this TLB entry
RPN points to an array of PTEs from which the TLB can then create direct
entries. Thus, in order to invalidate those when PTE pages are deleted,
we need the virtual address to pass to tlbilx or tlbivax instructions.
The old trick of sticking it somewhere in the PTE page struct page sucks
too much, the address is almost readily available in all call sites and
almost everybody implemets these as macros, so we may as well add the
argument everywhere. I added it to the pmd and pud variants for consistency.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: David Howells <dhowells@redhat.com> [MN10300 & FRV]
Acked-by: Nick Piggin <npiggin@suse.de>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com> [s390]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-07-22 05:44:28 +00:00
|
|
|
free_pte_range(tlb, pmd, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
} while (pmd++, addr = next, addr != end);
|
|
|
|
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
start &= PUD_MASK;
|
|
|
|
if (start < floor)
|
|
|
|
return;
|
|
|
|
if (ceiling) {
|
|
|
|
ceiling &= PUD_MASK;
|
|
|
|
if (!ceiling)
|
|
|
|
return;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
if (end - 1 > ceiling - 1)
|
|
|
|
return;
|
|
|
|
|
|
|
|
pmd = pmd_offset(pud, start);
|
|
|
|
pud_clear(pud);
|
mm: Pass virtual address to [__]p{te,ud,md}_free_tlb()
mm: Pass virtual address to [__]p{te,ud,md}_free_tlb()
Upcoming paches to support the new 64-bit "BookE" powerpc architecture
will need to have the virtual address corresponding to PTE page when
freeing it, due to the way the HW table walker works.
Basically, the TLB can be loaded with "large" pages that cover the whole
virtual space (well, sort-of, half of it actually) represented by a PTE
page, and which contain an "indirect" bit indicating that this TLB entry
RPN points to an array of PTEs from which the TLB can then create direct
entries. Thus, in order to invalidate those when PTE pages are deleted,
we need the virtual address to pass to tlbilx or tlbivax instructions.
The old trick of sticking it somewhere in the PTE page struct page sucks
too much, the address is almost readily available in all call sites and
almost everybody implemets these as macros, so we may as well add the
argument everywhere. I added it to the pmd and pud variants for consistency.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: David Howells <dhowells@redhat.com> [MN10300 & FRV]
Acked-by: Nick Piggin <npiggin@suse.de>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com> [s390]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-07-22 05:44:28 +00:00
|
|
|
pmd_free_tlb(tlb, pmd, start);
|
mm: account pmd page tables to the process
Dave noticed that unprivileged process can allocate significant amount of
memory -- >500 MiB on x86_64 -- and stay unnoticed by oom-killer and
memory cgroup. The trick is to allocate a lot of PMD page tables. Linux
kernel doesn't account PMD tables to the process, only PTE.
The use-cases below use few tricks to allocate a lot of PMD page tables
while keeping VmRSS and VmPTE low. oom_score for the process will be 0.
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/prctl.h>
#define PUD_SIZE (1UL << 30)
#define PMD_SIZE (1UL << 21)
#define NR_PUD 130000
int main(void)
{
char *addr = NULL;
unsigned long i;
prctl(PR_SET_THP_DISABLE);
for (i = 0; i < NR_PUD ; i++) {
addr = mmap(addr + PUD_SIZE, PUD_SIZE, PROT_WRITE|PROT_READ,
MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
if (addr == MAP_FAILED) {
perror("mmap");
break;
}
*addr = 'x';
munmap(addr, PMD_SIZE);
mmap(addr, PMD_SIZE, PROT_WRITE|PROT_READ,
MAP_ANONYMOUS|MAP_PRIVATE|MAP_FIXED, -1, 0);
if (addr == MAP_FAILED)
perror("re-mmap"), exit(1);
}
printf("PID %d consumed %lu KiB in PMD page tables\n",
getpid(), i * 4096 >> 10);
return pause();
}
The patch addresses the issue by account PMD tables to the process the
same way we account PTE.
The main place where PMD tables is accounted is __pmd_alloc() and
free_pmd_range(). But there're few corner cases:
- HugeTLB can share PMD page tables. The patch handles by accounting
the table to all processes who share it.
- x86 PAE pre-allocates few PMD tables on fork.
- Architectures with FIRST_USER_ADDRESS > 0. We need to adjust sanity
check on exit(2).
Accounting only happens on configuration where PMD page table's level is
present (PMD is not folded). As with nr_ptes we use per-mm counter. The
counter value is used to calculate baseline for badness score by
oom-killer.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: David Rientjes <rientjes@google.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:26:50 +00:00
|
|
|
mm_dec_nr_pmds(tlb->mm);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
static inline void free_pud_range(struct mmu_gather *tlb, p4d_t *p4d,
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
unsigned long floor, unsigned long ceiling)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pud_t *pud;
|
|
|
|
unsigned long next;
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
unsigned long start;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
start = addr;
|
2017-03-09 14:24:07 +00:00
|
|
|
pud = pud_offset(p4d, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
do {
|
|
|
|
next = pud_addr_end(addr, end);
|
|
|
|
if (pud_none_or_clear_bad(pud))
|
|
|
|
continue;
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
free_pmd_range(tlb, pud, addr, next, floor, ceiling);
|
2005-04-16 22:20:36 +00:00
|
|
|
} while (pud++, addr = next, addr != end);
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
start &= P4D_MASK;
|
|
|
|
if (start < floor)
|
|
|
|
return;
|
|
|
|
if (ceiling) {
|
|
|
|
ceiling &= P4D_MASK;
|
|
|
|
if (!ceiling)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
if (end - 1 > ceiling - 1)
|
|
|
|
return;
|
|
|
|
|
|
|
|
pud = pud_offset(p4d, start);
|
|
|
|
p4d_clear(p4d);
|
|
|
|
pud_free_tlb(tlb, pud, start);
|
2017-11-16 01:35:33 +00:00
|
|
|
mm_dec_nr_puds(tlb->mm);
|
2017-03-09 14:24:07 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void free_p4d_range(struct mmu_gather *tlb, pgd_t *pgd,
|
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
unsigned long floor, unsigned long ceiling)
|
|
|
|
{
|
|
|
|
p4d_t *p4d;
|
|
|
|
unsigned long next;
|
|
|
|
unsigned long start;
|
|
|
|
|
|
|
|
start = addr;
|
|
|
|
p4d = p4d_offset(pgd, addr);
|
|
|
|
do {
|
|
|
|
next = p4d_addr_end(addr, end);
|
|
|
|
if (p4d_none_or_clear_bad(p4d))
|
|
|
|
continue;
|
|
|
|
free_pud_range(tlb, p4d, addr, next, floor, ceiling);
|
|
|
|
} while (p4d++, addr = next, addr != end);
|
|
|
|
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
start &= PGDIR_MASK;
|
|
|
|
if (start < floor)
|
|
|
|
return;
|
|
|
|
if (ceiling) {
|
|
|
|
ceiling &= PGDIR_MASK;
|
|
|
|
if (!ceiling)
|
|
|
|
return;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
if (end - 1 > ceiling - 1)
|
|
|
|
return;
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d = p4d_offset(pgd, start);
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
pgd_clear(pgd);
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d_free_tlb(tlb, p4d, start);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
* This function frees user-level page tables of a process.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2008-07-24 04:27:10 +00:00
|
|
|
void free_pgd_range(struct mmu_gather *tlb,
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
unsigned long floor, unsigned long ceiling)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pgd_t *pgd;
|
|
|
|
unsigned long next;
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The next few lines have given us lots of grief...
|
|
|
|
*
|
|
|
|
* Why are we testing PMD* at this top level? Because often
|
|
|
|
* there will be no work to do at all, and we'd prefer not to
|
|
|
|
* go all the way down to the bottom just to discover that.
|
|
|
|
*
|
|
|
|
* Why all these "- 1"s? Because 0 represents both the bottom
|
|
|
|
* of the address space and the top of it (using -1 for the
|
|
|
|
* top wouldn't help much: the masks would do the wrong thing).
|
|
|
|
* The rule is that addr 0 and floor 0 refer to the bottom of
|
|
|
|
* the address space, but end 0 and ceiling 0 refer to the top
|
|
|
|
* Comparisons need to use "end - 1" and "ceiling - 1" (though
|
|
|
|
* that end 0 case should be mythical).
|
|
|
|
*
|
|
|
|
* Wherever addr is brought up or ceiling brought down, we must
|
|
|
|
* be careful to reject "the opposite 0" before it confuses the
|
|
|
|
* subsequent tests. But what about where end is brought down
|
|
|
|
* by PMD_SIZE below? no, end can't go down to 0 there.
|
|
|
|
*
|
|
|
|
* Whereas we round start (addr) and ceiling down, by different
|
|
|
|
* masks at different levels, in order to test whether a table
|
|
|
|
* now has no other vmas using it, so can be freed, we don't
|
|
|
|
* bother to round floor or end up - the tests don't need that.
|
|
|
|
*/
|
2005-04-16 22:20:36 +00:00
|
|
|
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
addr &= PMD_MASK;
|
|
|
|
if (addr < floor) {
|
|
|
|
addr += PMD_SIZE;
|
|
|
|
if (!addr)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
if (ceiling) {
|
|
|
|
ceiling &= PMD_MASK;
|
|
|
|
if (!ceiling)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
if (end - 1 > ceiling - 1)
|
|
|
|
end -= PMD_SIZE;
|
|
|
|
if (addr > end - 1)
|
|
|
|
return;
|
2016-12-13 00:42:40 +00:00
|
|
|
/*
|
|
|
|
* We add page table cache pages with PAGE_SIZE,
|
|
|
|
* (see pte_free_tlb()), flush the tlb if we need
|
|
|
|
*/
|
2018-08-31 12:46:08 +00:00
|
|
|
tlb_change_page_size(tlb, PAGE_SIZE);
|
2008-07-24 04:27:10 +00:00
|
|
|
pgd = pgd_offset(tlb->mm, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
do {
|
|
|
|
next = pgd_addr_end(addr, end);
|
|
|
|
if (pgd_none_or_clear_bad(pgd))
|
|
|
|
continue;
|
2017-03-09 14:24:07 +00:00
|
|
|
free_p4d_range(tlb, pgd, addr, next, floor, ceiling);
|
2005-04-16 22:20:36 +00:00
|
|
|
} while (pgd++, addr = next, addr != end);
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
}
|
|
|
|
|
2022-09-06 19:49:06 +00:00
|
|
|
void free_pgtables(struct mmu_gather *tlb, struct maple_tree *mt,
|
|
|
|
struct vm_area_struct *vma, unsigned long floor,
|
2023-02-27 17:36:18 +00:00
|
|
|
unsigned long ceiling, bool mm_wr_locked)
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
{
|
2022-09-06 19:49:06 +00:00
|
|
|
MA_STATE(mas, mt, vma->vm_end, vma->vm_end);
|
|
|
|
|
|
|
|
do {
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
unsigned long addr = vma->vm_start;
|
2022-09-06 19:49:06 +00:00
|
|
|
struct vm_area_struct *next;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Note: USER_PGTABLES_CEILING may be passed as ceiling and may
|
|
|
|
* be 0. This will underflow and is okay.
|
|
|
|
*/
|
|
|
|
next = mas_find(&mas, ceiling - 1);
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
|
[PATCH] mm: unlink vma before pagetables
In most places the descent from pgd to pud to pmd to pte holds mmap_sem
(exclusively or not), which ensures that free_pgtables cannot be freeing page
tables from any level at the same time. But truncation and reverse mapping
descend without mmap_sem.
No problem: just make sure that a vma is unlinked from its prio_tree (or
nonlinear list) and from its anon_vma list, after zapping the vma, but before
freeing its page tables. Then neither vmtruncate nor rmap can reach that vma
whose page tables are now volatile (nor do they need to reach it, since all
its page entries have been zapped by this stage).
The i_mmap_lock and anon_vma->lock already serialize this correctly; but the
locking hierarchy is such that we cannot take them while holding
page_table_lock. Well, we're trying to push that down anyway. So in this
patch, move anon_vma_unlink and unlink_file_vma into free_pgtables, at the
same time as moving page_table_lock around calls to unmap_vmas.
tlb_gather_mmu and tlb_finish_mmu then fall outside the page_table_lock, but
we made them preempt_disable and preempt_enable earlier; and a long source
audit of all the architectures has shown no problem with removing
page_table_lock from them. free_pgtables doesn't need page_table_lock for
itself, nor for what it calls; tlb->mm->nr_ptes is usually protected by
page_table_lock, but partly by non-exclusive mmap_sem - here it's decremented
with exclusive mmap_sem, or mm_users 0. update_hiwater_rss and
vm_unacct_memory don't need page_table_lock either.
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 01:16:29 +00:00
|
|
|
/*
|
2009-08-20 16:35:05 +00:00
|
|
|
* Hide vma from rmap and truncate_pagecache before freeing
|
|
|
|
* pgtables
|
[PATCH] mm: unlink vma before pagetables
In most places the descent from pgd to pud to pmd to pte holds mmap_sem
(exclusively or not), which ensures that free_pgtables cannot be freeing page
tables from any level at the same time. But truncation and reverse mapping
descend without mmap_sem.
No problem: just make sure that a vma is unlinked from its prio_tree (or
nonlinear list) and from its anon_vma list, after zapping the vma, but before
freeing its page tables. Then neither vmtruncate nor rmap can reach that vma
whose page tables are now volatile (nor do they need to reach it, since all
its page entries have been zapped by this stage).
The i_mmap_lock and anon_vma->lock already serialize this correctly; but the
locking hierarchy is such that we cannot take them while holding
page_table_lock. Well, we're trying to push that down anyway. So in this
patch, move anon_vma_unlink and unlink_file_vma into free_pgtables, at the
same time as moving page_table_lock around calls to unmap_vmas.
tlb_gather_mmu and tlb_finish_mmu then fall outside the page_table_lock, but
we made them preempt_disable and preempt_enable earlier; and a long source
audit of all the architectures has shown no problem with removing
page_table_lock from them. free_pgtables doesn't need page_table_lock for
itself, nor for what it calls; tlb->mm->nr_ptes is usually protected by
page_table_lock, but partly by non-exclusive mmap_sem - here it's decremented
with exclusive mmap_sem, or mm_users 0. update_hiwater_rss and
vm_unacct_memory don't need page_table_lock either.
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 01:16:29 +00:00
|
|
|
*/
|
2023-02-27 17:36:18 +00:00
|
|
|
if (mm_wr_locked)
|
|
|
|
vma_start_write(vma);
|
mm: change anon_vma linking to fix multi-process server scalability issue
The old anon_vma code can lead to scalability issues with heavily forking
workloads. Specifically, each anon_vma will be shared between the parent
process and all its child processes.
In a workload with 1000 child processes and a VMA with 1000 anonymous
pages per process that get COWed, this leads to a system with a million
anonymous pages in the same anon_vma, each of which is mapped in just one
of the 1000 processes. However, the current rmap code needs to walk them
all, leading to O(N) scanning complexity for each page.
This can result in systems where one CPU is walking the page tables of
1000 processes in page_referenced_one, while all other CPUs are stuck on
the anon_vma lock. This leads to catastrophic failure for a benchmark
like AIM7, where the total number of processes can reach in the tens of
thousands. Real workloads are still a factor 10 less process intensive
than AIM7, but they are catching up.
This patch changes the way anon_vmas and VMAs are linked, which allows us
to associate multiple anon_vmas with a VMA. At fork time, each child
process gets its own anon_vmas, in which its COWed pages will be
instantiated. The parents' anon_vma is also linked to the VMA, because
non-COWed pages could be present in any of the children.
This reduces rmap scanning complexity to O(1) for the pages of the 1000
child processes, with O(N) complexity for at most 1/N pages in the system.
This reduces the average scanning cost in heavily forking workloads from
O(N) to 2.
The only real complexity in this patch stems from the fact that linking a
VMA to anon_vmas now involves memory allocations. This means vma_adjust
can fail, if it needs to attach a VMA to anon_vma structures. This in
turn means error handling needs to be added to the calling functions.
A second source of complexity is that, because there can be multiple
anon_vmas, the anon_vma linking in vma_adjust can no longer be done under
"the" anon_vma lock. To prevent the rmap code from walking up an
incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit
flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h
to make sure it is impossible to compile a kernel that needs both symbolic
values for the same bitflag.
Some test results:
Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test
box with 16GB RAM and not quite enough IO), the system ends up running
>99% in system time, with every CPU on the same anon_vma lock in the
pageout code.
With these changes, AIM7 hits the cross-over point around 29.7k users.
This happens with ~99% IO wait time, there never seems to be any spike in
system time. The anon_vma lock contention appears to be resolved.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Rik van Riel <riel@redhat.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Larry Woodman <lwoodman@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 21:42:07 +00:00
|
|
|
unlink_anon_vmas(vma);
|
[PATCH] mm: unlink vma before pagetables
In most places the descent from pgd to pud to pmd to pte holds mmap_sem
(exclusively or not), which ensures that free_pgtables cannot be freeing page
tables from any level at the same time. But truncation and reverse mapping
descend without mmap_sem.
No problem: just make sure that a vma is unlinked from its prio_tree (or
nonlinear list) and from its anon_vma list, after zapping the vma, but before
freeing its page tables. Then neither vmtruncate nor rmap can reach that vma
whose page tables are now volatile (nor do they need to reach it, since all
its page entries have been zapped by this stage).
The i_mmap_lock and anon_vma->lock already serialize this correctly; but the
locking hierarchy is such that we cannot take them while holding
page_table_lock. Well, we're trying to push that down anyway. So in this
patch, move anon_vma_unlink and unlink_file_vma into free_pgtables, at the
same time as moving page_table_lock around calls to unmap_vmas.
tlb_gather_mmu and tlb_finish_mmu then fall outside the page_table_lock, but
we made them preempt_disable and preempt_enable earlier; and a long source
audit of all the architectures has shown no problem with removing
page_table_lock from them. free_pgtables doesn't need page_table_lock for
itself, nor for what it calls; tlb->mm->nr_ptes is usually protected by
page_table_lock, but partly by non-exclusive mmap_sem - here it's decremented
with exclusive mmap_sem, or mm_users 0. update_hiwater_rss and
vm_unacct_memory don't need page_table_lock either.
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 01:16:29 +00:00
|
|
|
unlink_file_vma(vma);
|
|
|
|
|
[PATCH] hugepage: Fix hugepage logic in free_pgtables()
free_pgtables() has special logic to call hugetlb_free_pgd_range() instead
of the normal free_pgd_range() on hugepage VMAs. However, the test it uses
to do so is incorrect: it calls is_hugepage_only_range on a hugepage sized
range at the start of the vma. is_hugepage_only_range() will return true
if the given range has any intersection with a hugepage address region, and
in this case the given region need not be hugepage aligned. So, for
example, this test can return true if called on, say, a 4k VMA immediately
preceding a (nicely aligned) hugepage VMA.
At present we get away with this because the powerpc version of
hugetlb_free_pgd_range() is just a call to free_pgd_range(). On ia64 (the
only other arch with a non-trivial is_hugepage_only_range()) we get away
with it for a different reason; the hugepage area is not contiguous with
the rest of the user address space, and VMAs are not permitted in between,
so the test can't return a false positive there.
Nonetheless this should be fixed. We do that in the patch below by
replacing the is_hugepage_only_range() test with an explicit test of the
VMA using is_vm_hugetlb_page().
This in turn changes behaviour for platforms where is_hugepage_only_range()
returns false always (everything except powerpc and ia64). We address this
by ensuring that hugetlb_free_pgd_range() is defined to be identical to
free_pgd_range() (instead of a no-op) on everything except ia64. Even so,
it will prevent some otherwise possible coalescing of calls down to
free_pgd_range(). Since this only happens for hugepage VMAs, removing this
small optimization seems unlikely to cause any trouble.
This patch causes no regressions on the libhugetlbfs testsuite - ppc64
POWER5 (8-way), ppc64 G5 (2-way) and i386 Pentium M (UP).
Signed-off-by: David Gibson <dwg@au1.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:57 +00:00
|
|
|
if (is_vm_hugetlb_page(vma)) {
|
2005-04-19 20:29:16 +00:00
|
|
|
hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
|
2017-02-24 22:59:01 +00:00
|
|
|
floor, next ? next->vm_start : ceiling);
|
2005-04-19 20:29:16 +00:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Optimization: gather nearby vmas into one call down
|
|
|
|
*/
|
|
|
|
while (next && next->vm_start <= vma->vm_end + PMD_SIZE
|
2006-03-22 08:08:58 +00:00
|
|
|
&& !is_vm_hugetlb_page(next)) {
|
2005-04-19 20:29:16 +00:00
|
|
|
vma = next;
|
2022-09-06 19:49:06 +00:00
|
|
|
next = mas_find(&mas, ceiling - 1);
|
2023-02-27 17:36:18 +00:00
|
|
|
if (mm_wr_locked)
|
|
|
|
vma_start_write(vma);
|
mm: change anon_vma linking to fix multi-process server scalability issue
The old anon_vma code can lead to scalability issues with heavily forking
workloads. Specifically, each anon_vma will be shared between the parent
process and all its child processes.
In a workload with 1000 child processes and a VMA with 1000 anonymous
pages per process that get COWed, this leads to a system with a million
anonymous pages in the same anon_vma, each of which is mapped in just one
of the 1000 processes. However, the current rmap code needs to walk them
all, leading to O(N) scanning complexity for each page.
This can result in systems where one CPU is walking the page tables of
1000 processes in page_referenced_one, while all other CPUs are stuck on
the anon_vma lock. This leads to catastrophic failure for a benchmark
like AIM7, where the total number of processes can reach in the tens of
thousands. Real workloads are still a factor 10 less process intensive
than AIM7, but they are catching up.
This patch changes the way anon_vmas and VMAs are linked, which allows us
to associate multiple anon_vmas with a VMA. At fork time, each child
process gets its own anon_vmas, in which its COWed pages will be
instantiated. The parents' anon_vma is also linked to the VMA, because
non-COWed pages could be present in any of the children.
This reduces rmap scanning complexity to O(1) for the pages of the 1000
child processes, with O(N) complexity for at most 1/N pages in the system.
This reduces the average scanning cost in heavily forking workloads from
O(N) to 2.
The only real complexity in this patch stems from the fact that linking a
VMA to anon_vmas now involves memory allocations. This means vma_adjust
can fail, if it needs to attach a VMA to anon_vma structures. This in
turn means error handling needs to be added to the calling functions.
A second source of complexity is that, because there can be multiple
anon_vmas, the anon_vma linking in vma_adjust can no longer be done under
"the" anon_vma lock. To prevent the rmap code from walking up an
incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit
flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h
to make sure it is impossible to compile a kernel that needs both symbolic
values for the same bitflag.
Some test results:
Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test
box with 16GB RAM and not quite enough IO), the system ends up running
>99% in system time, with every CPU on the same anon_vma lock in the
pageout code.
With these changes, AIM7 hits the cross-over point around 29.7k users.
This happens with ~99% IO wait time, there never seems to be any spike in
system time. The anon_vma lock contention appears to be resolved.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Rik van Riel <riel@redhat.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Larry Woodman <lwoodman@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-05 21:42:07 +00:00
|
|
|
unlink_anon_vmas(vma);
|
[PATCH] mm: unlink vma before pagetables
In most places the descent from pgd to pud to pmd to pte holds mmap_sem
(exclusively or not), which ensures that free_pgtables cannot be freeing page
tables from any level at the same time. But truncation and reverse mapping
descend without mmap_sem.
No problem: just make sure that a vma is unlinked from its prio_tree (or
nonlinear list) and from its anon_vma list, after zapping the vma, but before
freeing its page tables. Then neither vmtruncate nor rmap can reach that vma
whose page tables are now volatile (nor do they need to reach it, since all
its page entries have been zapped by this stage).
The i_mmap_lock and anon_vma->lock already serialize this correctly; but the
locking hierarchy is such that we cannot take them while holding
page_table_lock. Well, we're trying to push that down anyway. So in this
patch, move anon_vma_unlink and unlink_file_vma into free_pgtables, at the
same time as moving page_table_lock around calls to unmap_vmas.
tlb_gather_mmu and tlb_finish_mmu then fall outside the page_table_lock, but
we made them preempt_disable and preempt_enable earlier; and a long source
audit of all the architectures has shown no problem with removing
page_table_lock from them. free_pgtables doesn't need page_table_lock for
itself, nor for what it calls; tlb->mm->nr_ptes is usually protected by
page_table_lock, but partly by non-exclusive mmap_sem - here it's decremented
with exclusive mmap_sem, or mm_users 0. update_hiwater_rss and
vm_unacct_memory don't need page_table_lock either.
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 01:16:29 +00:00
|
|
|
unlink_file_vma(vma);
|
2005-04-19 20:29:16 +00:00
|
|
|
}
|
|
|
|
free_pgd_range(tlb, addr, vma->vm_end,
|
2017-02-24 22:59:01 +00:00
|
|
|
floor, next ? next->vm_start : ceiling);
|
2005-04-19 20:29:16 +00:00
|
|
|
}
|
[PATCH] freepgt: free_pgtables use vma list
Recent woes with some arches needing their own pgd_addr_end macro; and 4-level
clear_page_range regression since 2.6.10's clear_page_tables; and its
long-standing well-known inefficiency in searching throughout the higher-level
page tables for those few entries to clear and free: all can be blamed on
ignoring the list of vmas when we free page tables.
Replace exit_mmap's clear_page_range of the total user address space by
free_pgtables operating on the mm's vma list; unmap_region use it in the same
way, giving floor and ceiling beyond which it may not free tables. This
brings lmbench fork/exec/sh numbers back to 2.6.10 (unless preempt is enabled,
in which case latency fixes spoil unmap_vmas throughput).
Beware: the do_mmap_pgoff driver failure case must now use unmap_region
instead of zap_page_range, since a page table might have been allocated, and
can only be freed while it is touched by some vma.
Move free_pgtables from mmap.c to memory.c, where its lower levels are adapted
from the clear_page_range levels. (Most of free_pgtables' old code was
actually for a non-existent case, prev not properly set up, dating from before
hch gave us split_vma.) Pass mmu_gather** in the public interfaces, since we
might want to add latency lockdrops later; but no attempt to do so yet, going
by vma should itself reduce latency.
But what if is_hugepage_only_range? Those ia64 and ppc64 cases need careful
examination: put that off until a later patch of the series.
What of x86_64's 32bit vdso page __map_syscall32 maps outside any vma?
And the range to sparc64's flush_tlb_pgtables? It's less clear to me now that
we need to do more than is done here - every PMD_SIZE ever occupied will be
flushed, do we really have to flush every PGDIR_SIZE ever partially occupied?
A shame to complicate it unnecessarily.
Special thanks to David Miller for time spent repairing my ceilings.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-19 20:29:15 +00:00
|
|
|
vma = next;
|
2022-09-06 19:49:06 +00:00
|
|
|
} while (vma);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2021-11-05 20:38:38 +00:00
|
|
|
void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2021-11-05 20:38:38 +00:00
|
|
|
spinlock_t *ptl = pmd_lock(mm, pmd);
|
2005-10-30 01:16:22 +00:00
|
|
|
|
2011-01-13 23:46:43 +00:00
|
|
|
if (likely(pmd_none(*pmd))) { /* Has another populated it ? */
|
2017-11-16 01:35:37 +00:00
|
|
|
mm_inc_nr_ptes(mm);
|
2021-11-05 20:38:41 +00:00
|
|
|
/*
|
|
|
|
* Ensure all pte setup (eg. pte page lock and page clearing) are
|
|
|
|
* visible before the pte is made visible to other CPUs by being
|
|
|
|
* put into page tables.
|
|
|
|
*
|
|
|
|
* The other side of the story is the pointer chasing in the page
|
|
|
|
* table walking code (when walking the page table without locking;
|
|
|
|
* ie. most of the time). Fortunately, these data accesses consist
|
|
|
|
* of a chain of data-dependent loads, meaning most CPUs (alpha
|
|
|
|
* being the notable exception) will already guarantee loads are
|
|
|
|
* seen in-order. See the alpha page table accessors for the
|
|
|
|
* smp_rmb() barriers in page table walking code.
|
|
|
|
*/
|
|
|
|
smp_wmb(); /* Could be smp_wmb__xxx(before|after)_spin_lock */
|
2021-11-05 20:38:38 +00:00
|
|
|
pmd_populate(mm, pmd, *pte);
|
|
|
|
*pte = NULL;
|
2016-01-16 00:53:39 +00:00
|
|
|
}
|
2013-11-14 22:31:04 +00:00
|
|
|
spin_unlock(ptl);
|
2021-11-05 20:38:38 +00:00
|
|
|
}
|
|
|
|
|
mm: treewide: remove unused address argument from pte_alloc functions
Patch series "Add support for fast mremap".
This series speeds up the mremap(2) syscall by copying page tables at
the PMD level even for non-THP systems. There is concern that the extra
'address' argument that mremap passes to pte_alloc may do something
subtle architecture related in the future that may make the scheme not
work. Also we find that there is no point in passing the 'address' to
pte_alloc since its unused. This patch therefore removes this argument
tree-wide resulting in a nice negative diff as well. Also ensuring
along the way that the enabled architectures do not do anything funky
with the 'address' argument that goes unnoticed by the optimization.
Build and boot tested on x86-64. Build tested on arm64. The config
enablement patch for arm64 will be posted in the future after more
testing.
The changes were obtained by applying the following Coccinelle script.
(thanks Julia for answering all Coccinelle questions!).
Following fix ups were done manually:
* Removal of address argument from pte_fragment_alloc
* Removal of pte_alloc_one_fast definitions from m68k and microblaze.
// Options: --include-headers --no-includes
// Note: I split the 'identifier fn' line, so if you are manually
// running it, please unsplit it so it runs for you.
virtual patch
@pte_alloc_func_def depends on patch exists@
identifier E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
type T2;
@@
fn(...
- , T2 E2
)
{ ... }
@pte_alloc_func_proto_noarg depends on patch exists@
type T1, T2, T3, T4;
identifier fn =~ "^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1, T2);
+ T3 fn(T1);
|
- T3 fn(T1, T2, T4);
+ T3 fn(T1, T2);
)
@pte_alloc_func_proto depends on patch exists@
identifier E1, E2, E4;
type T1, T2, T3, T4;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1 E1, T2 E2);
+ T3 fn(T1 E1);
|
- T3 fn(T1 E1, T2 E2, T4 E4);
+ T3 fn(T1 E1, T2 E2);
)
@pte_alloc_func_call depends on patch exists@
expression E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
fn(...
-, E2
)
@pte_alloc_macro depends on patch exists@
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
identifier a, b, c;
expression e;
position p;
@@
(
- #define fn(a, b, c) e
+ #define fn(a, b) e
|
- #define fn(a, b) e
+ #define fn(a) e
)
Link: http://lkml.kernel.org/r/20181108181201.88826-2-joelaf@google.com
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Suggested-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: William Kucharski <william.kucharski@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-03 23:28:34 +00:00
|
|
|
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
mm: treewide: remove unused address argument from pte_alloc functions
Patch series "Add support for fast mremap".
This series speeds up the mremap(2) syscall by copying page tables at
the PMD level even for non-THP systems. There is concern that the extra
'address' argument that mremap passes to pte_alloc may do something
subtle architecture related in the future that may make the scheme not
work. Also we find that there is no point in passing the 'address' to
pte_alloc since its unused. This patch therefore removes this argument
tree-wide resulting in a nice negative diff as well. Also ensuring
along the way that the enabled architectures do not do anything funky
with the 'address' argument that goes unnoticed by the optimization.
Build and boot tested on x86-64. Build tested on arm64. The config
enablement patch for arm64 will be posted in the future after more
testing.
The changes were obtained by applying the following Coccinelle script.
(thanks Julia for answering all Coccinelle questions!).
Following fix ups were done manually:
* Removal of address argument from pte_fragment_alloc
* Removal of pte_alloc_one_fast definitions from m68k and microblaze.
// Options: --include-headers --no-includes
// Note: I split the 'identifier fn' line, so if you are manually
// running it, please unsplit it so it runs for you.
virtual patch
@pte_alloc_func_def depends on patch exists@
identifier E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
type T2;
@@
fn(...
- , T2 E2
)
{ ... }
@pte_alloc_func_proto_noarg depends on patch exists@
type T1, T2, T3, T4;
identifier fn =~ "^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1, T2);
+ T3 fn(T1);
|
- T3 fn(T1, T2, T4);
+ T3 fn(T1, T2);
)
@pte_alloc_func_proto depends on patch exists@
identifier E1, E2, E4;
type T1, T2, T3, T4;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1 E1, T2 E2);
+ T3 fn(T1 E1);
|
- T3 fn(T1 E1, T2 E2, T4 E4);
+ T3 fn(T1 E1, T2 E2);
)
@pte_alloc_func_call depends on patch exists@
expression E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
fn(...
-, E2
)
@pte_alloc_macro depends on patch exists@
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
identifier a, b, c;
expression e;
position p;
@@
(
- #define fn(a, b, c) e
+ #define fn(a, b) e
|
- #define fn(a, b) e
+ #define fn(a) e
)
Link: http://lkml.kernel.org/r/20181108181201.88826-2-joelaf@google.com
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Suggested-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: William Kucharski <william.kucharski@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-03 23:28:34 +00:00
|
|
|
pgtable_t new = pte_alloc_one(mm);
|
2005-10-30 01:16:22 +00:00
|
|
|
if (!new)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2021-11-05 20:38:38 +00:00
|
|
|
pmd_install(mm, pmd, &new);
|
2008-02-08 12:22:04 +00:00
|
|
|
if (new)
|
|
|
|
pte_free(mm, new);
|
2005-10-30 01:16:22 +00:00
|
|
|
return 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
mm: treewide: remove unused address argument from pte_alloc functions
Patch series "Add support for fast mremap".
This series speeds up the mremap(2) syscall by copying page tables at
the PMD level even for non-THP systems. There is concern that the extra
'address' argument that mremap passes to pte_alloc may do something
subtle architecture related in the future that may make the scheme not
work. Also we find that there is no point in passing the 'address' to
pte_alloc since its unused. This patch therefore removes this argument
tree-wide resulting in a nice negative diff as well. Also ensuring
along the way that the enabled architectures do not do anything funky
with the 'address' argument that goes unnoticed by the optimization.
Build and boot tested on x86-64. Build tested on arm64. The config
enablement patch for arm64 will be posted in the future after more
testing.
The changes were obtained by applying the following Coccinelle script.
(thanks Julia for answering all Coccinelle questions!).
Following fix ups were done manually:
* Removal of address argument from pte_fragment_alloc
* Removal of pte_alloc_one_fast definitions from m68k and microblaze.
// Options: --include-headers --no-includes
// Note: I split the 'identifier fn' line, so if you are manually
// running it, please unsplit it so it runs for you.
virtual patch
@pte_alloc_func_def depends on patch exists@
identifier E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
type T2;
@@
fn(...
- , T2 E2
)
{ ... }
@pte_alloc_func_proto_noarg depends on patch exists@
type T1, T2, T3, T4;
identifier fn =~ "^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1, T2);
+ T3 fn(T1);
|
- T3 fn(T1, T2, T4);
+ T3 fn(T1, T2);
)
@pte_alloc_func_proto depends on patch exists@
identifier E1, E2, E4;
type T1, T2, T3, T4;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1 E1, T2 E2);
+ T3 fn(T1 E1);
|
- T3 fn(T1 E1, T2 E2, T4 E4);
+ T3 fn(T1 E1, T2 E2);
)
@pte_alloc_func_call depends on patch exists@
expression E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
fn(...
-, E2
)
@pte_alloc_macro depends on patch exists@
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
identifier a, b, c;
expression e;
position p;
@@
(
- #define fn(a, b, c) e
+ #define fn(a, b) e
|
- #define fn(a, b) e
+ #define fn(a) e
)
Link: http://lkml.kernel.org/r/20181108181201.88826-2-joelaf@google.com
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Suggested-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: William Kucharski <william.kucharski@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-03 23:28:34 +00:00
|
|
|
int __pte_alloc_kernel(pmd_t *pmd)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
mm: treewide: remove unused address argument from pte_alloc functions
Patch series "Add support for fast mremap".
This series speeds up the mremap(2) syscall by copying page tables at
the PMD level even for non-THP systems. There is concern that the extra
'address' argument that mremap passes to pte_alloc may do something
subtle architecture related in the future that may make the scheme not
work. Also we find that there is no point in passing the 'address' to
pte_alloc since its unused. This patch therefore removes this argument
tree-wide resulting in a nice negative diff as well. Also ensuring
along the way that the enabled architectures do not do anything funky
with the 'address' argument that goes unnoticed by the optimization.
Build and boot tested on x86-64. Build tested on arm64. The config
enablement patch for arm64 will be posted in the future after more
testing.
The changes were obtained by applying the following Coccinelle script.
(thanks Julia for answering all Coccinelle questions!).
Following fix ups were done manually:
* Removal of address argument from pte_fragment_alloc
* Removal of pte_alloc_one_fast definitions from m68k and microblaze.
// Options: --include-headers --no-includes
// Note: I split the 'identifier fn' line, so if you are manually
// running it, please unsplit it so it runs for you.
virtual patch
@pte_alloc_func_def depends on patch exists@
identifier E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
type T2;
@@
fn(...
- , T2 E2
)
{ ... }
@pte_alloc_func_proto_noarg depends on patch exists@
type T1, T2, T3, T4;
identifier fn =~ "^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1, T2);
+ T3 fn(T1);
|
- T3 fn(T1, T2, T4);
+ T3 fn(T1, T2);
)
@pte_alloc_func_proto depends on patch exists@
identifier E1, E2, E4;
type T1, T2, T3, T4;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1 E1, T2 E2);
+ T3 fn(T1 E1);
|
- T3 fn(T1 E1, T2 E2, T4 E4);
+ T3 fn(T1 E1, T2 E2);
)
@pte_alloc_func_call depends on patch exists@
expression E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
fn(...
-, E2
)
@pte_alloc_macro depends on patch exists@
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
identifier a, b, c;
expression e;
position p;
@@
(
- #define fn(a, b, c) e
+ #define fn(a, b) e
|
- #define fn(a, b) e
+ #define fn(a) e
)
Link: http://lkml.kernel.org/r/20181108181201.88826-2-joelaf@google.com
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Suggested-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: William Kucharski <william.kucharski@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-03 23:28:34 +00:00
|
|
|
pte_t *new = pte_alloc_one_kernel(&init_mm);
|
2005-10-30 01:16:22 +00:00
|
|
|
if (!new)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
spin_lock(&init_mm.page_table_lock);
|
2011-01-13 23:46:43 +00:00
|
|
|
if (likely(pmd_none(*pmd))) { /* Has another populated it ? */
|
2021-11-05 20:38:41 +00:00
|
|
|
smp_wmb(); /* See comment in pmd_install() */
|
2005-10-30 01:16:22 +00:00
|
|
|
pmd_populate_kernel(&init_mm, pmd, new);
|
2008-02-08 12:22:04 +00:00
|
|
|
new = NULL;
|
2016-01-16 00:53:39 +00:00
|
|
|
}
|
2005-10-30 01:16:22 +00:00
|
|
|
spin_unlock(&init_mm.page_table_lock);
|
2008-02-08 12:22:04 +00:00
|
|
|
if (new)
|
|
|
|
pte_free_kernel(&init_mm, new);
|
2005-10-30 01:16:22 +00:00
|
|
|
return 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2010-03-05 21:41:39 +00:00
|
|
|
static inline void init_rss_vec(int *rss)
|
|
|
|
{
|
|
|
|
memset(rss, 0, sizeof(int) * NR_MM_COUNTERS);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void add_mm_rss_vec(struct mm_struct *mm, int *rss)
|
2005-10-30 01:16:05 +00:00
|
|
|
{
|
2010-03-05 21:41:39 +00:00
|
|
|
int i;
|
|
|
|
|
2010-03-05 21:41:40 +00:00
|
|
|
if (current->mm == mm)
|
2012-03-21 23:34:13 +00:00
|
|
|
sync_mm_rss(mm);
|
2010-03-05 21:41:39 +00:00
|
|
|
for (i = 0; i < NR_MM_COUNTERS; i++)
|
|
|
|
if (rss[i])
|
|
|
|
add_mm_counter(mm, i, rss[i]);
|
2005-10-30 01:16:05 +00:00
|
|
|
}
|
|
|
|
|
2005-10-30 01:16:12 +00:00
|
|
|
/*
|
2005-11-28 22:34:23 +00:00
|
|
|
* This function is called to print an error when a bad pte
|
|
|
|
* is found. For example, we might have a PFN-mapped pte in
|
|
|
|
* a region that doesn't allow it.
|
2005-10-30 01:16:12 +00:00
|
|
|
*
|
|
|
|
* The calling function must still handle the error.
|
|
|
|
*/
|
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
|
|
|
static void print_bad_pte(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
pte_t pte, struct page *page)
|
2005-10-30 01:16:12 +00:00
|
|
|
{
|
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
|
|
|
pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d_t *p4d = p4d_offset(pgd, addr);
|
|
|
|
pud_t *pud = pud_offset(p4d, addr);
|
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
|
|
|
pmd_t *pmd = pmd_offset(pud, addr);
|
|
|
|
struct address_space *mapping;
|
|
|
|
pgoff_t index;
|
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++;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
if (nr_unshown) {
|
2016-03-17 21:19:50 +00:00
|
|
|
pr_alert("BUG: Bad page map: %lu messages suppressed\n",
|
|
|
|
nr_unshown);
|
2009-01-06 22:40:12 +00:00
|
|
|
nr_unshown = 0;
|
|
|
|
}
|
|
|
|
nr_shown = 0;
|
|
|
|
}
|
|
|
|
if (nr_shown++ == 0)
|
|
|
|
resume = jiffies + 60 * HZ;
|
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
|
|
|
|
|
|
|
mapping = vma->vm_file ? vma->vm_file->f_mapping : NULL;
|
|
|
|
index = linear_page_index(vma, addr);
|
|
|
|
|
2016-03-17 21:19:50 +00:00
|
|
|
pr_alert("BUG: Bad page map in process %s pte:%08llx pmd:%08llx\n",
|
|
|
|
current->comm,
|
|
|
|
(long long)pte_val(pte), (long long)pmd_val(*pmd));
|
2010-03-10 23:20:43 +00:00
|
|
|
if (page)
|
2014-01-23 23:52:49 +00:00
|
|
|
dump_page(page, "bad pte");
|
2019-09-23 22:35:34 +00:00
|
|
|
pr_alert("addr:%px vm_flags:%08lx anon_vma:%px mapping:%px index:%lx\n",
|
2016-03-17 21:19:50 +00:00
|
|
|
(void *)addr, vma->vm_flags, vma->anon_vma, mapping, index);
|
2022-04-29 15:53:28 +00:00
|
|
|
pr_alert("file:%pD fault:%ps mmap:%ps read_folio:%ps\n",
|
2015-04-15 23:15:08 +00:00
|
|
|
vma->vm_file,
|
|
|
|
vma->vm_ops ? vma->vm_ops->fault : NULL,
|
|
|
|
vma->vm_file ? vma->vm_file->f_op->mmap : NULL,
|
2022-04-29 15:53:28 +00:00
|
|
|
mapping ? mapping->a_ops->read_folio : NULL);
|
2005-10-30 01:16:12 +00:00
|
|
|
dump_stack();
|
2013-01-21 06:47:39 +00:00
|
|
|
add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
|
2005-10-30 01:16:12 +00:00
|
|
|
}
|
|
|
|
|
[PATCH] unpaged: anon in VM_UNPAGED
copy_one_pte needs to copy the anonymous COWed pages in a VM_UNPAGED area,
zap_pte_range needs to free them, do_wp_page needs to COW them: just like
ordinary pages, not like the unpaged.
But recognizing them is a little subtle: because PageReserved is no longer a
condition for remap_pfn_range, we can now mmap all of /dev/mem (whether the
distro permits, and whether it's advisable on this or that architecture, is
another matter). So if we can see a PageAnon, it may not be ours to mess with
(or may be ours from elsewhere in the address space). I suspect there's an
entertaining insoluble self-referential problem here, but the page_is_anon
function does a good practical job, and MAP_PRIVATE PROT_WRITE VM_UNPAGED will
always be an odd choice.
In updating the comment on page_address_in_vma, noticed a potential NULL
dereference, in a path we don't actually take, but fixed it.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-22 05:32:18 +00:00
|
|
|
/*
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
* vm_normal_page -- This function gets the "struct page" associated with a pte.
|
2005-11-28 22:34:23 +00:00
|
|
|
*
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
* "Special" mappings do not wish to be associated with a "struct page" (either
|
|
|
|
* it doesn't exist, or it exists but they don't want to touch it). In this
|
|
|
|
* case, NULL is returned here. "Normal" mappings do have a struct page.
|
mm: introduce VM_MIXEDMAP
This series introduces some important infrastructure work. The overall result
is that:
1. We now support XIP backed filesystems using memory that have no
struct page allocated to them. And patches 6 and 7 actually implement
this for s390.
This is pretty important in a number of cases. As far as I understand,
in the case of virtualisation (eg. s390), each guest may mount a
readonly copy of the same filesystem (eg. the distro). Currently,
guests need to allocate struct pages for this image. So if you have
100 guests, you already need to allocate more memory for the struct
pages than the size of the image. I think. (Carsten?)
For other (eg. embedded) systems, you may have a very large non-
volatile filesystem. If you have to have struct pages for this, then
your RAM consumption will go up proportionally to fs size. Even
though it is just a small proportion, the RAM can be much more costly
eg in terms of power, so every KB less that Linux uses makes it more
attractive to a lot of these guys.
2. VM_MIXEDMAP allows us to support mappings where you actually do want
to refcount _some_ pages in the mapping, but not others, and support
COW on arbitrary (non-linear) mappings. Jared needs this for his NVRAM
filesystem in progress. Future iterations of this filesystem will
most likely want to migrate pages between pagecache and XIP backing,
which is where the requirement for mixed (some refcounted, some not)
comes from.
3. pte_special also has a peripheral usage that I need for my lockless
get_user_pages patch. That was shown to speed up "oltp" on db2 by
10% on a 2 socket system, which is kind of significant because they
scrounge for months to try to find 0.1% improvement on these
workloads. I'm hoping we might finally be faster than AIX on
pSeries with this :). My reference to lockless get_user_pages is not
meant to justify this patchset (which doesn't include lockless gup),
but just to show that pte_special is not some s390 specific thing that
should be hidden in arch code or xip code: I definitely want to use it
on at least x86 and powerpc as well.
This patch:
Introduce a new type of mapping, VM_MIXEDMAP. This is unlike VM_PFNMAP in
that it can support COW mappings of arbitrary ranges including ranges without
struct page *and* ranges with a struct page that we actually want to refcount
(PFNMAP can only support COW in those cases where the un-COW-ed translations
are mapped linearly in the virtual address, and can only support non
refcounted ranges).
VM_MIXEDMAP achieves this by refcounting all pfn_valid pages, and not
refcounting !pfn_valid pages (which is not an option for VM_PFNMAP, because it
needs to avoid refcounting pfn_valid pages eg. for /dev/mem mappings).
Signed-off-by: Jared Hulbert <jaredeh@gmail.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:12:58 +00:00
|
|
|
*
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
* There are 2 broad cases. Firstly, an architecture may define a pte_special()
|
|
|
|
* pte bit, in which case this function is trivial. Secondly, an architecture
|
|
|
|
* may not have a spare pte bit, which requires a more complicated scheme,
|
|
|
|
* described below.
|
|
|
|
*
|
|
|
|
* A raw VM_PFNMAP mapping (ie. one that is not COWed) is always considered a
|
|
|
|
* special mapping (even if there are underlying and valid "struct pages").
|
|
|
|
* COWed pages of a VM_PFNMAP are always normal.
|
2005-11-28 22:34:23 +00:00
|
|
|
*
|
mm: introduce VM_MIXEDMAP
This series introduces some important infrastructure work. The overall result
is that:
1. We now support XIP backed filesystems using memory that have no
struct page allocated to them. And patches 6 and 7 actually implement
this for s390.
This is pretty important in a number of cases. As far as I understand,
in the case of virtualisation (eg. s390), each guest may mount a
readonly copy of the same filesystem (eg. the distro). Currently,
guests need to allocate struct pages for this image. So if you have
100 guests, you already need to allocate more memory for the struct
pages than the size of the image. I think. (Carsten?)
For other (eg. embedded) systems, you may have a very large non-
volatile filesystem. If you have to have struct pages for this, then
your RAM consumption will go up proportionally to fs size. Even
though it is just a small proportion, the RAM can be much more costly
eg in terms of power, so every KB less that Linux uses makes it more
attractive to a lot of these guys.
2. VM_MIXEDMAP allows us to support mappings where you actually do want
to refcount _some_ pages in the mapping, but not others, and support
COW on arbitrary (non-linear) mappings. Jared needs this for his NVRAM
filesystem in progress. Future iterations of this filesystem will
most likely want to migrate pages between pagecache and XIP backing,
which is where the requirement for mixed (some refcounted, some not)
comes from.
3. pte_special also has a peripheral usage that I need for my lockless
get_user_pages patch. That was shown to speed up "oltp" on db2 by
10% on a 2 socket system, which is kind of significant because they
scrounge for months to try to find 0.1% improvement on these
workloads. I'm hoping we might finally be faster than AIX on
pSeries with this :). My reference to lockless get_user_pages is not
meant to justify this patchset (which doesn't include lockless gup),
but just to show that pte_special is not some s390 specific thing that
should be hidden in arch code or xip code: I definitely want to use it
on at least x86 and powerpc as well.
This patch:
Introduce a new type of mapping, VM_MIXEDMAP. This is unlike VM_PFNMAP in
that it can support COW mappings of arbitrary ranges including ranges without
struct page *and* ranges with a struct page that we actually want to refcount
(PFNMAP can only support COW in those cases where the un-COW-ed translations
are mapped linearly in the virtual address, and can only support non
refcounted ranges).
VM_MIXEDMAP achieves this by refcounting all pfn_valid pages, and not
refcounting !pfn_valid pages (which is not an option for VM_PFNMAP, because it
needs to avoid refcounting pfn_valid pages eg. for /dev/mem mappings).
Signed-off-by: Jared Hulbert <jaredeh@gmail.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:12:58 +00:00
|
|
|
* The way we recognize COWed pages within VM_PFNMAP mappings is through the
|
|
|
|
* rules set up by "remap_pfn_range()": the vma will have the VM_PFNMAP bit
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
* set, and the vm_pgoff will point to the first PFN mapped: thus every special
|
|
|
|
* mapping will always honor the rule
|
2005-11-28 22:34:23 +00:00
|
|
|
*
|
|
|
|
* pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
|
|
|
|
*
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
* And for normal mappings this is false.
|
|
|
|
*
|
|
|
|
* This restricts such mappings to be a linear translation from virtual address
|
|
|
|
* to pfn. To get around this restriction, we allow arbitrary mappings so long
|
|
|
|
* as the vma is not a COW mapping; in that case, we know that all ptes are
|
|
|
|
* special (because none can have been COWed).
|
mm: introduce VM_MIXEDMAP
This series introduces some important infrastructure work. The overall result
is that:
1. We now support XIP backed filesystems using memory that have no
struct page allocated to them. And patches 6 and 7 actually implement
this for s390.
This is pretty important in a number of cases. As far as I understand,
in the case of virtualisation (eg. s390), each guest may mount a
readonly copy of the same filesystem (eg. the distro). Currently,
guests need to allocate struct pages for this image. So if you have
100 guests, you already need to allocate more memory for the struct
pages than the size of the image. I think. (Carsten?)
For other (eg. embedded) systems, you may have a very large non-
volatile filesystem. If you have to have struct pages for this, then
your RAM consumption will go up proportionally to fs size. Even
though it is just a small proportion, the RAM can be much more costly
eg in terms of power, so every KB less that Linux uses makes it more
attractive to a lot of these guys.
2. VM_MIXEDMAP allows us to support mappings where you actually do want
to refcount _some_ pages in the mapping, but not others, and support
COW on arbitrary (non-linear) mappings. Jared needs this for his NVRAM
filesystem in progress. Future iterations of this filesystem will
most likely want to migrate pages between pagecache and XIP backing,
which is where the requirement for mixed (some refcounted, some not)
comes from.
3. pte_special also has a peripheral usage that I need for my lockless
get_user_pages patch. That was shown to speed up "oltp" on db2 by
10% on a 2 socket system, which is kind of significant because they
scrounge for months to try to find 0.1% improvement on these
workloads. I'm hoping we might finally be faster than AIX on
pSeries with this :). My reference to lockless get_user_pages is not
meant to justify this patchset (which doesn't include lockless gup),
but just to show that pte_special is not some s390 specific thing that
should be hidden in arch code or xip code: I definitely want to use it
on at least x86 and powerpc as well.
This patch:
Introduce a new type of mapping, VM_MIXEDMAP. This is unlike VM_PFNMAP in
that it can support COW mappings of arbitrary ranges including ranges without
struct page *and* ranges with a struct page that we actually want to refcount
(PFNMAP can only support COW in those cases where the un-COW-ed translations
are mapped linearly in the virtual address, and can only support non
refcounted ranges).
VM_MIXEDMAP achieves this by refcounting all pfn_valid pages, and not
refcounting !pfn_valid pages (which is not an option for VM_PFNMAP, because it
needs to avoid refcounting pfn_valid pages eg. for /dev/mem mappings).
Signed-off-by: Jared Hulbert <jaredeh@gmail.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:12:58 +00:00
|
|
|
*
|
|
|
|
*
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
* In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
|
mm: introduce VM_MIXEDMAP
This series introduces some important infrastructure work. The overall result
is that:
1. We now support XIP backed filesystems using memory that have no
struct page allocated to them. And patches 6 and 7 actually implement
this for s390.
This is pretty important in a number of cases. As far as I understand,
in the case of virtualisation (eg. s390), each guest may mount a
readonly copy of the same filesystem (eg. the distro). Currently,
guests need to allocate struct pages for this image. So if you have
100 guests, you already need to allocate more memory for the struct
pages than the size of the image. I think. (Carsten?)
For other (eg. embedded) systems, you may have a very large non-
volatile filesystem. If you have to have struct pages for this, then
your RAM consumption will go up proportionally to fs size. Even
though it is just a small proportion, the RAM can be much more costly
eg in terms of power, so every KB less that Linux uses makes it more
attractive to a lot of these guys.
2. VM_MIXEDMAP allows us to support mappings where you actually do want
to refcount _some_ pages in the mapping, but not others, and support
COW on arbitrary (non-linear) mappings. Jared needs this for his NVRAM
filesystem in progress. Future iterations of this filesystem will
most likely want to migrate pages between pagecache and XIP backing,
which is where the requirement for mixed (some refcounted, some not)
comes from.
3. pte_special also has a peripheral usage that I need for my lockless
get_user_pages patch. That was shown to speed up "oltp" on db2 by
10% on a 2 socket system, which is kind of significant because they
scrounge for months to try to find 0.1% improvement on these
workloads. I'm hoping we might finally be faster than AIX on
pSeries with this :). My reference to lockless get_user_pages is not
meant to justify this patchset (which doesn't include lockless gup),
but just to show that pte_special is not some s390 specific thing that
should be hidden in arch code or xip code: I definitely want to use it
on at least x86 and powerpc as well.
This patch:
Introduce a new type of mapping, VM_MIXEDMAP. This is unlike VM_PFNMAP in
that it can support COW mappings of arbitrary ranges including ranges without
struct page *and* ranges with a struct page that we actually want to refcount
(PFNMAP can only support COW in those cases where the un-COW-ed translations
are mapped linearly in the virtual address, and can only support non
refcounted ranges).
VM_MIXEDMAP achieves this by refcounting all pfn_valid pages, and not
refcounting !pfn_valid pages (which is not an option for VM_PFNMAP, because it
needs to avoid refcounting pfn_valid pages eg. for /dev/mem mappings).
Signed-off-by: Jared Hulbert <jaredeh@gmail.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:12:58 +00:00
|
|
|
*
|
|
|
|
* VM_MIXEDMAP mappings can likewise contain memory with or without "struct
|
|
|
|
* page" backing, however the difference is that _all_ pages with a struct
|
|
|
|
* page (that is, those where pfn_valid is true) are refcounted and considered
|
|
|
|
* normal pages by the VM. The disadvantage is that pages are refcounted
|
|
|
|
* (which can be slower and simply not an option for some PFNMAP users). The
|
|
|
|
* advantage is that we don't have to follow the strict linearity rule of
|
|
|
|
* PFNMAP mappings in order to support COWable mappings.
|
|
|
|
*
|
[PATCH] unpaged: anon in VM_UNPAGED
copy_one_pte needs to copy the anonymous COWed pages in a VM_UNPAGED area,
zap_pte_range needs to free them, do_wp_page needs to COW them: just like
ordinary pages, not like the unpaged.
But recognizing them is a little subtle: because PageReserved is no longer a
condition for remap_pfn_range, we can now mmap all of /dev/mem (whether the
distro permits, and whether it's advisable on this or that architecture, is
another matter). So if we can see a PageAnon, it may not be ours to mess with
(or may be ours from elsewhere in the address space). I suspect there's an
entertaining insoluble self-referential problem here, but the page_is_anon
function does a good practical job, and MAP_PRIVATE PROT_WRITE VM_UNPAGED will
always be an odd choice.
In updating the comment on page_address_in_vma, noticed a potential NULL
dereference, in a path we don't actually take, but fixed it.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-22 05:32:18 +00:00
|
|
|
*/
|
2019-06-13 20:50:49 +00:00
|
|
|
struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
pte_t pte)
|
[PATCH] unpaged: anon in VM_UNPAGED
copy_one_pte needs to copy the anonymous COWed pages in a VM_UNPAGED area,
zap_pte_range needs to free them, do_wp_page needs to COW them: just like
ordinary pages, not like the unpaged.
But recognizing them is a little subtle: because PageReserved is no longer a
condition for remap_pfn_range, we can now mmap all of /dev/mem (whether the
distro permits, and whether it's advisable on this or that architecture, is
another matter). So if we can see a PageAnon, it may not be ours to mess with
(or may be ours from elsewhere in the address space). I suspect there's an
entertaining insoluble self-referential problem here, but the page_is_anon
function does a good practical job, and MAP_PRIVATE PROT_WRITE VM_UNPAGED will
always be an odd choice.
In updating the comment on page_address_in_vma, noticed a potential NULL
dereference, in a path we don't actually take, but fixed it.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-22 05:32:18 +00:00
|
|
|
{
|
2009-01-06 22:40:09 +00:00
|
|
|
unsigned long pfn = pte_pfn(pte);
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
|
2018-06-08 00:06:12 +00:00
|
|
|
if (IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL)) {
|
x86,mm: fix pte_special versus pte_numa
Sasha Levin has shown oopses on ffffea0003480048 and ffffea0003480008 at
mm/memory.c:1132, running Trinity on different 3.16-rc-next kernels:
where zap_pte_range() checks page->mapping to see if PageAnon(page).
Those addresses fit struct pages for pfns d2001 and d2000, and in each
dump a register or a stack slot showed d2001730 or d2000730: pte flags
0x730 are PCD ACCESSED PROTNONE SPECIAL IOMAP; and Sasha's e820 map has
a hole between cfffffff and 100000000, which would need special access.
Commit c46a7c817e66 ("x86: define _PAGE_NUMA by reusing software bits on
the PMD and PTE levels") has broken vm_normal_page(): a PROTNONE SPECIAL
pte no longer passes the pte_special() test, so zap_pte_range() goes on
to try to access a non-existent struct page.
Fix this by refining pte_special() (SPECIAL with PRESENT or PROTNONE) to
complement pte_numa() (SPECIAL with neither PRESENT nor PROTNONE). A
hint that this was a problem was that c46a7c817e66 added pte_numa() test
to vm_normal_page(), and moved its is_zero_pfn() test from slow to fast
path: This was papering over a pte_special() snag when the zero page was
encountered during zap. This patch reverts vm_normal_page() to how it
was before, relying on pte_special().
It still appears that this patch may be incomplete: aren't there other
places which need to be handling PROTNONE along with PRESENT? For
example, pte_mknuma() clears _PAGE_PRESENT and sets _PAGE_NUMA, but on a
PROT_NONE area, that would make it pte_special(). This is side-stepped
by the fact that NUMA hinting faults skipped PROT_NONE VMAs and there
are no grounds where a NUMA hinting fault on a PROT_NONE VMA would be
interesting.
Fixes: c46a7c817e66 ("x86: define _PAGE_NUMA by reusing software bits on the PMD and PTE levels")
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Tested-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
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>
2014-08-29 22:18:44 +00:00
|
|
|
if (likely(!pte_special(pte)))
|
2009-01-06 22:40:09 +00:00
|
|
|
goto check_pfn;
|
2014-12-18 14:48:15 +00:00
|
|
|
if (vma->vm_ops && vma->vm_ops->find_special_page)
|
|
|
|
return vma->vm_ops->find_special_page(vma, addr);
|
2009-09-22 00:03:30 +00:00
|
|
|
if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
|
|
|
|
return NULL;
|
2017-09-08 23:12:24 +00:00
|
|
|
if (is_zero_pfn(pfn))
|
|
|
|
return NULL;
|
2018-08-17 22:43:40 +00:00
|
|
|
if (pte_devmap(pte))
|
2022-07-15 15:05:11 +00:00
|
|
|
/*
|
|
|
|
* NOTE: New users of ZONE_DEVICE will not set pte_devmap()
|
|
|
|
* and will have refcounts incremented on their struct pages
|
|
|
|
* when they are inserted into PTEs, thus they are safe to
|
|
|
|
* return here. Legacy ZONE_DEVICE pages that set pte_devmap()
|
|
|
|
* do not have refcounts. Example of legacy ZONE_DEVICE is
|
|
|
|
* MEMORY_DEVICE_FS_DAX type in pmem or virtio_fs drivers.
|
|
|
|
*/
|
2018-08-17 22:43:40 +00:00
|
|
|
return NULL;
|
|
|
|
|
2017-09-08 23:12:24 +00:00
|
|
|
print_bad_pte(vma, addr, pte, NULL);
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2018-06-08 00:06:12 +00:00
|
|
|
/* !CONFIG_ARCH_HAS_PTE_SPECIAL case follows: */
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
|
mm: introduce VM_MIXEDMAP
This series introduces some important infrastructure work. The overall result
is that:
1. We now support XIP backed filesystems using memory that have no
struct page allocated to them. And patches 6 and 7 actually implement
this for s390.
This is pretty important in a number of cases. As far as I understand,
in the case of virtualisation (eg. s390), each guest may mount a
readonly copy of the same filesystem (eg. the distro). Currently,
guests need to allocate struct pages for this image. So if you have
100 guests, you already need to allocate more memory for the struct
pages than the size of the image. I think. (Carsten?)
For other (eg. embedded) systems, you may have a very large non-
volatile filesystem. If you have to have struct pages for this, then
your RAM consumption will go up proportionally to fs size. Even
though it is just a small proportion, the RAM can be much more costly
eg in terms of power, so every KB less that Linux uses makes it more
attractive to a lot of these guys.
2. VM_MIXEDMAP allows us to support mappings where you actually do want
to refcount _some_ pages in the mapping, but not others, and support
COW on arbitrary (non-linear) mappings. Jared needs this for his NVRAM
filesystem in progress. Future iterations of this filesystem will
most likely want to migrate pages between pagecache and XIP backing,
which is where the requirement for mixed (some refcounted, some not)
comes from.
3. pte_special also has a peripheral usage that I need for my lockless
get_user_pages patch. That was shown to speed up "oltp" on db2 by
10% on a 2 socket system, which is kind of significant because they
scrounge for months to try to find 0.1% improvement on these
workloads. I'm hoping we might finally be faster than AIX on
pSeries with this :). My reference to lockless get_user_pages is not
meant to justify this patchset (which doesn't include lockless gup),
but just to show that pte_special is not some s390 specific thing that
should be hidden in arch code or xip code: I definitely want to use it
on at least x86 and powerpc as well.
This patch:
Introduce a new type of mapping, VM_MIXEDMAP. This is unlike VM_PFNMAP in
that it can support COW mappings of arbitrary ranges including ranges without
struct page *and* ranges with a struct page that we actually want to refcount
(PFNMAP can only support COW in those cases where the un-COW-ed translations
are mapped linearly in the virtual address, and can only support non
refcounted ranges).
VM_MIXEDMAP achieves this by refcounting all pfn_valid pages, and not
refcounting !pfn_valid pages (which is not an option for VM_PFNMAP, because it
needs to avoid refcounting pfn_valid pages eg. for /dev/mem mappings).
Signed-off-by: Jared Hulbert <jaredeh@gmail.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:12:58 +00:00
|
|
|
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
|
|
|
|
if (vma->vm_flags & VM_MIXEDMAP) {
|
|
|
|
if (!pfn_valid(pfn))
|
|
|
|
return NULL;
|
|
|
|
goto out;
|
|
|
|
} else {
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
unsigned long off;
|
|
|
|
off = (addr - vma->vm_start) >> PAGE_SHIFT;
|
mm: introduce VM_MIXEDMAP
This series introduces some important infrastructure work. The overall result
is that:
1. We now support XIP backed filesystems using memory that have no
struct page allocated to them. And patches 6 and 7 actually implement
this for s390.
This is pretty important in a number of cases. As far as I understand,
in the case of virtualisation (eg. s390), each guest may mount a
readonly copy of the same filesystem (eg. the distro). Currently,
guests need to allocate struct pages for this image. So if you have
100 guests, you already need to allocate more memory for the struct
pages than the size of the image. I think. (Carsten?)
For other (eg. embedded) systems, you may have a very large non-
volatile filesystem. If you have to have struct pages for this, then
your RAM consumption will go up proportionally to fs size. Even
though it is just a small proportion, the RAM can be much more costly
eg in terms of power, so every KB less that Linux uses makes it more
attractive to a lot of these guys.
2. VM_MIXEDMAP allows us to support mappings where you actually do want
to refcount _some_ pages in the mapping, but not others, and support
COW on arbitrary (non-linear) mappings. Jared needs this for his NVRAM
filesystem in progress. Future iterations of this filesystem will
most likely want to migrate pages between pagecache and XIP backing,
which is where the requirement for mixed (some refcounted, some not)
comes from.
3. pte_special also has a peripheral usage that I need for my lockless
get_user_pages patch. That was shown to speed up "oltp" on db2 by
10% on a 2 socket system, which is kind of significant because they
scrounge for months to try to find 0.1% improvement on these
workloads. I'm hoping we might finally be faster than AIX on
pSeries with this :). My reference to lockless get_user_pages is not
meant to justify this patchset (which doesn't include lockless gup),
but just to show that pte_special is not some s390 specific thing that
should be hidden in arch code or xip code: I definitely want to use it
on at least x86 and powerpc as well.
This patch:
Introduce a new type of mapping, VM_MIXEDMAP. This is unlike VM_PFNMAP in
that it can support COW mappings of arbitrary ranges including ranges without
struct page *and* ranges with a struct page that we actually want to refcount
(PFNMAP can only support COW in those cases where the un-COW-ed translations
are mapped linearly in the virtual address, and can only support non
refcounted ranges).
VM_MIXEDMAP achieves this by refcounting all pfn_valid pages, and not
refcounting !pfn_valid pages (which is not an option for VM_PFNMAP, because it
needs to avoid refcounting pfn_valid pages eg. for /dev/mem mappings).
Signed-off-by: Jared Hulbert <jaredeh@gmail.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:12:58 +00:00
|
|
|
if (pfn == vma->vm_pgoff + off)
|
|
|
|
return NULL;
|
|
|
|
if (!is_cow_mapping(vma->vm_flags))
|
|
|
|
return NULL;
|
|
|
|
}
|
2005-11-28 22:34:23 +00:00
|
|
|
}
|
|
|
|
|
x86,mm: fix pte_special versus pte_numa
Sasha Levin has shown oopses on ffffea0003480048 and ffffea0003480008 at
mm/memory.c:1132, running Trinity on different 3.16-rc-next kernels:
where zap_pte_range() checks page->mapping to see if PageAnon(page).
Those addresses fit struct pages for pfns d2001 and d2000, and in each
dump a register or a stack slot showed d2001730 or d2000730: pte flags
0x730 are PCD ACCESSED PROTNONE SPECIAL IOMAP; and Sasha's e820 map has
a hole between cfffffff and 100000000, which would need special access.
Commit c46a7c817e66 ("x86: define _PAGE_NUMA by reusing software bits on
the PMD and PTE levels") has broken vm_normal_page(): a PROTNONE SPECIAL
pte no longer passes the pte_special() test, so zap_pte_range() goes on
to try to access a non-existent struct page.
Fix this by refining pte_special() (SPECIAL with PRESENT or PROTNONE) to
complement pte_numa() (SPECIAL with neither PRESENT nor PROTNONE). A
hint that this was a problem was that c46a7c817e66 added pte_numa() test
to vm_normal_page(), and moved its is_zero_pfn() test from slow to fast
path: This was papering over a pte_special() snag when the zero page was
encountered during zap. This patch reverts vm_normal_page() to how it
was before, relying on pte_special().
It still appears that this patch may be incomplete: aren't there other
places which need to be handling PROTNONE along with PRESENT? For
example, pte_mknuma() clears _PAGE_PRESENT and sets _PAGE_NUMA, but on a
PROT_NONE area, that would make it pte_special(). This is side-stepped
by the fact that NUMA hinting faults skipped PROT_NONE VMAs and there
are no grounds where a NUMA hinting fault on a PROT_NONE VMA would be
interesting.
Fixes: c46a7c817e66 ("x86: define _PAGE_NUMA by reusing software bits on the PMD and PTE levels")
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Tested-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
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>
2014-08-29 22:18:44 +00:00
|
|
|
if (is_zero_pfn(pfn))
|
|
|
|
return NULL;
|
2018-06-08 00:06:12 +00:00
|
|
|
|
2009-01-06 22:40:09 +00:00
|
|
|
check_pfn:
|
|
|
|
if (unlikely(pfn > highest_memmap_pfn)) {
|
|
|
|
print_bad_pte(vma, addr, pte, NULL);
|
|
|
|
return NULL;
|
|
|
|
}
|
2005-11-28 22:34:23 +00:00
|
|
|
|
|
|
|
/*
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
* NOTE! We still have PageReserved() pages in the page tables.
|
|
|
|
* eg. VDSO mappings can cause them to exist.
|
2005-11-28 22:34:23 +00:00
|
|
|
*/
|
mm: introduce VM_MIXEDMAP
This series introduces some important infrastructure work. The overall result
is that:
1. We now support XIP backed filesystems using memory that have no
struct page allocated to them. And patches 6 and 7 actually implement
this for s390.
This is pretty important in a number of cases. As far as I understand,
in the case of virtualisation (eg. s390), each guest may mount a
readonly copy of the same filesystem (eg. the distro). Currently,
guests need to allocate struct pages for this image. So if you have
100 guests, you already need to allocate more memory for the struct
pages than the size of the image. I think. (Carsten?)
For other (eg. embedded) systems, you may have a very large non-
volatile filesystem. If you have to have struct pages for this, then
your RAM consumption will go up proportionally to fs size. Even
though it is just a small proportion, the RAM can be much more costly
eg in terms of power, so every KB less that Linux uses makes it more
attractive to a lot of these guys.
2. VM_MIXEDMAP allows us to support mappings where you actually do want
to refcount _some_ pages in the mapping, but not others, and support
COW on arbitrary (non-linear) mappings. Jared needs this for his NVRAM
filesystem in progress. Future iterations of this filesystem will
most likely want to migrate pages between pagecache and XIP backing,
which is where the requirement for mixed (some refcounted, some not)
comes from.
3. pte_special also has a peripheral usage that I need for my lockless
get_user_pages patch. That was shown to speed up "oltp" on db2 by
10% on a 2 socket system, which is kind of significant because they
scrounge for months to try to find 0.1% improvement on these
workloads. I'm hoping we might finally be faster than AIX on
pSeries with this :). My reference to lockless get_user_pages is not
meant to justify this patchset (which doesn't include lockless gup),
but just to show that pte_special is not some s390 specific thing that
should be hidden in arch code or xip code: I definitely want to use it
on at least x86 and powerpc as well.
This patch:
Introduce a new type of mapping, VM_MIXEDMAP. This is unlike VM_PFNMAP in
that it can support COW mappings of arbitrary ranges including ranges without
struct page *and* ranges with a struct page that we actually want to refcount
(PFNMAP can only support COW in those cases where the un-COW-ed translations
are mapped linearly in the virtual address, and can only support non
refcounted ranges).
VM_MIXEDMAP achieves this by refcounting all pfn_valid pages, and not
refcounting !pfn_valid pages (which is not an option for VM_PFNMAP, because it
needs to avoid refcounting pfn_valid pages eg. for /dev/mem mappings).
Signed-off-by: Jared Hulbert <jaredeh@gmail.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:12:58 +00:00
|
|
|
out:
|
2005-11-28 22:34:23 +00:00
|
|
|
return pfn_to_page(pfn);
|
[PATCH] unpaged: anon in VM_UNPAGED
copy_one_pte needs to copy the anonymous COWed pages in a VM_UNPAGED area,
zap_pte_range needs to free them, do_wp_page needs to COW them: just like
ordinary pages, not like the unpaged.
But recognizing them is a little subtle: because PageReserved is no longer a
condition for remap_pfn_range, we can now mmap all of /dev/mem (whether the
distro permits, and whether it's advisable on this or that architecture, is
another matter). So if we can see a PageAnon, it may not be ours to mess with
(or may be ours from elsewhere in the address space). I suspect there's an
entertaining insoluble self-referential problem here, but the page_is_anon
function does a good practical job, and MAP_PRIVATE PROT_WRITE VM_UNPAGED will
always be an odd choice.
In updating the comment on page_address_in_vma, noticed a potential NULL
dereference, in a path we don't actually take, but fixed it.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-22 05:32:18 +00:00
|
|
|
}
|
|
|
|
|
2022-12-21 18:08:45 +00:00
|
|
|
struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
pte_t pte)
|
|
|
|
{
|
|
|
|
struct page *page = vm_normal_page(vma, addr, pte);
|
|
|
|
|
|
|
|
if (page)
|
|
|
|
return page_folio(page);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2016-04-28 23:18:35 +00:00
|
|
|
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
|
|
|
|
struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
pmd_t pmd)
|
|
|
|
{
|
|
|
|
unsigned long pfn = pmd_pfn(pmd);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* There is no pmd_special() but there may be special pmds, e.g.
|
|
|
|
* in a direct-access (dax) mapping, so let's just replicate the
|
2018-06-08 00:06:12 +00:00
|
|
|
* !CONFIG_ARCH_HAS_PTE_SPECIAL case from vm_normal_page() here.
|
2016-04-28 23:18:35 +00:00
|
|
|
*/
|
|
|
|
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
|
|
|
|
if (vma->vm_flags & VM_MIXEDMAP) {
|
|
|
|
if (!pfn_valid(pfn))
|
|
|
|
return NULL;
|
|
|
|
goto out;
|
|
|
|
} else {
|
|
|
|
unsigned long off;
|
|
|
|
off = (addr - vma->vm_start) >> PAGE_SHIFT;
|
|
|
|
if (pfn == vma->vm_pgoff + off)
|
|
|
|
return NULL;
|
|
|
|
if (!is_cow_mapping(vma->vm_flags))
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-08-17 22:43:40 +00:00
|
|
|
if (pmd_devmap(pmd))
|
|
|
|
return NULL;
|
2019-12-05 00:49:56 +00:00
|
|
|
if (is_huge_zero_pmd(pmd))
|
2016-04-28 23:18:35 +00:00
|
|
|
return NULL;
|
|
|
|
if (unlikely(pfn > highest_memmap_pfn))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* NOTE! We still have PageReserved() pages in the page tables.
|
|
|
|
* eg. VDSO mappings can cause them to exist.
|
|
|
|
*/
|
|
|
|
out:
|
|
|
|
return pfn_to_page(pfn);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2021-07-01 01:54:25 +00:00
|
|
|
static void restore_exclusive_pte(struct vm_area_struct *vma,
|
|
|
|
struct page *page, unsigned long address,
|
|
|
|
pte_t *ptep)
|
|
|
|
{
|
|
|
|
pte_t pte;
|
|
|
|
swp_entry_t entry;
|
|
|
|
|
|
|
|
pte = pte_mkold(mk_pte(page, READ_ONCE(vma->vm_page_prot)));
|
|
|
|
if (pte_swp_soft_dirty(*ptep))
|
|
|
|
pte = pte_mksoft_dirty(pte);
|
|
|
|
|
|
|
|
entry = pte_to_swp_entry(*ptep);
|
|
|
|
if (pte_swp_uffd_wp(*ptep))
|
|
|
|
pte = pte_mkuffd_wp(pte);
|
|
|
|
else if (is_writable_device_exclusive_entry(entry))
|
|
|
|
pte = maybe_mkwrite(pte_mkdirty(pte), vma);
|
|
|
|
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
VM_BUG_ON(pte_write(pte) && !(PageAnon(page) && PageAnonExclusive(page)));
|
|
|
|
|
2021-07-01 01:54:25 +00:00
|
|
|
/*
|
|
|
|
* No need to take a page reference as one was already
|
|
|
|
* created when the swap entry was made.
|
|
|
|
*/
|
|
|
|
if (PageAnon(page))
|
2022-05-10 01:20:43 +00:00
|
|
|
page_add_anon_rmap(page, vma, address, RMAP_NONE);
|
2021-07-01 01:54:25 +00:00
|
|
|
else
|
|
|
|
/*
|
|
|
|
* Currently device exclusive access only supports anonymous
|
|
|
|
* memory so the entry shouldn't point to a filebacked page.
|
|
|
|
*/
|
2022-07-16 08:18:16 +00:00
|
|
|
WARN_ON_ONCE(1);
|
2021-07-01 01:54:25 +00:00
|
|
|
|
2022-01-14 22:06:29 +00:00
|
|
|
set_pte_at(vma->vm_mm, address, ptep, pte);
|
|
|
|
|
2021-07-01 01:54:25 +00:00
|
|
|
/*
|
|
|
|
* No need to invalidate - it was non-present before. However
|
|
|
|
* secondary CPUs may have mappings that need invalidating.
|
|
|
|
*/
|
|
|
|
update_mmu_cache(vma, address, ptep);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Tries to restore an exclusive pte if the page lock can be acquired without
|
|
|
|
* sleeping.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
try_restore_exclusive_pte(pte_t *src_pte, struct vm_area_struct *vma,
|
|
|
|
unsigned long addr)
|
|
|
|
{
|
|
|
|
swp_entry_t entry = pte_to_swp_entry(*src_pte);
|
|
|
|
struct page *page = pfn_swap_entry_to_page(entry);
|
|
|
|
|
|
|
|
if (trylock_page(page)) {
|
|
|
|
restore_exclusive_pte(vma, page, addr, src_pte);
|
|
|
|
unlock_page(page);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return -EBUSY;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* copy one vm_area from one task to the other. Assumes the page tables
|
|
|
|
* already present in the new task to be cleared in the whole range
|
|
|
|
* covered by this vma.
|
|
|
|
*/
|
|
|
|
|
2020-09-23 16:56:59 +00:00
|
|
|
static unsigned long
|
|
|
|
copy_nonpresent_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
|
mm/userfaultfd: fix uffd-wp special cases for fork()
We tried to do something similar in b569a1760782 ("userfaultfd: wp: drop
_PAGE_UFFD_WP properly when fork") previously, but it's not doing it all
right.. A few fixes around the code path:
1. We were referencing VM_UFFD_WP vm_flags on the _old_ vma rather
than the new vma. That's overlooked in b569a1760782, so it won't work
as expected. Thanks to the recent rework on fork code
(7a4830c380f3a8b3), we can easily get the new vma now, so switch the
checks to that.
2. Dropping the uffd-wp bit in copy_huge_pmd() could be wrong if the
huge pmd is a migration huge pmd. When it happens, instead of using
pmd_uffd_wp(), we should use pmd_swp_uffd_wp(). The fix is simply to
handle them separately.
3. Forget to carry over uffd-wp bit for a write migration huge pmd
entry. This also happens in copy_huge_pmd(), where we converted a
write huge migration entry into a read one.
4. In copy_nonpresent_pte(), drop uffd-wp if necessary for swap ptes.
5. In copy_present_page() when COW is enforced when fork(), we also
need to pass over the uffd-wp bit if VM_UFFD_WP is armed on the new
vma, and when the pte to be copied has uffd-wp bit set.
Remove the comment in copy_present_pte() about this. It won't help a huge
lot to only comment there, but comment everywhere would be an overkill.
Let's assume the commit messages would help.
[peterx@redhat.com: fix a few thp pmd missing uffd-wp bit]
Link: https://lkml.kernel.org/r/20210428225030.9708-4-peterx@redhat.com
Link: https://lkml.kernel.org/r/20210428225030.9708-3-peterx@redhat.com
Fixes: b569a1760782f ("userfaultfd: wp: drop _PAGE_UFFD_WP properly when fork")
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Brian Geffon <bgeffon@google.com>
Cc: "Dr . David Alan Gilbert" <dgilbert@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Joe Perches <joe@perches.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Lokesh Gidra <lokeshgidra@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Oliver Upton <oupton@google.com>
Cc: Shaohua Li <shli@fb.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Wang Qing <wangqing@vivo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:49:02 +00:00
|
|
|
pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *dst_vma,
|
|
|
|
struct vm_area_struct *src_vma, unsigned long addr, int *rss)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
mm/userfaultfd: fix uffd-wp special cases for fork()
We tried to do something similar in b569a1760782 ("userfaultfd: wp: drop
_PAGE_UFFD_WP properly when fork") previously, but it's not doing it all
right.. A few fixes around the code path:
1. We were referencing VM_UFFD_WP vm_flags on the _old_ vma rather
than the new vma. That's overlooked in b569a1760782, so it won't work
as expected. Thanks to the recent rework on fork code
(7a4830c380f3a8b3), we can easily get the new vma now, so switch the
checks to that.
2. Dropping the uffd-wp bit in copy_huge_pmd() could be wrong if the
huge pmd is a migration huge pmd. When it happens, instead of using
pmd_uffd_wp(), we should use pmd_swp_uffd_wp(). The fix is simply to
handle them separately.
3. Forget to carry over uffd-wp bit for a write migration huge pmd
entry. This also happens in copy_huge_pmd(), where we converted a
write huge migration entry into a read one.
4. In copy_nonpresent_pte(), drop uffd-wp if necessary for swap ptes.
5. In copy_present_page() when COW is enforced when fork(), we also
need to pass over the uffd-wp bit if VM_UFFD_WP is armed on the new
vma, and when the pte to be copied has uffd-wp bit set.
Remove the comment in copy_present_pte() about this. It won't help a huge
lot to only comment there, but comment everywhere would be an overkill.
Let's assume the commit messages would help.
[peterx@redhat.com: fix a few thp pmd missing uffd-wp bit]
Link: https://lkml.kernel.org/r/20210428225030.9708-4-peterx@redhat.com
Link: https://lkml.kernel.org/r/20210428225030.9708-3-peterx@redhat.com
Fixes: b569a1760782f ("userfaultfd: wp: drop _PAGE_UFFD_WP properly when fork")
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Brian Geffon <bgeffon@google.com>
Cc: "Dr . David Alan Gilbert" <dgilbert@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Joe Perches <joe@perches.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Lokesh Gidra <lokeshgidra@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Oliver Upton <oupton@google.com>
Cc: Shaohua Li <shli@fb.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Wang Qing <wangqing@vivo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:49:02 +00:00
|
|
|
unsigned long vm_flags = dst_vma->vm_flags;
|
2005-04-16 22:20:36 +00:00
|
|
|
pte_t pte = *src_pte;
|
|
|
|
struct page *page;
|
2020-09-23 16:56:59 +00:00
|
|
|
swp_entry_t entry = pte_to_swp_entry(pte);
|
|
|
|
|
|
|
|
if (likely(!non_swap_entry(entry))) {
|
|
|
|
if (swap_duplicate(entry) < 0)
|
2021-07-01 01:54:22 +00:00
|
|
|
return -EIO;
|
2020-09-23 16:56:59 +00:00
|
|
|
|
|
|
|
/* make sure dst_mm is on swapoff's mmlist. */
|
|
|
|
if (unlikely(list_empty(&dst_mm->mmlist))) {
|
|
|
|
spin_lock(&mmlist_lock);
|
|
|
|
if (list_empty(&dst_mm->mmlist))
|
|
|
|
list_add(&dst_mm->mmlist,
|
|
|
|
&src_mm->mmlist);
|
|
|
|
spin_unlock(&mmlist_lock);
|
|
|
|
}
|
mm/swap: remember PG_anon_exclusive via a swp pte bit
Patch series "mm: COW fixes part 3: reliable GUP R/W FOLL_GET of anonymous pages", v2.
This series fixes memory corruptions when a GUP R/W reference (FOLL_WRITE
| FOLL_GET) was taken on an anonymous page and COW logic fails to detect
exclusivity of the page to then replacing the anonymous page by a copy in
the page table: The GUP reference lost synchronicity with the pages mapped
into the page tables. This series focuses on x86, arm64, s390x and
ppc64/book3s -- other architectures are fairly easy to support by
implementing __HAVE_ARCH_PTE_SWP_EXCLUSIVE.
This primarily fixes the O_DIRECT memory corruptions that can happen on
concurrent swapout, whereby we lose DMA reads to a page (modifying the
user page by writing to it).
O_DIRECT currently uses FOLL_GET for short-term (!FOLL_LONGTERM) DMA
from/to a user page. In the long run, we want to convert it to properly
use FOLL_PIN, and John is working on it, but that might take a while and
might not be easy to backport. In the meantime, let's restore what used
to work before we started modifying our COW logic: make R/W FOLL_GET
references reliable as long as there is no fork() after GUP involved.
This is just the natural follow-up of part 2, that will also further
reduce "wrong COW" on the swapin path, for example, when we cannot remove
a page from the swapcache due to concurrent writeback, or if we have two
threads faulting on the same swapped-out page. Fixing O_DIRECT is just a
nice side-product
This issue, including other related COW issues, has been summarized in [3]
under 2):
"
2. Intra Process Memory Corruptions due to Wrong COW (FOLL_GET)
It was discovered that we can create a memory corruption by reading a
file via O_DIRECT to a part (e.g., first 512 bytes) of a page,
concurrently writing to an unrelated part (e.g., last byte) of the same
page, and concurrently write-protecting the page via clear_refs
SOFTDIRTY tracking [6].
For the reproducer, the issue is that O_DIRECT grabs a reference of the
target page (via FOLL_GET) and clear_refs write-protects the relevant
page table entry. On successive write access to the page from the
process itself, we wrongly COW the page when resolving the write fault,
resulting in a loss of synchronicity and consequently a memory corruption.
While some people might think that using clear_refs in this combination
is a corner cases, it turns out to be a more generic problem unfortunately.
For example, it was just recently discovered that we can similarly
create a memory corruption without clear_refs, simply by concurrently
swapping out the buffer pages [7]. Note that we nowadays even use the
swap infrastructure in Linux without an actual swap disk/partition: the
prime example is zram which is enabled as default under Fedora [10].
The root issue is that a write-fault on a page that has additional
references results in a COW and thereby a loss of synchronicity
and consequently a memory corruption if two parties believe they are
referencing the same page.
"
We don't particularly care about R/O FOLL_GET references: they were never
reliable and O_DIRECT doesn't expect to observe modifications from a page
after DMA was started.
Note that:
* this only fixes the issue on x86, arm64, s390x and ppc64/book3s
("enterprise architectures"). Other architectures have to implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE to achieve the same.
* this does *not * consider any kind of fork() after taking the reference:
fork() after GUP never worked reliably with FOLL_GET.
* Not losing PG_anon_exclusive during swapout was the last remaining
piece. KSM already makes sure that there are no other references on
a page before considering it for sharing. Page migration maintains
PG_anon_exclusive and simply fails when there are additional references
(freezing the refcount fails). Only swapout code dropped the
PG_anon_exclusive flag because it requires more work to remember +
restore it.
With this series in place, most COW issues of [3] are fixed on said
architectures. Other architectures can implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE fairly easily.
[1] https://lkml.kernel.org/r/20220329160440.193848-1-david@redhat.com
[2] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[3] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 8):
Currently, we clear PG_anon_exclusive in try_to_unmap() and forget about
it. We do this, to keep fork() logic on swap entries easy and efficient:
for example, if we wouldn't clear it when unmapping, we'd have to lookup
the page in the swapcache for each and every swap entry during fork() and
clear PG_anon_exclusive if set.
Instead, we want to store that information directly in the swap pte,
protected by the page table lock, similarly to how we handle
SWP_MIGRATION_READ_EXCLUSIVE for migration entries. However, for actual
swap entries, we don't want to mess with the swap type (e.g., still one
bit) because it overcomplicates swap code.
In try_to_unmap(), we already reject to unmap in case the page might be
pinned, because we must not lose PG_anon_exclusive on pinned pages ever.
Checking if there are other unexpected references reliably *before*
completely unmapping a page is unfortunately not really possible: THP
heavily overcomplicate the situation. Once fully unmapped it's easier --
we, for example, make sure that there are no unexpected references *after*
unmapping a page before starting writeback on that page.
So, we currently might end up unmapping a page and clearing
PG_anon_exclusive if that page has additional references, for example, due
to a FOLL_GET.
do_swap_page() has to re-determine if a page is exclusive, which will
easily fail if there are other references on a page, most prominently GUP
references via FOLL_GET. This can currently result in memory corruptions
when taking a FOLL_GET | FOLL_WRITE reference on a page even when fork()
is never involved: try_to_unmap() will succeed, and when refaulting the
page, it cannot be marked exclusive and will get replaced by a copy in the
page tables on the next write access, resulting in writes via the GUP
reference to the page being lost.
In an ideal world, everybody that uses GUP and wants to modify page
content, such as O_DIRECT, would properly use FOLL_PIN. However, that
conversion will take a while. It's easier to fix what used to work in the
past (FOLL_GET | FOLL_WRITE) remembering PG_anon_exclusive. In addition,
by remembering PG_anon_exclusive we can further reduce unnecessary COW in
some cases, so it's the natural thing to do.
So let's transfer the PG_anon_exclusive information to the swap pte and
store it via an architecture-dependant pte bit; use that information when
restoring the swap pte in do_swap_page() and unuse_pte(). During fork(),
we simply have to clear the pte bit and are done.
Of course, there is one corner case to handle: swap backends that don't
support concurrent page modifications while the page is under writeback.
Special case these, and drop the exclusive marker. Add a comment why that
is just fine (also, reuse_swap_page() would have done the same in the
past).
In the future, we'll hopefully have all architectures support
__HAVE_ARCH_PTE_SWP_EXCLUSIVE, such that we can get rid of the empty stubs
and the define completely. Then, we can also convert
SWP_MIGRATION_READ_EXCLUSIVE. For architectures it's fairly easy to
support: either simply use a yet unused pte bit that can be used for swap
entries, steal one from the arch type bits if they exceed 5, or steal one
from the offset bits.
Note: R/O FOLL_GET references were never really reliable, especially when
taking one on a shared page and then writing to the page (e.g., GUP after
fork()). FOLL_GET, including R/W references, were never really reliable
once fork was involved (e.g., GUP before fork(), GUP during fork()). KSM
steps back in case it stumbles over unexpected references and is,
therefore, fine.
[david@redhat.com: fix SWP_STABLE_WRITES test]
Link: https://lkml.kernel.org/r/ac725bcb-313a-4fff-250a-68ba9a8f85fb@redhat.comLink: https://lkml.kernel.org/r/20220329164329.208407-1-david@redhat.com
Link: https://lkml.kernel.org/r/20220329164329.208407-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
/* Mark the swap entry as shared. */
|
|
|
|
if (pte_swp_exclusive(*src_pte)) {
|
|
|
|
pte = pte_swp_clear_exclusive(*src_pte);
|
|
|
|
set_pte_at(src_mm, addr, src_pte, pte);
|
|
|
|
}
|
2020-09-23 16:56:59 +00:00
|
|
|
rss[MM_SWAPENTS]++;
|
|
|
|
} else if (is_migration_entry(entry)) {
|
2021-07-01 01:54:06 +00:00
|
|
|
page = pfn_swap_entry_to_page(entry);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2020-09-23 16:56:59 +00:00
|
|
|
rss[mm_counter(page)]++;
|
2017-09-08 23:11:43 +00:00
|
|
|
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
if (!is_readable_migration_entry(entry) &&
|
2020-09-23 16:56:59 +00:00
|
|
|
is_cow_mapping(vm_flags)) {
|
2017-09-08 23:11:43 +00:00
|
|
|
/*
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
* COW mappings require pages in both parent and child
|
|
|
|
* to be set to read. A previously exclusive entry is
|
|
|
|
* now shared.
|
2017-09-08 23:11:43 +00:00
|
|
|
*/
|
2021-07-01 01:54:09 +00:00
|
|
|
entry = make_readable_migration_entry(
|
|
|
|
swp_offset(entry));
|
2020-09-23 16:56:59 +00:00
|
|
|
pte = swp_entry_to_pte(entry);
|
|
|
|
if (pte_swp_soft_dirty(*src_pte))
|
|
|
|
pte = pte_swp_mksoft_dirty(pte);
|
|
|
|
if (pte_swp_uffd_wp(*src_pte))
|
|
|
|
pte = pte_swp_mkuffd_wp(pte);
|
|
|
|
set_pte_at(src_mm, addr, src_pte, pte);
|
|
|
|
}
|
|
|
|
} else if (is_device_private_entry(entry)) {
|
2021-07-01 01:54:06 +00:00
|
|
|
page = pfn_swap_entry_to_page(entry);
|
2017-09-08 23:11:43 +00:00
|
|
|
|
2020-09-23 16:56:59 +00:00
|
|
|
/*
|
|
|
|
* Update rss count even for unaddressable pages, as
|
|
|
|
* they should treated just like normal pages in this
|
|
|
|
* respect.
|
|
|
|
*
|
|
|
|
* We will likely want to have some new rss counters
|
|
|
|
* for unaddressable pages, at some point. But for now
|
|
|
|
* keep things as they are.
|
|
|
|
*/
|
|
|
|
get_page(page);
|
|
|
|
rss[mm_counter(page)]++;
|
2022-05-10 01:20:43 +00:00
|
|
|
/* Cannot fail as these pages cannot get pinned. */
|
|
|
|
BUG_ON(page_try_dup_anon_rmap(page, false, src_vma));
|
2020-09-23 16:56:59 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We do not preserve soft-dirty information, because so
|
|
|
|
* far, checkpoint/restore is the only feature that
|
|
|
|
* requires that. And checkpoint/restore does not work
|
|
|
|
* when a device driver is involved (you cannot easily
|
|
|
|
* save and restore device driver state).
|
|
|
|
*/
|
2021-07-01 01:54:09 +00:00
|
|
|
if (is_writable_device_private_entry(entry) &&
|
2020-09-23 16:56:59 +00:00
|
|
|
is_cow_mapping(vm_flags)) {
|
2021-07-01 01:54:09 +00:00
|
|
|
entry = make_readable_device_private_entry(
|
|
|
|
swp_offset(entry));
|
2020-09-23 16:56:59 +00:00
|
|
|
pte = swp_entry_to_pte(entry);
|
|
|
|
if (pte_swp_uffd_wp(*src_pte))
|
|
|
|
pte = pte_swp_mkuffd_wp(pte);
|
|
|
|
set_pte_at(src_mm, addr, src_pte, pte);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2021-07-01 01:54:25 +00:00
|
|
|
} else if (is_device_exclusive_entry(entry)) {
|
|
|
|
/*
|
|
|
|
* Make device exclusive entries present by restoring the
|
|
|
|
* original entry then copying as for a present pte. Device
|
|
|
|
* exclusive entries currently only support private writable
|
|
|
|
* (ie. COW) mappings.
|
|
|
|
*/
|
|
|
|
VM_BUG_ON(!is_cow_mapping(src_vma->vm_flags));
|
|
|
|
if (try_restore_exclusive_pte(src_pte, src_vma, addr))
|
|
|
|
return -EBUSY;
|
|
|
|
return -ENOENT;
|
2022-05-13 03:22:53 +00:00
|
|
|
} else if (is_pte_marker_entry(entry)) {
|
2022-12-14 20:04:53 +00:00
|
|
|
if (is_swapin_error_entry(entry) || userfaultfd_wp(dst_vma))
|
2022-12-14 20:04:52 +00:00
|
|
|
set_pte_at(dst_mm, addr, dst_pte, pte);
|
2022-05-13 03:22:53 +00:00
|
|
|
return 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
mm/userfaultfd: fix uffd-wp special cases for fork()
We tried to do something similar in b569a1760782 ("userfaultfd: wp: drop
_PAGE_UFFD_WP properly when fork") previously, but it's not doing it all
right.. A few fixes around the code path:
1. We were referencing VM_UFFD_WP vm_flags on the _old_ vma rather
than the new vma. That's overlooked in b569a1760782, so it won't work
as expected. Thanks to the recent rework on fork code
(7a4830c380f3a8b3), we can easily get the new vma now, so switch the
checks to that.
2. Dropping the uffd-wp bit in copy_huge_pmd() could be wrong if the
huge pmd is a migration huge pmd. When it happens, instead of using
pmd_uffd_wp(), we should use pmd_swp_uffd_wp(). The fix is simply to
handle them separately.
3. Forget to carry over uffd-wp bit for a write migration huge pmd
entry. This also happens in copy_huge_pmd(), where we converted a
write huge migration entry into a read one.
4. In copy_nonpresent_pte(), drop uffd-wp if necessary for swap ptes.
5. In copy_present_page() when COW is enforced when fork(), we also
need to pass over the uffd-wp bit if VM_UFFD_WP is armed on the new
vma, and when the pte to be copied has uffd-wp bit set.
Remove the comment in copy_present_pte() about this. It won't help a huge
lot to only comment there, but comment everywhere would be an overkill.
Let's assume the commit messages would help.
[peterx@redhat.com: fix a few thp pmd missing uffd-wp bit]
Link: https://lkml.kernel.org/r/20210428225030.9708-4-peterx@redhat.com
Link: https://lkml.kernel.org/r/20210428225030.9708-3-peterx@redhat.com
Fixes: b569a1760782f ("userfaultfd: wp: drop _PAGE_UFFD_WP properly when fork")
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Brian Geffon <bgeffon@google.com>
Cc: "Dr . David Alan Gilbert" <dgilbert@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Joe Perches <joe@perches.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Lokesh Gidra <lokeshgidra@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Oliver Upton <oupton@google.com>
Cc: Shaohua Li <shli@fb.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Wang Qing <wangqing@vivo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:49:02 +00:00
|
|
|
if (!userfaultfd_wp(dst_vma))
|
|
|
|
pte = pte_swp_clear_uffd_wp(pte);
|
2020-09-23 16:56:59 +00:00
|
|
|
set_pte_at(dst_mm, addr, dst_pte, pte);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2020-09-25 22:25:59 +00:00
|
|
|
/*
|
2022-05-10 01:20:42 +00:00
|
|
|
* Copy a present and normal page.
|
2020-09-25 22:25:59 +00:00
|
|
|
*
|
2022-05-10 01:20:42 +00:00
|
|
|
* NOTE! The usual case is that this isn't required;
|
|
|
|
* instead, the caller can just increase the page refcount
|
|
|
|
* and re-use the pte the traditional way.
|
2020-09-25 22:25:59 +00:00
|
|
|
*
|
|
|
|
* And if we need a pre-allocated page but don't yet have
|
|
|
|
* one, return a negative error to let the preallocation
|
|
|
|
* code know so that it can do so outside the page table
|
|
|
|
* lock.
|
|
|
|
*/
|
|
|
|
static inline int
|
2020-10-13 23:54:21 +00:00
|
|
|
copy_present_page(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
|
|
|
|
pte_t *dst_pte, pte_t *src_pte, unsigned long addr, int *rss,
|
2023-01-16 19:18:12 +00:00
|
|
|
struct folio **prealloc, struct page *page)
|
2020-09-25 22:25:59 +00:00
|
|
|
{
|
2023-01-16 19:18:12 +00:00
|
|
|
struct folio *new_folio;
|
2022-05-10 01:20:42 +00:00
|
|
|
pte_t pte;
|
2020-09-25 22:25:59 +00:00
|
|
|
|
2023-01-16 19:18:12 +00:00
|
|
|
new_folio = *prealloc;
|
|
|
|
if (!new_folio)
|
2020-09-25 22:25:59 +00:00
|
|
|
return -EAGAIN;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We have a prealloc page, all good! Take it
|
|
|
|
* over and copy the page & arm it.
|
|
|
|
*/
|
|
|
|
*prealloc = NULL;
|
2023-01-16 19:18:12 +00:00
|
|
|
copy_user_highpage(&new_folio->page, page, addr, src_vma);
|
|
|
|
__folio_mark_uptodate(new_folio);
|
|
|
|
folio_add_new_anon_rmap(new_folio, dst_vma, addr);
|
|
|
|
folio_add_lru_vma(new_folio, dst_vma);
|
|
|
|
rss[MM_ANONPAGES]++;
|
2020-09-25 22:25:59 +00:00
|
|
|
|
|
|
|
/* All done, just insert the new page copy in the child */
|
2023-01-16 19:18:12 +00:00
|
|
|
pte = mk_pte(&new_folio->page, dst_vma->vm_page_prot);
|
2020-10-13 23:54:21 +00:00
|
|
|
pte = maybe_mkwrite(pte_mkdirty(pte), dst_vma);
|
mm/userfaultfd: fix uffd-wp special cases for fork()
We tried to do something similar in b569a1760782 ("userfaultfd: wp: drop
_PAGE_UFFD_WP properly when fork") previously, but it's not doing it all
right.. A few fixes around the code path:
1. We were referencing VM_UFFD_WP vm_flags on the _old_ vma rather
than the new vma. That's overlooked in b569a1760782, so it won't work
as expected. Thanks to the recent rework on fork code
(7a4830c380f3a8b3), we can easily get the new vma now, so switch the
checks to that.
2. Dropping the uffd-wp bit in copy_huge_pmd() could be wrong if the
huge pmd is a migration huge pmd. When it happens, instead of using
pmd_uffd_wp(), we should use pmd_swp_uffd_wp(). The fix is simply to
handle them separately.
3. Forget to carry over uffd-wp bit for a write migration huge pmd
entry. This also happens in copy_huge_pmd(), where we converted a
write huge migration entry into a read one.
4. In copy_nonpresent_pte(), drop uffd-wp if necessary for swap ptes.
5. In copy_present_page() when COW is enforced when fork(), we also
need to pass over the uffd-wp bit if VM_UFFD_WP is armed on the new
vma, and when the pte to be copied has uffd-wp bit set.
Remove the comment in copy_present_pte() about this. It won't help a huge
lot to only comment there, but comment everywhere would be an overkill.
Let's assume the commit messages would help.
[peterx@redhat.com: fix a few thp pmd missing uffd-wp bit]
Link: https://lkml.kernel.org/r/20210428225030.9708-4-peterx@redhat.com
Link: https://lkml.kernel.org/r/20210428225030.9708-3-peterx@redhat.com
Fixes: b569a1760782f ("userfaultfd: wp: drop _PAGE_UFFD_WP properly when fork")
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Brian Geffon <bgeffon@google.com>
Cc: "Dr . David Alan Gilbert" <dgilbert@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Joe Perches <joe@perches.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Lokesh Gidra <lokeshgidra@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Oliver Upton <oupton@google.com>
Cc: Shaohua Li <shli@fb.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Wang Qing <wangqing@vivo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:49:02 +00:00
|
|
|
if (userfaultfd_pte_wp(dst_vma, *src_pte))
|
|
|
|
/* Uffd-wp needs to be delivered to dest pte as well */
|
2022-12-14 20:15:33 +00:00
|
|
|
pte = pte_mkuffd_wp(pte);
|
2020-10-13 23:54:21 +00:00
|
|
|
set_pte_at(dst_vma->vm_mm, addr, dst_pte, pte);
|
2020-09-25 22:25:59 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Copy one pte. Returns 0 if succeeded, or -EAGAIN if one preallocated page
|
|
|
|
* is required to copy this pte.
|
|
|
|
*/
|
|
|
|
static inline int
|
2020-10-13 23:54:21 +00:00
|
|
|
copy_present_pte(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
|
|
|
|
pte_t *dst_pte, pte_t *src_pte, unsigned long addr, int *rss,
|
2023-01-16 19:18:12 +00:00
|
|
|
struct folio **prealloc)
|
2020-09-23 16:56:59 +00:00
|
|
|
{
|
2020-10-13 23:54:21 +00:00
|
|
|
struct mm_struct *src_mm = src_vma->vm_mm;
|
|
|
|
unsigned long vm_flags = src_vma->vm_flags;
|
2020-09-23 16:56:59 +00:00
|
|
|
pte_t pte = *src_pte;
|
|
|
|
struct page *page;
|
2023-01-16 19:18:13 +00:00
|
|
|
struct folio *folio;
|
2020-09-23 16:56:59 +00:00
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
page = vm_normal_page(src_vma, addr, pte);
|
2023-01-16 19:18:13 +00:00
|
|
|
if (page)
|
|
|
|
folio = page_folio(page);
|
|
|
|
if (page && folio_test_anon(folio)) {
|
2022-05-10 01:20:42 +00:00
|
|
|
/*
|
|
|
|
* If this page may have been pinned by the parent process,
|
|
|
|
* copy the page immediately for the child so that we'll always
|
|
|
|
* guarantee the pinned page won't be randomly replaced in the
|
|
|
|
* future.
|
|
|
|
*/
|
2023-01-16 19:18:13 +00:00
|
|
|
folio_get(folio);
|
2022-05-10 01:20:43 +00:00
|
|
|
if (unlikely(page_try_dup_anon_rmap(page, false, src_vma))) {
|
2023-01-16 19:18:13 +00:00
|
|
|
/* Page may be pinned, we have to copy. */
|
|
|
|
folio_put(folio);
|
2022-05-10 01:20:43 +00:00
|
|
|
return copy_present_page(dst_vma, src_vma, dst_pte, src_pte,
|
|
|
|
addr, rss, prealloc, page);
|
|
|
|
}
|
2023-01-16 19:18:12 +00:00
|
|
|
rss[MM_ANONPAGES]++;
|
2022-05-10 01:20:42 +00:00
|
|
|
} else if (page) {
|
2023-01-16 19:18:13 +00:00
|
|
|
folio_get(folio);
|
2022-05-10 01:20:43 +00:00
|
|
|
page_dup_file_rmap(page, false);
|
2023-01-16 19:18:12 +00:00
|
|
|
rss[mm_counter_file(page)]++;
|
2020-09-25 22:25:59 +00:00
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* If it's a COW mapping, write protect it both
|
|
|
|
* in the parent and the child
|
|
|
|
*/
|
2018-07-09 20:19:49 +00:00
|
|
|
if (is_cow_mapping(vm_flags) && pte_write(pte)) {
|
2005-04-16 22:20:36 +00:00
|
|
|
ptep_set_wrprotect(src_mm, addr, src_pte);
|
2006-10-01 06:29:30 +00:00
|
|
|
pte = pte_wrprotect(pte);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2023-01-16 19:18:13 +00:00
|
|
|
VM_BUG_ON(page && folio_test_anon(folio) && PageAnonExclusive(page));
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If it's a shared mapping, mark it clean in
|
|
|
|
* the child
|
|
|
|
*/
|
|
|
|
if (vm_flags & VM_SHARED)
|
|
|
|
pte = pte_mkclean(pte);
|
|
|
|
pte = pte_mkold(pte);
|
2005-11-28 22:34:23 +00:00
|
|
|
|
mm/userfaultfd: fix uffd-wp special cases for fork()
We tried to do something similar in b569a1760782 ("userfaultfd: wp: drop
_PAGE_UFFD_WP properly when fork") previously, but it's not doing it all
right.. A few fixes around the code path:
1. We were referencing VM_UFFD_WP vm_flags on the _old_ vma rather
than the new vma. That's overlooked in b569a1760782, so it won't work
as expected. Thanks to the recent rework on fork code
(7a4830c380f3a8b3), we can easily get the new vma now, so switch the
checks to that.
2. Dropping the uffd-wp bit in copy_huge_pmd() could be wrong if the
huge pmd is a migration huge pmd. When it happens, instead of using
pmd_uffd_wp(), we should use pmd_swp_uffd_wp(). The fix is simply to
handle them separately.
3. Forget to carry over uffd-wp bit for a write migration huge pmd
entry. This also happens in copy_huge_pmd(), where we converted a
write huge migration entry into a read one.
4. In copy_nonpresent_pte(), drop uffd-wp if necessary for swap ptes.
5. In copy_present_page() when COW is enforced when fork(), we also
need to pass over the uffd-wp bit if VM_UFFD_WP is armed on the new
vma, and when the pte to be copied has uffd-wp bit set.
Remove the comment in copy_present_pte() about this. It won't help a huge
lot to only comment there, but comment everywhere would be an overkill.
Let's assume the commit messages would help.
[peterx@redhat.com: fix a few thp pmd missing uffd-wp bit]
Link: https://lkml.kernel.org/r/20210428225030.9708-4-peterx@redhat.com
Link: https://lkml.kernel.org/r/20210428225030.9708-3-peterx@redhat.com
Fixes: b569a1760782f ("userfaultfd: wp: drop _PAGE_UFFD_WP properly when fork")
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Brian Geffon <bgeffon@google.com>
Cc: "Dr . David Alan Gilbert" <dgilbert@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Joe Perches <joe@perches.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Lokesh Gidra <lokeshgidra@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Oliver Upton <oupton@google.com>
Cc: Shaohua Li <shli@fb.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Wang Qing <wangqing@vivo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:49:02 +00:00
|
|
|
if (!userfaultfd_wp(dst_vma))
|
2020-04-07 03:05:53 +00:00
|
|
|
pte = pte_clear_uffd_wp(pte);
|
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
set_pte_at(dst_vma->vm_mm, addr, dst_pte, pte);
|
2020-09-25 22:25:59 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2023-01-16 19:18:12 +00:00
|
|
|
static inline struct folio *page_copy_prealloc(struct mm_struct *src_mm,
|
|
|
|
struct vm_area_struct *vma, unsigned long addr)
|
2020-09-25 22:25:59 +00:00
|
|
|
{
|
2023-01-16 19:18:12 +00:00
|
|
|
struct folio *new_folio;
|
2020-09-25 22:25:59 +00:00
|
|
|
|
2023-01-16 19:18:12 +00:00
|
|
|
new_folio = vma_alloc_folio(GFP_HIGHUSER_MOVABLE, 0, vma, addr, false);
|
|
|
|
if (!new_folio)
|
2020-09-25 22:25:59 +00:00
|
|
|
return NULL;
|
|
|
|
|
2023-01-16 19:18:12 +00:00
|
|
|
if (mem_cgroup_charge(new_folio, src_mm, GFP_KERNEL)) {
|
|
|
|
folio_put(new_folio);
|
2020-09-25 22:25:59 +00:00
|
|
|
return NULL;
|
2005-11-28 22:34:23 +00:00
|
|
|
}
|
2023-03-02 11:58:31 +00:00
|
|
|
folio_throttle_swaprate(new_folio, GFP_KERNEL);
|
2005-10-30 01:16:05 +00:00
|
|
|
|
2023-01-16 19:18:12 +00:00
|
|
|
return new_folio;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
static int
|
|
|
|
copy_pte_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
|
|
|
|
pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr,
|
|
|
|
unsigned long end)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2020-10-13 23:54:21 +00:00
|
|
|
struct mm_struct *dst_mm = dst_vma->vm_mm;
|
|
|
|
struct mm_struct *src_mm = src_vma->vm_mm;
|
2009-10-26 23:50:23 +00:00
|
|
|
pte_t *orig_src_pte, *orig_dst_pte;
|
2005-04-16 22:20:36 +00:00
|
|
|
pte_t *src_pte, *dst_pte;
|
2005-10-30 01:16:23 +00:00
|
|
|
spinlock_t *src_ptl, *dst_ptl;
|
2020-09-25 22:25:59 +00:00
|
|
|
int progress, ret = 0;
|
2010-03-05 21:41:39 +00:00
|
|
|
int rss[NR_MM_COUNTERS];
|
swap_info: swap count continuations
Swap is duplicated (reference count incremented by one) whenever the same
swap page is inserted into another mm (when forking finds a swap entry in
place of a pte, or when reclaim unmaps a pte to insert the swap entry).
swap_info_struct's vmalloc'ed swap_map is the array of these reference
counts: but what happens when the unsigned short (or unsigned char since
the preceding patch) is full? (and its high bit is kept for a cache flag)
We then lose track of it, never freeing, leaving it in use until swapoff:
at which point we _hope_ that a single pass will have found all instances,
assume there are no more, and will lose user data if we're wrong.
Swapping of KSM pages has not yet been enabled; but it is implemented,
and makes it very easy for a user to overflow the maximum swap count:
possible with ordinary process pages, but unlikely, even when pid_max
has been raised from PID_MAX_DEFAULT.
This patch implements swap count continuations: when the count overflows,
a continuation page is allocated and linked to the original vmalloc'ed
map page, and this used to hold the continuation counts for that entry
and its neighbours. These continuation pages are seldom referenced:
the common paths all work on the original swap_map, only referring to
a continuation page when the low "digit" of a count is incremented or
decremented through SWAP_MAP_MAX.
Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.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>
2009-12-15 01:58:46 +00:00
|
|
|
swp_entry_t entry = (swp_entry_t){0};
|
2023-01-16 19:18:12 +00:00
|
|
|
struct folio *prealloc = NULL;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
again:
|
2020-09-25 22:25:59 +00:00
|
|
|
progress = 0;
|
2010-03-05 21:41:39 +00:00
|
|
|
init_rss_vec(rss);
|
|
|
|
|
2005-10-30 01:16:23 +00:00
|
|
|
dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
|
2020-09-25 22:25:59 +00:00
|
|
|
if (!dst_pte) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2010-10-26 21:21:52 +00:00
|
|
|
src_pte = pte_offset_map(src_pmd, addr);
|
[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
|
|
|
src_ptl = pte_lockptr(src_mm, src_pmd);
|
2006-07-03 07:25:08 +00:00
|
|
|
spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
|
2009-10-26 23:50:23 +00:00
|
|
|
orig_src_pte = src_pte;
|
|
|
|
orig_dst_pte = dst_pte;
|
2006-10-01 06:29:33 +00:00
|
|
|
arch_enter_lazy_mmu_mode();
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
do {
|
|
|
|
/*
|
|
|
|
* We are holding two locks at this point - either of them
|
|
|
|
* could generate latencies in another task on another CPU.
|
|
|
|
*/
|
2005-10-30 01:15:53 +00:00
|
|
|
if (progress >= 32) {
|
|
|
|
progress = 0;
|
|
|
|
if (need_resched() ||
|
2008-01-30 12:31:20 +00:00
|
|
|
spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
|
2005-10-30 01:15:53 +00:00
|
|
|
break;
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
if (pte_none(*src_pte)) {
|
|
|
|
progress++;
|
|
|
|
continue;
|
|
|
|
}
|
2020-09-23 17:04:16 +00:00
|
|
|
if (unlikely(!pte_present(*src_pte))) {
|
2021-07-01 01:54:22 +00:00
|
|
|
ret = copy_nonpresent_pte(dst_mm, src_mm,
|
|
|
|
dst_pte, src_pte,
|
|
|
|
dst_vma, src_vma,
|
|
|
|
addr, rss);
|
|
|
|
if (ret == -EIO) {
|
|
|
|
entry = pte_to_swp_entry(*src_pte);
|
2020-09-23 17:04:16 +00:00
|
|
|
break;
|
2021-07-01 01:54:25 +00:00
|
|
|
} else if (ret == -EBUSY) {
|
|
|
|
break;
|
|
|
|
} else if (!ret) {
|
|
|
|
progress += 8;
|
|
|
|
continue;
|
2021-07-01 01:54:22 +00:00
|
|
|
}
|
2021-07-01 01:54:25 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Device exclusive entry restored, continue by copying
|
|
|
|
* the now present pte.
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(ret != -ENOENT);
|
2020-09-23 17:04:16 +00:00
|
|
|
}
|
2020-09-25 22:25:59 +00:00
|
|
|
/* copy_present_pte() will clear `*prealloc' if consumed */
|
2020-10-13 23:54:21 +00:00
|
|
|
ret = copy_present_pte(dst_vma, src_vma, dst_pte, src_pte,
|
|
|
|
addr, rss, &prealloc);
|
2020-09-25 22:25:59 +00:00
|
|
|
/*
|
|
|
|
* If we need a pre-allocated page for this pte, drop the
|
|
|
|
* locks, allocate, and try again.
|
|
|
|
*/
|
|
|
|
if (unlikely(ret == -EAGAIN))
|
|
|
|
break;
|
|
|
|
if (unlikely(prealloc)) {
|
|
|
|
/*
|
|
|
|
* pre-alloc page cannot be reused by next time so as
|
|
|
|
* to strictly follow mempolicy (e.g., alloc_page_vma()
|
|
|
|
* will allocate page according to address). This
|
|
|
|
* could only happen if one pinned pte changed.
|
|
|
|
*/
|
2023-01-16 19:18:12 +00:00
|
|
|
folio_put(prealloc);
|
2020-09-25 22:25:59 +00:00
|
|
|
prealloc = NULL;
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
progress += 8;
|
|
|
|
} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);
|
|
|
|
|
2006-10-01 06:29:33 +00:00
|
|
|
arch_leave_lazy_mmu_mode();
|
2005-10-30 01:16:23 +00:00
|
|
|
spin_unlock(src_ptl);
|
2010-10-26 21:21:52 +00:00
|
|
|
pte_unmap(orig_src_pte);
|
2010-03-05 21:41:39 +00:00
|
|
|
add_mm_rss_vec(dst_mm, rss);
|
2009-10-26 23:50:23 +00:00
|
|
|
pte_unmap_unlock(orig_dst_pte, dst_ptl);
|
2005-10-30 01:16:23 +00:00
|
|
|
cond_resched();
|
swap_info: swap count continuations
Swap is duplicated (reference count incremented by one) whenever the same
swap page is inserted into another mm (when forking finds a swap entry in
place of a pte, or when reclaim unmaps a pte to insert the swap entry).
swap_info_struct's vmalloc'ed swap_map is the array of these reference
counts: but what happens when the unsigned short (or unsigned char since
the preceding patch) is full? (and its high bit is kept for a cache flag)
We then lose track of it, never freeing, leaving it in use until swapoff:
at which point we _hope_ that a single pass will have found all instances,
assume there are no more, and will lose user data if we're wrong.
Swapping of KSM pages has not yet been enabled; but it is implemented,
and makes it very easy for a user to overflow the maximum swap count:
possible with ordinary process pages, but unlikely, even when pid_max
has been raised from PID_MAX_DEFAULT.
This patch implements swap count continuations: when the count overflows,
a continuation page is allocated and linked to the original vmalloc'ed
map page, and this used to hold the continuation counts for that entry
and its neighbours. These continuation pages are seldom referenced:
the common paths all work on the original swap_map, only referring to
a continuation page when the low "digit" of a count is incremented or
decremented through SWAP_MAP_MAX.
Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.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>
2009-12-15 01:58:46 +00:00
|
|
|
|
2021-07-01 01:54:22 +00:00
|
|
|
if (ret == -EIO) {
|
|
|
|
VM_WARN_ON_ONCE(!entry.val);
|
2020-09-25 22:25:59 +00:00
|
|
|
if (add_swap_count_continuation(entry, GFP_KERNEL) < 0) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
entry.val = 0;
|
2021-07-01 01:54:25 +00:00
|
|
|
} else if (ret == -EBUSY) {
|
|
|
|
goto out;
|
2021-07-01 01:54:22 +00:00
|
|
|
} else if (ret == -EAGAIN) {
|
2020-10-13 23:54:21 +00:00
|
|
|
prealloc = page_copy_prealloc(src_mm, src_vma, addr);
|
2020-09-25 22:25:59 +00:00
|
|
|
if (!prealloc)
|
swap_info: swap count continuations
Swap is duplicated (reference count incremented by one) whenever the same
swap page is inserted into another mm (when forking finds a swap entry in
place of a pte, or when reclaim unmaps a pte to insert the swap entry).
swap_info_struct's vmalloc'ed swap_map is the array of these reference
counts: but what happens when the unsigned short (or unsigned char since
the preceding patch) is full? (and its high bit is kept for a cache flag)
We then lose track of it, never freeing, leaving it in use until swapoff:
at which point we _hope_ that a single pass will have found all instances,
assume there are no more, and will lose user data if we're wrong.
Swapping of KSM pages has not yet been enabled; but it is implemented,
and makes it very easy for a user to overflow the maximum swap count:
possible with ordinary process pages, but unlikely, even when pid_max
has been raised from PID_MAX_DEFAULT.
This patch implements swap count continuations: when the count overflows,
a continuation page is allocated and linked to the original vmalloc'ed
map page, and this used to hold the continuation counts for that entry
and its neighbours. These continuation pages are seldom referenced:
the common paths all work on the original swap_map, only referring to
a continuation page when the low "digit" of a count is incremented or
decremented through SWAP_MAP_MAX.
Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.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>
2009-12-15 01:58:46 +00:00
|
|
|
return -ENOMEM;
|
2021-07-01 01:54:22 +00:00
|
|
|
} else if (ret) {
|
|
|
|
VM_WARN_ON_ONCE(1);
|
swap_info: swap count continuations
Swap is duplicated (reference count incremented by one) whenever the same
swap page is inserted into another mm (when forking finds a swap entry in
place of a pte, or when reclaim unmaps a pte to insert the swap entry).
swap_info_struct's vmalloc'ed swap_map is the array of these reference
counts: but what happens when the unsigned short (or unsigned char since
the preceding patch) is full? (and its high bit is kept for a cache flag)
We then lose track of it, never freeing, leaving it in use until swapoff:
at which point we _hope_ that a single pass will have found all instances,
assume there are no more, and will lose user data if we're wrong.
Swapping of KSM pages has not yet been enabled; but it is implemented,
and makes it very easy for a user to overflow the maximum swap count:
possible with ordinary process pages, but unlikely, even when pid_max
has been raised from PID_MAX_DEFAULT.
This patch implements swap count continuations: when the count overflows,
a continuation page is allocated and linked to the original vmalloc'ed
map page, and this used to hold the continuation counts for that entry
and its neighbours. These continuation pages are seldom referenced:
the common paths all work on the original swap_map, only referring to
a continuation page when the low "digit" of a count is incremented or
decremented through SWAP_MAP_MAX.
Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.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>
2009-12-15 01:58:46 +00:00
|
|
|
}
|
2021-07-01 01:54:22 +00:00
|
|
|
|
|
|
|
/* We've captured and resolved the error. Reset, try again. */
|
|
|
|
ret = 0;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (addr != end)
|
|
|
|
goto again;
|
2020-09-25 22:25:59 +00:00
|
|
|
out:
|
|
|
|
if (unlikely(prealloc))
|
2023-01-16 19:18:12 +00:00
|
|
|
folio_put(prealloc);
|
2020-09-25 22:25:59 +00:00
|
|
|
return ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
static inline int
|
|
|
|
copy_pmd_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
|
|
|
|
pud_t *dst_pud, pud_t *src_pud, unsigned long addr,
|
|
|
|
unsigned long end)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2020-10-13 23:54:21 +00:00
|
|
|
struct mm_struct *dst_mm = dst_vma->vm_mm;
|
|
|
|
struct mm_struct *src_mm = src_vma->vm_mm;
|
2005-04-16 22:20:36 +00:00
|
|
|
pmd_t *src_pmd, *dst_pmd;
|
|
|
|
unsigned long next;
|
|
|
|
|
|
|
|
dst_pmd = pmd_alloc(dst_mm, dst_pud, addr);
|
|
|
|
if (!dst_pmd)
|
|
|
|
return -ENOMEM;
|
|
|
|
src_pmd = pmd_offset(src_pud, addr);
|
|
|
|
do {
|
|
|
|
next = pmd_addr_end(addr, end);
|
mm: thp: check pmd migration entry in common path
When THP migration is being used, memory management code needs to handle
pmd migration entries properly. This patch uses !pmd_present() or
is_swap_pmd() (depending on whether pmd_none() needs separate code or
not) to check pmd migration entries at the places where a pmd entry is
present.
Since pmd-related code uses split_huge_page(), split_huge_pmd(),
pmd_trans_huge(), pmd_trans_unstable(), or
pmd_none_or_trans_huge_or_clear_bad(), this patch:
1. adds pmd migration entry split code in split_huge_pmd(),
2. takes care of pmd migration entries whenever pmd_trans_huge() is present,
3. makes pmd_none_or_trans_huge_or_clear_bad() pmd migration entry aware.
Since split_huge_page() uses split_huge_pmd() and pmd_trans_unstable()
is equivalent to pmd_none_or_trans_huge_or_clear_bad(), we do not change
them.
Until this commit, a pmd entry should be:
1. pointing to a pte page,
2. is_swap_pmd(),
3. pmd_trans_huge(),
4. pmd_devmap(), or
5. pmd_none().
Signed-off-by: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Nellans <dnellans@nvidia.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:11:01 +00:00
|
|
|
if (is_swap_pmd(*src_pmd) || pmd_trans_huge(*src_pmd)
|
|
|
|
|| pmd_devmap(*src_pmd)) {
|
thp: transparent hugepage core
Lately I've been working to make KVM use hugepages transparently without
the usual restrictions of hugetlbfs. Some of the restrictions I'd like to
see removed:
1) hugepages have to be swappable or the guest physical memory remains
locked in RAM and can't be paged out to swap
2) if a hugepage allocation fails, regular pages should be allocated
instead and mixed in the same vma without any failure and without
userland noticing
3) if some task quits and more hugepages become available in the
buddy, guest physical memory backed by regular pages should be
relocated on hugepages automatically in regions under
madvise(MADV_HUGEPAGE) (ideally event driven by waking up the
kernel deamon if the order=HPAGE_PMD_SHIFT-PAGE_SHIFT list becomes
not null)
4) avoidance of reservation and maximization of use of hugepages whenever
possible. Reservation (needed to avoid runtime fatal faliures) may be ok for
1 machine with 1 database with 1 database cache with 1 database cache size
known at boot time. It's definitely not feasible with a virtualization
hypervisor usage like RHEV-H that runs an unknown number of virtual machines
with an unknown size of each virtual machine with an unknown amount of
pagecache that could be potentially useful in the host for guest not using
O_DIRECT (aka cache=off).
hugepages in the virtualization hypervisor (and also in the guest!) are
much more important than in a regular host not using virtualization,
becasue with NPT/EPT they decrease the tlb-miss cacheline accesses from 24
to 19 in case only the hypervisor uses transparent hugepages, and they
decrease the tlb-miss cacheline accesses from 19 to 15 in case both the
linux hypervisor and the linux guest both uses this patch (though the
guest will limit the addition speedup to anonymous regions only for
now...). Even more important is that the tlb miss handler is much slower
on a NPT/EPT guest than for a regular shadow paging or no-virtualization
scenario. So maximizing the amount of virtual memory cached by the TLB
pays off significantly more with NPT/EPT than without (even if there would
be no significant speedup in the tlb-miss runtime).
The first (and more tedious) part of this work requires allowing the VM to
handle anonymous hugepages mixed with regular pages transparently on
regular anonymous vmas. This is what this patch tries to achieve in the
least intrusive possible way. We want hugepages and hugetlb to be used in
a way so that all applications can benefit without changes (as usual we
leverage the KVM virtualization design: by improving the Linux VM at
large, KVM gets the performance boost too).
The most important design choice is: always fallback to 4k allocation if
the hugepage allocation fails! This is the _very_ opposite of some large
pagecache patches that failed with -EIO back then if a 64k (or similar)
allocation failed...
Second important decision (to reduce the impact of the feature on the
existing pagetable handling code) is that at any time we can split an
hugepage into 512 regular pages and it has to be done with an operation
that can't fail. This way the reliability of the swapping isn't decreased
(no need to allocate memory when we are short on memory to swap) and it's
trivial to plug a split_huge_page* one-liner where needed without
polluting the VM. Over time we can teach mprotect, mremap and friends to
handle pmd_trans_huge natively without calling split_huge_page*. The fact
it can't fail isn't just for swap: if split_huge_page would return -ENOMEM
(instead of the current void) we'd need to rollback the mprotect from the
middle of it (ideally including undoing the split_vma) which would be a
big change and in the very wrong direction (it'd likely be simpler not to
call split_huge_page at all and to teach mprotect and friends to handle
hugepages instead of rolling them back from the middle). In short the
very value of split_huge_page is that it can't fail.
The collapsing and madvise(MADV_HUGEPAGE) part will remain separated and
incremental and it'll just be an "harmless" addition later if this initial
part is agreed upon. It also should be noted that locking-wise replacing
regular pages with hugepages is going to be very easy if compared to what
I'm doing below in split_huge_page, as it will only happen when
page_count(page) matches page_mapcount(page) if we can take the PG_lock
and mmap_sem in write mode. collapse_huge_page will be a "best effort"
that (unlike split_huge_page) can fail at the minimal sign of trouble and
we can try again later. collapse_huge_page will be similar to how KSM
works and the madvise(MADV_HUGEPAGE) will work similar to
madvise(MADV_MERGEABLE).
The default I like is that transparent hugepages are used at page fault
time. This can be changed with
/sys/kernel/mm/transparent_hugepage/enabled. The control knob can be set
to three values "always", "madvise", "never" which mean respectively that
hugepages are always used, or only inside madvise(MADV_HUGEPAGE) regions,
or never used. /sys/kernel/mm/transparent_hugepage/defrag instead
controls if the hugepage allocation should defrag memory aggressively
"always", only inside "madvise" regions, or "never".
The pmd_trans_splitting/pmd_trans_huge locking is very solid. The
put_page (from get_user_page users that can't use mmu notifier like
O_DIRECT) that runs against a __split_huge_page_refcount instead was a
pain to serialize in a way that would result always in a coherent page
count for both tail and head. I think my locking solution with a
compound_lock taken only after the page_first is valid and is still a
PageHead should be safe but it surely needs review from SMP race point of
view. In short there is no current existing way to serialize the O_DIRECT
final put_page against split_huge_page_refcount so I had to invent a new
one (O_DIRECT loses knowledge on the mapping status by the time gup_fast
returns so...). And I didn't want to impact all gup/gup_fast users for
now, maybe if we change the gup interface substantially we can avoid this
locking, I admit I didn't think too much about it because changing the gup
unpinning interface would be invasive.
If we ignored O_DIRECT we could stick to the existing compound refcounting
code, by simply adding a get_user_pages_fast_flags(foll_flags) where KVM
(and any other mmu notifier user) would call it without FOLL_GET (and if
FOLL_GET isn't set we'd just BUG_ON if nobody registered itself in the
current task mmu notifier list yet). But O_DIRECT is fundamental for
decent performance of virtualized I/O on fast storage so we can't avoid it
to solve the race of put_page against split_huge_page_refcount to achieve
a complete hugepage feature for KVM.
Swap and oom works fine (well just like with regular pages ;). MMU
notifier is handled transparently too, with the exception of the young bit
on the pmd, that didn't have a range check but I think KVM will be fine
because the whole point of hugepages is that EPT/NPT will also use a huge
pmd when they notice gup returns pages with PageCompound set, so they
won't care of a range and there's just the pmd young bit to check in that
case.
NOTE: in some cases if the L2 cache is small, this may slowdown and waste
memory during COWs because 4M of memory are accessed in a single fault
instead of 8k (the payoff is that after COW the program can run faster).
So we might want to switch the copy_huge_page (and clear_huge_page too) to
not temporal stores. I also extensively researched ways to avoid this
cache trashing with a full prefault logic that would cow in 8k/16k/32k/64k
up to 1M (I can send those patches that fully implemented prefault) but I
concluded they're not worth it and they add an huge additional complexity
and they remove all tlb benefits until the full hugepage has been faulted
in, to save a little bit of memory and some cache during app startup, but
they still don't improve substantially the cache-trashing during startup
if the prefault happens in >4k chunks. One reason is that those 4k pte
entries copied are still mapped on a perfectly cache-colored hugepage, so
the trashing is the worst one can generate in those copies (cow of 4k page
copies aren't so well colored so they trashes less, but again this results
in software running faster after the page fault). Those prefault patches
allowed things like a pte where post-cow pages were local 4k regular anon
pages and the not-yet-cowed pte entries were pointing in the middle of
some hugepage mapped read-only. If it doesn't payoff substantially with
todays hardware it will payoff even less in the future with larger l2
caches, and the prefault logic would blot the VM a lot. If one is
emebdded transparent_hugepage can be disabled during boot with sysfs or
with the boot commandline parameter transparent_hugepage=0 (or
transparent_hugepage=2 to restrict hugepages inside madvise regions) that
will ensure not a single hugepage is allocated at boot time. It is simple
enough to just disable transparent hugepage globally and let transparent
hugepages be allocated selectively by applications in the MADV_HUGEPAGE
region (both at page fault time, and if enabled with the
collapse_huge_page too through the kernel daemon).
This patch supports only hugepages mapped in the pmd, archs that have
smaller hugepages will not fit in this patch alone. Also some archs like
power have certain tlb limits that prevents mixing different page size in
the same regions so they will not fit in this framework that requires
"graceful fallback" to basic PAGE_SIZE in case of physical memory
fragmentation. hugetlbfs remains a perfect fit for those because its
software limits happen to match the hardware limits. hugetlbfs also
remains a perfect fit for hugepage sizes like 1GByte that cannot be hoped
to be found not fragmented after a certain system uptime and that would be
very expensive to defragment with relocation, so requiring reservation.
hugetlbfs is the "reservation way", the point of transparent hugepages is
not to have any reservation at all and maximizing the use of cache and
hugepages at all times automatically.
Some performance result:
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largep
ages3
memset page fault 1566023
memset tlb miss 453854
memset second tlb miss 453321
random access tlb miss 41635
random access second tlb miss 41658
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largepages3
memset page fault 1566471
memset tlb miss 453375
memset second tlb miss 453320
random access tlb miss 41636
random access second tlb miss 41637
vmx andrea # ./largepages3
memset page fault 1566642
memset tlb miss 453417
memset second tlb miss 453313
random access tlb miss 41630
random access second tlb miss 41647
vmx andrea # ./largepages3
memset page fault 1566872
memset tlb miss 453418
memset second tlb miss 453315
random access tlb miss 41618
random access second tlb miss 41659
vmx andrea # echo 0 > /proc/sys/vm/transparent_hugepage
vmx andrea # ./largepages3
memset page fault 2182476
memset tlb miss 460305
memset second tlb miss 460179
random access tlb miss 44483
random access second tlb miss 44186
vmx andrea # ./largepages3
memset page fault 2182791
memset tlb miss 460742
memset second tlb miss 459962
random access tlb miss 43981
random access second tlb miss 43988
============
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#define SIZE (3UL*1024*1024*1024)
int main()
{
char *p = malloc(SIZE), *p2;
struct timeval before, after;
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset page fault %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
return 0;
}
============
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:46:52 +00:00
|
|
|
int err;
|
2020-10-13 23:54:21 +00:00
|
|
|
VM_BUG_ON_VMA(next-addr != HPAGE_PMD_SIZE, src_vma);
|
mm/userfaultfd: fix uffd-wp special cases for fork()
We tried to do something similar in b569a1760782 ("userfaultfd: wp: drop
_PAGE_UFFD_WP properly when fork") previously, but it's not doing it all
right.. A few fixes around the code path:
1. We were referencing VM_UFFD_WP vm_flags on the _old_ vma rather
than the new vma. That's overlooked in b569a1760782, so it won't work
as expected. Thanks to the recent rework on fork code
(7a4830c380f3a8b3), we can easily get the new vma now, so switch the
checks to that.
2. Dropping the uffd-wp bit in copy_huge_pmd() could be wrong if the
huge pmd is a migration huge pmd. When it happens, instead of using
pmd_uffd_wp(), we should use pmd_swp_uffd_wp(). The fix is simply to
handle them separately.
3. Forget to carry over uffd-wp bit for a write migration huge pmd
entry. This also happens in copy_huge_pmd(), where we converted a
write huge migration entry into a read one.
4. In copy_nonpresent_pte(), drop uffd-wp if necessary for swap ptes.
5. In copy_present_page() when COW is enforced when fork(), we also
need to pass over the uffd-wp bit if VM_UFFD_WP is armed on the new
vma, and when the pte to be copied has uffd-wp bit set.
Remove the comment in copy_present_pte() about this. It won't help a huge
lot to only comment there, but comment everywhere would be an overkill.
Let's assume the commit messages would help.
[peterx@redhat.com: fix a few thp pmd missing uffd-wp bit]
Link: https://lkml.kernel.org/r/20210428225030.9708-4-peterx@redhat.com
Link: https://lkml.kernel.org/r/20210428225030.9708-3-peterx@redhat.com
Fixes: b569a1760782f ("userfaultfd: wp: drop _PAGE_UFFD_WP properly when fork")
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Brian Geffon <bgeffon@google.com>
Cc: "Dr . David Alan Gilbert" <dgilbert@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Joe Perches <joe@perches.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Lokesh Gidra <lokeshgidra@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Oliver Upton <oupton@google.com>
Cc: Shaohua Li <shli@fb.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Wang Qing <wangqing@vivo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:49:02 +00:00
|
|
|
err = copy_huge_pmd(dst_mm, src_mm, dst_pmd, src_pmd,
|
|
|
|
addr, dst_vma, src_vma);
|
thp: transparent hugepage core
Lately I've been working to make KVM use hugepages transparently without
the usual restrictions of hugetlbfs. Some of the restrictions I'd like to
see removed:
1) hugepages have to be swappable or the guest physical memory remains
locked in RAM and can't be paged out to swap
2) if a hugepage allocation fails, regular pages should be allocated
instead and mixed in the same vma without any failure and without
userland noticing
3) if some task quits and more hugepages become available in the
buddy, guest physical memory backed by regular pages should be
relocated on hugepages automatically in regions under
madvise(MADV_HUGEPAGE) (ideally event driven by waking up the
kernel deamon if the order=HPAGE_PMD_SHIFT-PAGE_SHIFT list becomes
not null)
4) avoidance of reservation and maximization of use of hugepages whenever
possible. Reservation (needed to avoid runtime fatal faliures) may be ok for
1 machine with 1 database with 1 database cache with 1 database cache size
known at boot time. It's definitely not feasible with a virtualization
hypervisor usage like RHEV-H that runs an unknown number of virtual machines
with an unknown size of each virtual machine with an unknown amount of
pagecache that could be potentially useful in the host for guest not using
O_DIRECT (aka cache=off).
hugepages in the virtualization hypervisor (and also in the guest!) are
much more important than in a regular host not using virtualization,
becasue with NPT/EPT they decrease the tlb-miss cacheline accesses from 24
to 19 in case only the hypervisor uses transparent hugepages, and they
decrease the tlb-miss cacheline accesses from 19 to 15 in case both the
linux hypervisor and the linux guest both uses this patch (though the
guest will limit the addition speedup to anonymous regions only for
now...). Even more important is that the tlb miss handler is much slower
on a NPT/EPT guest than for a regular shadow paging or no-virtualization
scenario. So maximizing the amount of virtual memory cached by the TLB
pays off significantly more with NPT/EPT than without (even if there would
be no significant speedup in the tlb-miss runtime).
The first (and more tedious) part of this work requires allowing the VM to
handle anonymous hugepages mixed with regular pages transparently on
regular anonymous vmas. This is what this patch tries to achieve in the
least intrusive possible way. We want hugepages and hugetlb to be used in
a way so that all applications can benefit without changes (as usual we
leverage the KVM virtualization design: by improving the Linux VM at
large, KVM gets the performance boost too).
The most important design choice is: always fallback to 4k allocation if
the hugepage allocation fails! This is the _very_ opposite of some large
pagecache patches that failed with -EIO back then if a 64k (or similar)
allocation failed...
Second important decision (to reduce the impact of the feature on the
existing pagetable handling code) is that at any time we can split an
hugepage into 512 regular pages and it has to be done with an operation
that can't fail. This way the reliability of the swapping isn't decreased
(no need to allocate memory when we are short on memory to swap) and it's
trivial to plug a split_huge_page* one-liner where needed without
polluting the VM. Over time we can teach mprotect, mremap and friends to
handle pmd_trans_huge natively without calling split_huge_page*. The fact
it can't fail isn't just for swap: if split_huge_page would return -ENOMEM
(instead of the current void) we'd need to rollback the mprotect from the
middle of it (ideally including undoing the split_vma) which would be a
big change and in the very wrong direction (it'd likely be simpler not to
call split_huge_page at all and to teach mprotect and friends to handle
hugepages instead of rolling them back from the middle). In short the
very value of split_huge_page is that it can't fail.
The collapsing and madvise(MADV_HUGEPAGE) part will remain separated and
incremental and it'll just be an "harmless" addition later if this initial
part is agreed upon. It also should be noted that locking-wise replacing
regular pages with hugepages is going to be very easy if compared to what
I'm doing below in split_huge_page, as it will only happen when
page_count(page) matches page_mapcount(page) if we can take the PG_lock
and mmap_sem in write mode. collapse_huge_page will be a "best effort"
that (unlike split_huge_page) can fail at the minimal sign of trouble and
we can try again later. collapse_huge_page will be similar to how KSM
works and the madvise(MADV_HUGEPAGE) will work similar to
madvise(MADV_MERGEABLE).
The default I like is that transparent hugepages are used at page fault
time. This can be changed with
/sys/kernel/mm/transparent_hugepage/enabled. The control knob can be set
to three values "always", "madvise", "never" which mean respectively that
hugepages are always used, or only inside madvise(MADV_HUGEPAGE) regions,
or never used. /sys/kernel/mm/transparent_hugepage/defrag instead
controls if the hugepage allocation should defrag memory aggressively
"always", only inside "madvise" regions, or "never".
The pmd_trans_splitting/pmd_trans_huge locking is very solid. The
put_page (from get_user_page users that can't use mmu notifier like
O_DIRECT) that runs against a __split_huge_page_refcount instead was a
pain to serialize in a way that would result always in a coherent page
count for both tail and head. I think my locking solution with a
compound_lock taken only after the page_first is valid and is still a
PageHead should be safe but it surely needs review from SMP race point of
view. In short there is no current existing way to serialize the O_DIRECT
final put_page against split_huge_page_refcount so I had to invent a new
one (O_DIRECT loses knowledge on the mapping status by the time gup_fast
returns so...). And I didn't want to impact all gup/gup_fast users for
now, maybe if we change the gup interface substantially we can avoid this
locking, I admit I didn't think too much about it because changing the gup
unpinning interface would be invasive.
If we ignored O_DIRECT we could stick to the existing compound refcounting
code, by simply adding a get_user_pages_fast_flags(foll_flags) where KVM
(and any other mmu notifier user) would call it without FOLL_GET (and if
FOLL_GET isn't set we'd just BUG_ON if nobody registered itself in the
current task mmu notifier list yet). But O_DIRECT is fundamental for
decent performance of virtualized I/O on fast storage so we can't avoid it
to solve the race of put_page against split_huge_page_refcount to achieve
a complete hugepage feature for KVM.
Swap and oom works fine (well just like with regular pages ;). MMU
notifier is handled transparently too, with the exception of the young bit
on the pmd, that didn't have a range check but I think KVM will be fine
because the whole point of hugepages is that EPT/NPT will also use a huge
pmd when they notice gup returns pages with PageCompound set, so they
won't care of a range and there's just the pmd young bit to check in that
case.
NOTE: in some cases if the L2 cache is small, this may slowdown and waste
memory during COWs because 4M of memory are accessed in a single fault
instead of 8k (the payoff is that after COW the program can run faster).
So we might want to switch the copy_huge_page (and clear_huge_page too) to
not temporal stores. I also extensively researched ways to avoid this
cache trashing with a full prefault logic that would cow in 8k/16k/32k/64k
up to 1M (I can send those patches that fully implemented prefault) but I
concluded they're not worth it and they add an huge additional complexity
and they remove all tlb benefits until the full hugepage has been faulted
in, to save a little bit of memory and some cache during app startup, but
they still don't improve substantially the cache-trashing during startup
if the prefault happens in >4k chunks. One reason is that those 4k pte
entries copied are still mapped on a perfectly cache-colored hugepage, so
the trashing is the worst one can generate in those copies (cow of 4k page
copies aren't so well colored so they trashes less, but again this results
in software running faster after the page fault). Those prefault patches
allowed things like a pte where post-cow pages were local 4k regular anon
pages and the not-yet-cowed pte entries were pointing in the middle of
some hugepage mapped read-only. If it doesn't payoff substantially with
todays hardware it will payoff even less in the future with larger l2
caches, and the prefault logic would blot the VM a lot. If one is
emebdded transparent_hugepage can be disabled during boot with sysfs or
with the boot commandline parameter transparent_hugepage=0 (or
transparent_hugepage=2 to restrict hugepages inside madvise regions) that
will ensure not a single hugepage is allocated at boot time. It is simple
enough to just disable transparent hugepage globally and let transparent
hugepages be allocated selectively by applications in the MADV_HUGEPAGE
region (both at page fault time, and if enabled with the
collapse_huge_page too through the kernel daemon).
This patch supports only hugepages mapped in the pmd, archs that have
smaller hugepages will not fit in this patch alone. Also some archs like
power have certain tlb limits that prevents mixing different page size in
the same regions so they will not fit in this framework that requires
"graceful fallback" to basic PAGE_SIZE in case of physical memory
fragmentation. hugetlbfs remains a perfect fit for those because its
software limits happen to match the hardware limits. hugetlbfs also
remains a perfect fit for hugepage sizes like 1GByte that cannot be hoped
to be found not fragmented after a certain system uptime and that would be
very expensive to defragment with relocation, so requiring reservation.
hugetlbfs is the "reservation way", the point of transparent hugepages is
not to have any reservation at all and maximizing the use of cache and
hugepages at all times automatically.
Some performance result:
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largep
ages3
memset page fault 1566023
memset tlb miss 453854
memset second tlb miss 453321
random access tlb miss 41635
random access second tlb miss 41658
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largepages3
memset page fault 1566471
memset tlb miss 453375
memset second tlb miss 453320
random access tlb miss 41636
random access second tlb miss 41637
vmx andrea # ./largepages3
memset page fault 1566642
memset tlb miss 453417
memset second tlb miss 453313
random access tlb miss 41630
random access second tlb miss 41647
vmx andrea # ./largepages3
memset page fault 1566872
memset tlb miss 453418
memset second tlb miss 453315
random access tlb miss 41618
random access second tlb miss 41659
vmx andrea # echo 0 > /proc/sys/vm/transparent_hugepage
vmx andrea # ./largepages3
memset page fault 2182476
memset tlb miss 460305
memset second tlb miss 460179
random access tlb miss 44483
random access second tlb miss 44186
vmx andrea # ./largepages3
memset page fault 2182791
memset tlb miss 460742
memset second tlb miss 459962
random access tlb miss 43981
random access second tlb miss 43988
============
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#define SIZE (3UL*1024*1024*1024)
int main()
{
char *p = malloc(SIZE), *p2;
struct timeval before, after;
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset page fault %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
return 0;
}
============
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:46:52 +00:00
|
|
|
if (err == -ENOMEM)
|
|
|
|
return -ENOMEM;
|
|
|
|
if (!err)
|
|
|
|
continue;
|
|
|
|
/* fall through */
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
if (pmd_none_or_clear_bad(src_pmd))
|
|
|
|
continue;
|
2020-10-13 23:54:21 +00:00
|
|
|
if (copy_pte_range(dst_vma, src_vma, dst_pmd, src_pmd,
|
|
|
|
addr, next))
|
2005-04-16 22:20:36 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
} while (dst_pmd++, src_pmd++, addr = next, addr != end);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
static inline int
|
|
|
|
copy_pud_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
|
|
|
|
p4d_t *dst_p4d, p4d_t *src_p4d, unsigned long addr,
|
|
|
|
unsigned long end)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2020-10-13 23:54:21 +00:00
|
|
|
struct mm_struct *dst_mm = dst_vma->vm_mm;
|
|
|
|
struct mm_struct *src_mm = src_vma->vm_mm;
|
2005-04-16 22:20:36 +00:00
|
|
|
pud_t *src_pud, *dst_pud;
|
|
|
|
unsigned long next;
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
dst_pud = pud_alloc(dst_mm, dst_p4d, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (!dst_pud)
|
|
|
|
return -ENOMEM;
|
2017-03-09 14:24:07 +00:00
|
|
|
src_pud = pud_offset(src_p4d, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
do {
|
|
|
|
next = pud_addr_end(addr, end);
|
2017-02-24 22:57:02 +00:00
|
|
|
if (pud_trans_huge(*src_pud) || pud_devmap(*src_pud)) {
|
|
|
|
int err;
|
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
VM_BUG_ON_VMA(next-addr != HPAGE_PUD_SIZE, src_vma);
|
2017-02-24 22:57:02 +00:00
|
|
|
err = copy_huge_pud(dst_mm, src_mm,
|
2020-10-13 23:54:21 +00:00
|
|
|
dst_pud, src_pud, addr, src_vma);
|
2017-02-24 22:57:02 +00:00
|
|
|
if (err == -ENOMEM)
|
|
|
|
return -ENOMEM;
|
|
|
|
if (!err)
|
|
|
|
continue;
|
|
|
|
/* fall through */
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
if (pud_none_or_clear_bad(src_pud))
|
|
|
|
continue;
|
2020-10-13 23:54:21 +00:00
|
|
|
if (copy_pmd_range(dst_vma, src_vma, dst_pud, src_pud,
|
|
|
|
addr, next))
|
2005-04-16 22:20:36 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
} while (dst_pud++, src_pud++, addr = next, addr != end);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
static inline int
|
|
|
|
copy_p4d_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
|
|
|
|
pgd_t *dst_pgd, pgd_t *src_pgd, unsigned long addr,
|
|
|
|
unsigned long end)
|
2017-03-09 14:24:07 +00:00
|
|
|
{
|
2020-10-13 23:54:21 +00:00
|
|
|
struct mm_struct *dst_mm = dst_vma->vm_mm;
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d_t *src_p4d, *dst_p4d;
|
|
|
|
unsigned long next;
|
|
|
|
|
|
|
|
dst_p4d = p4d_alloc(dst_mm, dst_pgd, addr);
|
|
|
|
if (!dst_p4d)
|
|
|
|
return -ENOMEM;
|
|
|
|
src_p4d = p4d_offset(src_pgd, addr);
|
|
|
|
do {
|
|
|
|
next = p4d_addr_end(addr, end);
|
|
|
|
if (p4d_none_or_clear_bad(src_p4d))
|
|
|
|
continue;
|
2020-10-13 23:54:21 +00:00
|
|
|
if (copy_pud_range(dst_vma, src_vma, dst_p4d, src_p4d,
|
|
|
|
addr, next))
|
2017-03-09 14:24:07 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
} while (dst_p4d++, src_p4d++, addr = next, addr != end);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2022-05-13 03:22:53 +00:00
|
|
|
/*
|
|
|
|
* Return true if the vma needs to copy the pgtable during this fork(). Return
|
|
|
|
* false when we can speed up fork() by allowing lazy page faults later until
|
|
|
|
* when the child accesses the memory range.
|
|
|
|
*/
|
2022-05-13 03:22:55 +00:00
|
|
|
static bool
|
2022-05-13 03:22:53 +00:00
|
|
|
vma_needs_copy(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Always copy pgtables when dst_vma has uffd-wp enabled even if it's
|
|
|
|
* file-backed (e.g. shmem). Because when uffd-wp is enabled, pgtable
|
|
|
|
* contains uffd-wp protection information, that's something we can't
|
|
|
|
* retrieve from page cache, and skip copying will lose those info.
|
|
|
|
*/
|
|
|
|
if (userfaultfd_wp(dst_vma))
|
|
|
|
return true;
|
|
|
|
|
2022-06-21 23:56:20 +00:00
|
|
|
if (src_vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
|
2022-05-13 03:22:53 +00:00
|
|
|
return true;
|
|
|
|
|
|
|
|
if (src_vma->anon_vma)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Don't copy ptes where a page fault will fill them correctly. Fork
|
|
|
|
* becomes much lighter when there are big shared or private readonly
|
|
|
|
* mappings. The tradeoff is that copy_page_range is more efficient
|
|
|
|
* than faulting.
|
|
|
|
*/
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
int
|
|
|
|
copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pgd_t *src_pgd, *dst_pgd;
|
|
|
|
unsigned long next;
|
2020-10-13 23:54:21 +00:00
|
|
|
unsigned long addr = src_vma->vm_start;
|
|
|
|
unsigned long end = src_vma->vm_end;
|
|
|
|
struct mm_struct *dst_mm = dst_vma->vm_mm;
|
|
|
|
struct mm_struct *src_mm = src_vma->vm_mm;
|
2018-12-28 08:38:09 +00:00
|
|
|
struct mmu_notifier_range range;
|
mm: move all mmu notifier invocations to be done outside the PT lock
In order to allow sleeping during mmu notifier calls, we need to avoid
invoking them under the page table spinlock. This patch solves the
problem by calling invalidate_page notification after releasing the lock
(but before freeing the page itself), or by wrapping the page invalidation
with calls to invalidate_range_begin and invalidate_range_end.
To prevent accidental changes to the invalidate_range_end arguments after
the call to invalidate_range_begin, the patch introduces a convention of
saving the arguments in consistently named locals:
unsigned long mmun_start; /* For mmu_notifiers */
unsigned long mmun_end; /* For mmu_notifiers */
...
mmun_start = ...
mmun_end = ...
mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
...
mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
The patch changes code to use this convention for all calls to
mmu_notifier_invalidate_range_start/end, except those where the calls are
close enough so that anyone who glances at the code can see the values
aren't changing.
This patchset is a preliminary step towards on-demand paging design to be
added to the RDMA stack.
Why do we want on-demand paging for Infiniband?
Applications register memory with an RDMA adapter using system calls,
and subsequently post IO operations that refer to the corresponding
virtual addresses directly to HW. Until now, this was achieved by
pinning the memory during the registration calls. The goal of on demand
paging is to avoid pinning the pages of registered memory regions (MRs).
This will allow users the same flexibility they get when swapping any
other part of their processes address spaces. Instead of requiring the
entire MR to fit in physical memory, we can allow the MR to be larger,
and only fit the current working set in physical memory.
Why should anyone care? What problems are users currently experiencing?
This can make programming with RDMA much simpler. Today, developers
that are working with more data than their RAM can hold need either to
deregister and reregister memory regions throughout their process's
life, or keep a single memory region and copy the data to it. On demand
paging will allow these developers to register a single MR at the
beginning of their process's life, and let the operating system manage
which pages needs to be fetched at a given time. In the future, we
might be able to provide a single memory access key for each process
that would provide the entire process's address as one large memory
region, and the developers wouldn't need to register memory regions at
all.
Is there any prospect that any other subsystems will utilise these
infrastructural changes? If so, which and how, etc?
As for other subsystems, I understand that XPMEM wanted to sleep in
MMU notifiers, as Christoph Lameter wrote at
http://lkml.indiana.edu/hypermail/linux/kernel/0802.1/0460.html and
perhaps Andrea knows about other use cases.
Scheduling in mmu notifications is required since we need to sync the
hardware with the secondary page tables change. A TLB flush of an IO
device is inherently slower than a CPU TLB flush, so our design works by
sending the invalidation request to the device, and waiting for an
interrupt before exiting the mmu notifier handler.
Avi said:
kvm may be a buyer. kvm::mmu_lock, which serializes guest page
faults, also protects long operations such as destroying large ranges.
It would be good to convert it into a spinlock, but as it is used inside
mmu notifiers, this cannot be done.
(there are alternatives, such as keeping the spinlock and using a
generation counter to do the teardown in O(1), which is what the "may"
is doing up there).
[akpm@linux-foundation.orgpossible speed tweak in hugetlb_cow(), cleanups]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Haggai Eran <haggaie@mellanox.com>
Cc: Shachar Raindel <raindel@mellanox.com>
Cc: Liran Liss <liranl@mellanox.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Avi Kivity <avi@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-08 23:33:33 +00:00
|
|
|
bool is_cow;
|
mmu-notifiers: core
With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages.
There are secondary MMUs (with secondary sptes and secondary tlbs) too.
sptes in the kvm case are shadow pagetables, but when I say spte in
mmu-notifier context, I mean "secondary pte". In GRU case there's no
actual secondary pte and there's only a secondary tlb because the GRU
secondary MMU has no knowledge about sptes and every secondary tlb miss
event in the MMU always generates a page fault that has to be resolved by
the CPU (this is not the case of KVM where the a secondary tlb miss will
walk sptes in hardware and it will refill the secondary tlb transparently
to software if the corresponding spte is present). The same way
zap_page_range has to invalidate the pte before freeing the page, the spte
(and secondary tlb) must also be invalidated before any page is freed and
reused.
Currently we take a page_count pin on every page mapped by sptes, but that
means the pages can't be swapped whenever they're mapped by any spte
because they're part of the guest working set. Furthermore a spte unmap
event can immediately lead to a page to be freed when the pin is released
(so requiring the same complex and relatively slow tlb_gather smp safe
logic we have in zap_page_range and that can be avoided completely if the
spte unmap event doesn't require an unpin of the page previously mapped in
the secondary MMU).
The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know
when the VM is swapping or freeing or doing anything on the primary MMU so
that the secondary MMU code can drop sptes before the pages are freed,
avoiding all page pinning and allowing 100% reliable swapping of guest
physical address space. Furthermore it avoids the code that teardown the
mappings of the secondary MMU, to implement a logic like tlb_gather in
zap_page_range that would require many IPI to flush other cpu tlbs, for
each fixed number of spte unmapped.
To make an example: if what happens on the primary MMU is a protection
downgrade (from writeable to wrprotect) the secondary MMU mappings will be
invalidated, and the next secondary-mmu-page-fault will call
get_user_pages and trigger a do_wp_page through get_user_pages if it
called get_user_pages with write=1, and it'll re-establishing an updated
spte or secondary-tlb-mapping on the copied page. Or it will setup a
readonly spte or readonly tlb mapping if it's a guest-read, if it calls
get_user_pages with write=0. This is just an example.
This allows to map any page pointed by any pte (and in turn visible in the
primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an
full MMU with both sptes and secondary-tlb like the shadow-pagetable layer
with kvm), or a remote DMA in software like XPMEM (hence needing of
schedule in XPMEM code to send the invalidate to the remote node, while no
need to schedule in kvm/gru as it's an immediate event like invalidating
primary-mmu pte).
At least for KVM without this patch it's impossible to swap guests
reliably. And having this feature and removing the page pin allows
several other optimizations that simplify life considerably.
Dependencies:
1) mm_take_all_locks() to register the mmu notifier when the whole VM
isn't doing anything with "mm". This allows mmu notifier users to keep
track if the VM is in the middle of the invalidate_range_begin/end
critical section with an atomic counter incraese in range_begin and
decreased in range_end. No secondary MMU page fault is allowed to map
any spte or secondary tlb reference, while the VM is in the middle of
range_begin/end as any page returned by get_user_pages in that critical
section could later immediately be freed without any further
->invalidate_page notification (invalidate_range_begin/end works on
ranges and ->invalidate_page isn't called immediately before freeing
the page). To stop all page freeing and pagetable overwrites the
mmap_sem must be taken in write mode and all other anon_vma/i_mmap
locks must be taken too.
2) It'd be a waste to add branches in the VM if nobody could possibly
run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if
CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of
mmu notifiers, but this already allows to compile a KVM external module
against a kernel with mmu notifiers enabled and from the next pull from
kvm.git we'll start using them. And GRU/XPMEM will also be able to
continue the development by enabling KVM=m in their config, until they
submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can
also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n).
This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM
are all =n.
The mmu_notifier_register call can fail because mm_take_all_locks may be
interrupted by a signal and return -EINTR. Because mmu_notifier_reigster
is used when a driver startup, a failure can be gracefully handled. Here
an example of the change applied to kvm to register the mmu notifiers.
Usually when a driver startups other allocations are required anyway and
-ENOMEM failure paths exists already.
struct kvm *kvm_arch_create_vm(void)
{
struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ int err;
if (!kvm)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
+ kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops;
+ err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm);
+ if (err) {
+ kfree(kvm);
+ return ERR_PTR(err);
+ }
+
return kvm;
}
mmu_notifier_unregister returns void and it's reliable.
The patch also adds a few needed but missing includes that would prevent
kernel to compile after these changes on non-x86 archs (x86 didn't need
them by luck).
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix mm/filemap_xip.c build]
[akpm@linux-foundation.org: fix mm/mmu_notifier.c build]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Jack Steiner <steiner@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Kanoj Sarcar <kanojsarcar@yahoo.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Avi Kivity <avi@qumranet.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Chris Wright <chrisw@redhat.com>
Cc: Marcelo Tosatti <marcelo@kvack.org>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: Izik Eidus <izike@qumranet.com>
Cc: Anthony Liguori <aliguori@us.ibm.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>
2008-07-28 22:46:29 +00:00
|
|
|
int ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2022-05-13 03:22:53 +00:00
|
|
|
if (!vma_needs_copy(dst_vma, src_vma))
|
2015-02-10 22:10:04 +00:00
|
|
|
return 0;
|
2005-08-28 06:49:11 +00:00
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
if (is_vm_hugetlb_page(src_vma))
|
2022-05-13 03:22:55 +00:00
|
|
|
return copy_hugetlb_page_range(dst_mm, src_mm, dst_vma, src_vma);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2020-10-13 23:54:21 +00:00
|
|
|
if (unlikely(src_vma->vm_flags & VM_PFNMAP)) {
|
2008-12-18 19:41:29 +00:00
|
|
|
/*
|
|
|
|
* We do not free on error cases below as remove_vma
|
|
|
|
* gets called on error from higher level routine
|
|
|
|
*/
|
2020-10-13 23:54:21 +00:00
|
|
|
ret = track_pfn_copy(src_vma);
|
2008-12-18 19:41:29 +00:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
mmu-notifiers: core
With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages.
There are secondary MMUs (with secondary sptes and secondary tlbs) too.
sptes in the kvm case are shadow pagetables, but when I say spte in
mmu-notifier context, I mean "secondary pte". In GRU case there's no
actual secondary pte and there's only a secondary tlb because the GRU
secondary MMU has no knowledge about sptes and every secondary tlb miss
event in the MMU always generates a page fault that has to be resolved by
the CPU (this is not the case of KVM where the a secondary tlb miss will
walk sptes in hardware and it will refill the secondary tlb transparently
to software if the corresponding spte is present). The same way
zap_page_range has to invalidate the pte before freeing the page, the spte
(and secondary tlb) must also be invalidated before any page is freed and
reused.
Currently we take a page_count pin on every page mapped by sptes, but that
means the pages can't be swapped whenever they're mapped by any spte
because they're part of the guest working set. Furthermore a spte unmap
event can immediately lead to a page to be freed when the pin is released
(so requiring the same complex and relatively slow tlb_gather smp safe
logic we have in zap_page_range and that can be avoided completely if the
spte unmap event doesn't require an unpin of the page previously mapped in
the secondary MMU).
The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know
when the VM is swapping or freeing or doing anything on the primary MMU so
that the secondary MMU code can drop sptes before the pages are freed,
avoiding all page pinning and allowing 100% reliable swapping of guest
physical address space. Furthermore it avoids the code that teardown the
mappings of the secondary MMU, to implement a logic like tlb_gather in
zap_page_range that would require many IPI to flush other cpu tlbs, for
each fixed number of spte unmapped.
To make an example: if what happens on the primary MMU is a protection
downgrade (from writeable to wrprotect) the secondary MMU mappings will be
invalidated, and the next secondary-mmu-page-fault will call
get_user_pages and trigger a do_wp_page through get_user_pages if it
called get_user_pages with write=1, and it'll re-establishing an updated
spte or secondary-tlb-mapping on the copied page. Or it will setup a
readonly spte or readonly tlb mapping if it's a guest-read, if it calls
get_user_pages with write=0. This is just an example.
This allows to map any page pointed by any pte (and in turn visible in the
primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an
full MMU with both sptes and secondary-tlb like the shadow-pagetable layer
with kvm), or a remote DMA in software like XPMEM (hence needing of
schedule in XPMEM code to send the invalidate to the remote node, while no
need to schedule in kvm/gru as it's an immediate event like invalidating
primary-mmu pte).
At least for KVM without this patch it's impossible to swap guests
reliably. And having this feature and removing the page pin allows
several other optimizations that simplify life considerably.
Dependencies:
1) mm_take_all_locks() to register the mmu notifier when the whole VM
isn't doing anything with "mm". This allows mmu notifier users to keep
track if the VM is in the middle of the invalidate_range_begin/end
critical section with an atomic counter incraese in range_begin and
decreased in range_end. No secondary MMU page fault is allowed to map
any spte or secondary tlb reference, while the VM is in the middle of
range_begin/end as any page returned by get_user_pages in that critical
section could later immediately be freed without any further
->invalidate_page notification (invalidate_range_begin/end works on
ranges and ->invalidate_page isn't called immediately before freeing
the page). To stop all page freeing and pagetable overwrites the
mmap_sem must be taken in write mode and all other anon_vma/i_mmap
locks must be taken too.
2) It'd be a waste to add branches in the VM if nobody could possibly
run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if
CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of
mmu notifiers, but this already allows to compile a KVM external module
against a kernel with mmu notifiers enabled and from the next pull from
kvm.git we'll start using them. And GRU/XPMEM will also be able to
continue the development by enabling KVM=m in their config, until they
submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can
also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n).
This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM
are all =n.
The mmu_notifier_register call can fail because mm_take_all_locks may be
interrupted by a signal and return -EINTR. Because mmu_notifier_reigster
is used when a driver startup, a failure can be gracefully handled. Here
an example of the change applied to kvm to register the mmu notifiers.
Usually when a driver startups other allocations are required anyway and
-ENOMEM failure paths exists already.
struct kvm *kvm_arch_create_vm(void)
{
struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ int err;
if (!kvm)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
+ kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops;
+ err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm);
+ if (err) {
+ kfree(kvm);
+ return ERR_PTR(err);
+ }
+
return kvm;
}
mmu_notifier_unregister returns void and it's reliable.
The patch also adds a few needed but missing includes that would prevent
kernel to compile after these changes on non-x86 archs (x86 didn't need
them by luck).
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix mm/filemap_xip.c build]
[akpm@linux-foundation.org: fix mm/mmu_notifier.c build]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Jack Steiner <steiner@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Kanoj Sarcar <kanojsarcar@yahoo.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Avi Kivity <avi@qumranet.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Chris Wright <chrisw@redhat.com>
Cc: Marcelo Tosatti <marcelo@kvack.org>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: Izik Eidus <izike@qumranet.com>
Cc: Anthony Liguori <aliguori@us.ibm.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>
2008-07-28 22:46:29 +00:00
|
|
|
/*
|
|
|
|
* We need to invalidate the secondary MMU mappings only when
|
|
|
|
* there could be a permission downgrade on the ptes of the
|
|
|
|
* parent mm. And a permission downgrade will only happen if
|
|
|
|
* is_cow_mapping() returns true.
|
|
|
|
*/
|
2020-10-13 23:54:21 +00:00
|
|
|
is_cow = is_cow_mapping(src_vma->vm_flags);
|
2018-12-28 08:38:09 +00:00
|
|
|
|
|
|
|
if (is_cow) {
|
2019-05-14 00:20:53 +00:00
|
|
|
mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_PAGE,
|
2023-01-10 02:57:22 +00:00
|
|
|
0, src_mm, addr, end);
|
2018-12-28 08:38:09 +00:00
|
|
|
mmu_notifier_invalidate_range_start(&range);
|
2020-12-15 03:05:44 +00:00
|
|
|
/*
|
|
|
|
* Disabling preemption is not needed for the write side, as
|
|
|
|
* the read side doesn't spin, but goes to the mmap_lock.
|
|
|
|
*
|
|
|
|
* Use the raw variant of the seqcount_t write API to avoid
|
|
|
|
* lockdep complaining about preemptibility.
|
|
|
|
*/
|
|
|
|
mmap_assert_write_locked(src_mm);
|
|
|
|
raw_write_seqcount_begin(&src_mm->write_protect_seq);
|
2018-12-28 08:38:09 +00:00
|
|
|
}
|
mmu-notifiers: core
With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages.
There are secondary MMUs (with secondary sptes and secondary tlbs) too.
sptes in the kvm case are shadow pagetables, but when I say spte in
mmu-notifier context, I mean "secondary pte". In GRU case there's no
actual secondary pte and there's only a secondary tlb because the GRU
secondary MMU has no knowledge about sptes and every secondary tlb miss
event in the MMU always generates a page fault that has to be resolved by
the CPU (this is not the case of KVM where the a secondary tlb miss will
walk sptes in hardware and it will refill the secondary tlb transparently
to software if the corresponding spte is present). The same way
zap_page_range has to invalidate the pte before freeing the page, the spte
(and secondary tlb) must also be invalidated before any page is freed and
reused.
Currently we take a page_count pin on every page mapped by sptes, but that
means the pages can't be swapped whenever they're mapped by any spte
because they're part of the guest working set. Furthermore a spte unmap
event can immediately lead to a page to be freed when the pin is released
(so requiring the same complex and relatively slow tlb_gather smp safe
logic we have in zap_page_range and that can be avoided completely if the
spte unmap event doesn't require an unpin of the page previously mapped in
the secondary MMU).
The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know
when the VM is swapping or freeing or doing anything on the primary MMU so
that the secondary MMU code can drop sptes before the pages are freed,
avoiding all page pinning and allowing 100% reliable swapping of guest
physical address space. Furthermore it avoids the code that teardown the
mappings of the secondary MMU, to implement a logic like tlb_gather in
zap_page_range that would require many IPI to flush other cpu tlbs, for
each fixed number of spte unmapped.
To make an example: if what happens on the primary MMU is a protection
downgrade (from writeable to wrprotect) the secondary MMU mappings will be
invalidated, and the next secondary-mmu-page-fault will call
get_user_pages and trigger a do_wp_page through get_user_pages if it
called get_user_pages with write=1, and it'll re-establishing an updated
spte or secondary-tlb-mapping on the copied page. Or it will setup a
readonly spte or readonly tlb mapping if it's a guest-read, if it calls
get_user_pages with write=0. This is just an example.
This allows to map any page pointed by any pte (and in turn visible in the
primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an
full MMU with both sptes and secondary-tlb like the shadow-pagetable layer
with kvm), or a remote DMA in software like XPMEM (hence needing of
schedule in XPMEM code to send the invalidate to the remote node, while no
need to schedule in kvm/gru as it's an immediate event like invalidating
primary-mmu pte).
At least for KVM without this patch it's impossible to swap guests
reliably. And having this feature and removing the page pin allows
several other optimizations that simplify life considerably.
Dependencies:
1) mm_take_all_locks() to register the mmu notifier when the whole VM
isn't doing anything with "mm". This allows mmu notifier users to keep
track if the VM is in the middle of the invalidate_range_begin/end
critical section with an atomic counter incraese in range_begin and
decreased in range_end. No secondary MMU page fault is allowed to map
any spte or secondary tlb reference, while the VM is in the middle of
range_begin/end as any page returned by get_user_pages in that critical
section could later immediately be freed without any further
->invalidate_page notification (invalidate_range_begin/end works on
ranges and ->invalidate_page isn't called immediately before freeing
the page). To stop all page freeing and pagetable overwrites the
mmap_sem must be taken in write mode and all other anon_vma/i_mmap
locks must be taken too.
2) It'd be a waste to add branches in the VM if nobody could possibly
run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if
CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of
mmu notifiers, but this already allows to compile a KVM external module
against a kernel with mmu notifiers enabled and from the next pull from
kvm.git we'll start using them. And GRU/XPMEM will also be able to
continue the development by enabling KVM=m in their config, until they
submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can
also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n).
This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM
are all =n.
The mmu_notifier_register call can fail because mm_take_all_locks may be
interrupted by a signal and return -EINTR. Because mmu_notifier_reigster
is used when a driver startup, a failure can be gracefully handled. Here
an example of the change applied to kvm to register the mmu notifiers.
Usually when a driver startups other allocations are required anyway and
-ENOMEM failure paths exists already.
struct kvm *kvm_arch_create_vm(void)
{
struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ int err;
if (!kvm)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
+ kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops;
+ err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm);
+ if (err) {
+ kfree(kvm);
+ return ERR_PTR(err);
+ }
+
return kvm;
}
mmu_notifier_unregister returns void and it's reliable.
The patch also adds a few needed but missing includes that would prevent
kernel to compile after these changes on non-x86 archs (x86 didn't need
them by luck).
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix mm/filemap_xip.c build]
[akpm@linux-foundation.org: fix mm/mmu_notifier.c build]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Jack Steiner <steiner@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Kanoj Sarcar <kanojsarcar@yahoo.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Avi Kivity <avi@qumranet.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Chris Wright <chrisw@redhat.com>
Cc: Marcelo Tosatti <marcelo@kvack.org>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: Izik Eidus <izike@qumranet.com>
Cc: Anthony Liguori <aliguori@us.ibm.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>
2008-07-28 22:46:29 +00:00
|
|
|
|
|
|
|
ret = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
dst_pgd = pgd_offset(dst_mm, addr);
|
|
|
|
src_pgd = pgd_offset(src_mm, addr);
|
|
|
|
do {
|
|
|
|
next = pgd_addr_end(addr, end);
|
|
|
|
if (pgd_none_or_clear_bad(src_pgd))
|
|
|
|
continue;
|
2020-10-13 23:54:21 +00:00
|
|
|
if (unlikely(copy_p4d_range(dst_vma, src_vma, dst_pgd, src_pgd,
|
|
|
|
addr, next))) {
|
2023-02-17 02:56:15 +00:00
|
|
|
untrack_pfn_clear(dst_vma);
|
mmu-notifiers: core
With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages.
There are secondary MMUs (with secondary sptes and secondary tlbs) too.
sptes in the kvm case are shadow pagetables, but when I say spte in
mmu-notifier context, I mean "secondary pte". In GRU case there's no
actual secondary pte and there's only a secondary tlb because the GRU
secondary MMU has no knowledge about sptes and every secondary tlb miss
event in the MMU always generates a page fault that has to be resolved by
the CPU (this is not the case of KVM where the a secondary tlb miss will
walk sptes in hardware and it will refill the secondary tlb transparently
to software if the corresponding spte is present). The same way
zap_page_range has to invalidate the pte before freeing the page, the spte
(and secondary tlb) must also be invalidated before any page is freed and
reused.
Currently we take a page_count pin on every page mapped by sptes, but that
means the pages can't be swapped whenever they're mapped by any spte
because they're part of the guest working set. Furthermore a spte unmap
event can immediately lead to a page to be freed when the pin is released
(so requiring the same complex and relatively slow tlb_gather smp safe
logic we have in zap_page_range and that can be avoided completely if the
spte unmap event doesn't require an unpin of the page previously mapped in
the secondary MMU).
The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know
when the VM is swapping or freeing or doing anything on the primary MMU so
that the secondary MMU code can drop sptes before the pages are freed,
avoiding all page pinning and allowing 100% reliable swapping of guest
physical address space. Furthermore it avoids the code that teardown the
mappings of the secondary MMU, to implement a logic like tlb_gather in
zap_page_range that would require many IPI to flush other cpu tlbs, for
each fixed number of spte unmapped.
To make an example: if what happens on the primary MMU is a protection
downgrade (from writeable to wrprotect) the secondary MMU mappings will be
invalidated, and the next secondary-mmu-page-fault will call
get_user_pages and trigger a do_wp_page through get_user_pages if it
called get_user_pages with write=1, and it'll re-establishing an updated
spte or secondary-tlb-mapping on the copied page. Or it will setup a
readonly spte or readonly tlb mapping if it's a guest-read, if it calls
get_user_pages with write=0. This is just an example.
This allows to map any page pointed by any pte (and in turn visible in the
primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an
full MMU with both sptes and secondary-tlb like the shadow-pagetable layer
with kvm), or a remote DMA in software like XPMEM (hence needing of
schedule in XPMEM code to send the invalidate to the remote node, while no
need to schedule in kvm/gru as it's an immediate event like invalidating
primary-mmu pte).
At least for KVM without this patch it's impossible to swap guests
reliably. And having this feature and removing the page pin allows
several other optimizations that simplify life considerably.
Dependencies:
1) mm_take_all_locks() to register the mmu notifier when the whole VM
isn't doing anything with "mm". This allows mmu notifier users to keep
track if the VM is in the middle of the invalidate_range_begin/end
critical section with an atomic counter incraese in range_begin and
decreased in range_end. No secondary MMU page fault is allowed to map
any spte or secondary tlb reference, while the VM is in the middle of
range_begin/end as any page returned by get_user_pages in that critical
section could later immediately be freed without any further
->invalidate_page notification (invalidate_range_begin/end works on
ranges and ->invalidate_page isn't called immediately before freeing
the page). To stop all page freeing and pagetable overwrites the
mmap_sem must be taken in write mode and all other anon_vma/i_mmap
locks must be taken too.
2) It'd be a waste to add branches in the VM if nobody could possibly
run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if
CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of
mmu notifiers, but this already allows to compile a KVM external module
against a kernel with mmu notifiers enabled and from the next pull from
kvm.git we'll start using them. And GRU/XPMEM will also be able to
continue the development by enabling KVM=m in their config, until they
submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can
also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n).
This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM
are all =n.
The mmu_notifier_register call can fail because mm_take_all_locks may be
interrupted by a signal and return -EINTR. Because mmu_notifier_reigster
is used when a driver startup, a failure can be gracefully handled. Here
an example of the change applied to kvm to register the mmu notifiers.
Usually when a driver startups other allocations are required anyway and
-ENOMEM failure paths exists already.
struct kvm *kvm_arch_create_vm(void)
{
struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ int err;
if (!kvm)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
+ kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops;
+ err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm);
+ if (err) {
+ kfree(kvm);
+ return ERR_PTR(err);
+ }
+
return kvm;
}
mmu_notifier_unregister returns void and it's reliable.
The patch also adds a few needed but missing includes that would prevent
kernel to compile after these changes on non-x86 archs (x86 didn't need
them by luck).
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix mm/filemap_xip.c build]
[akpm@linux-foundation.org: fix mm/mmu_notifier.c build]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Jack Steiner <steiner@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Kanoj Sarcar <kanojsarcar@yahoo.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Avi Kivity <avi@qumranet.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Chris Wright <chrisw@redhat.com>
Cc: Marcelo Tosatti <marcelo@kvack.org>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: Izik Eidus <izike@qumranet.com>
Cc: Anthony Liguori <aliguori@us.ibm.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>
2008-07-28 22:46:29 +00:00
|
|
|
ret = -ENOMEM;
|
|
|
|
break;
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
} while (dst_pgd++, src_pgd++, addr = next, addr != end);
|
mmu-notifiers: core
With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages.
There are secondary MMUs (with secondary sptes and secondary tlbs) too.
sptes in the kvm case are shadow pagetables, but when I say spte in
mmu-notifier context, I mean "secondary pte". In GRU case there's no
actual secondary pte and there's only a secondary tlb because the GRU
secondary MMU has no knowledge about sptes and every secondary tlb miss
event in the MMU always generates a page fault that has to be resolved by
the CPU (this is not the case of KVM where the a secondary tlb miss will
walk sptes in hardware and it will refill the secondary tlb transparently
to software if the corresponding spte is present). The same way
zap_page_range has to invalidate the pte before freeing the page, the spte
(and secondary tlb) must also be invalidated before any page is freed and
reused.
Currently we take a page_count pin on every page mapped by sptes, but that
means the pages can't be swapped whenever they're mapped by any spte
because they're part of the guest working set. Furthermore a spte unmap
event can immediately lead to a page to be freed when the pin is released
(so requiring the same complex and relatively slow tlb_gather smp safe
logic we have in zap_page_range and that can be avoided completely if the
spte unmap event doesn't require an unpin of the page previously mapped in
the secondary MMU).
The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know
when the VM is swapping or freeing or doing anything on the primary MMU so
that the secondary MMU code can drop sptes before the pages are freed,
avoiding all page pinning and allowing 100% reliable swapping of guest
physical address space. Furthermore it avoids the code that teardown the
mappings of the secondary MMU, to implement a logic like tlb_gather in
zap_page_range that would require many IPI to flush other cpu tlbs, for
each fixed number of spte unmapped.
To make an example: if what happens on the primary MMU is a protection
downgrade (from writeable to wrprotect) the secondary MMU mappings will be
invalidated, and the next secondary-mmu-page-fault will call
get_user_pages and trigger a do_wp_page through get_user_pages if it
called get_user_pages with write=1, and it'll re-establishing an updated
spte or secondary-tlb-mapping on the copied page. Or it will setup a
readonly spte or readonly tlb mapping if it's a guest-read, if it calls
get_user_pages with write=0. This is just an example.
This allows to map any page pointed by any pte (and in turn visible in the
primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an
full MMU with both sptes and secondary-tlb like the shadow-pagetable layer
with kvm), or a remote DMA in software like XPMEM (hence needing of
schedule in XPMEM code to send the invalidate to the remote node, while no
need to schedule in kvm/gru as it's an immediate event like invalidating
primary-mmu pte).
At least for KVM without this patch it's impossible to swap guests
reliably. And having this feature and removing the page pin allows
several other optimizations that simplify life considerably.
Dependencies:
1) mm_take_all_locks() to register the mmu notifier when the whole VM
isn't doing anything with "mm". This allows mmu notifier users to keep
track if the VM is in the middle of the invalidate_range_begin/end
critical section with an atomic counter incraese in range_begin and
decreased in range_end. No secondary MMU page fault is allowed to map
any spte or secondary tlb reference, while the VM is in the middle of
range_begin/end as any page returned by get_user_pages in that critical
section could later immediately be freed without any further
->invalidate_page notification (invalidate_range_begin/end works on
ranges and ->invalidate_page isn't called immediately before freeing
the page). To stop all page freeing and pagetable overwrites the
mmap_sem must be taken in write mode and all other anon_vma/i_mmap
locks must be taken too.
2) It'd be a waste to add branches in the VM if nobody could possibly
run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if
CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of
mmu notifiers, but this already allows to compile a KVM external module
against a kernel with mmu notifiers enabled and from the next pull from
kvm.git we'll start using them. And GRU/XPMEM will also be able to
continue the development by enabling KVM=m in their config, until they
submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can
also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n).
This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM
are all =n.
The mmu_notifier_register call can fail because mm_take_all_locks may be
interrupted by a signal and return -EINTR. Because mmu_notifier_reigster
is used when a driver startup, a failure can be gracefully handled. Here
an example of the change applied to kvm to register the mmu notifiers.
Usually when a driver startups other allocations are required anyway and
-ENOMEM failure paths exists already.
struct kvm *kvm_arch_create_vm(void)
{
struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ int err;
if (!kvm)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
+ kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops;
+ err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm);
+ if (err) {
+ kfree(kvm);
+ return ERR_PTR(err);
+ }
+
return kvm;
}
mmu_notifier_unregister returns void and it's reliable.
The patch also adds a few needed but missing includes that would prevent
kernel to compile after these changes on non-x86 archs (x86 didn't need
them by luck).
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix mm/filemap_xip.c build]
[akpm@linux-foundation.org: fix mm/mmu_notifier.c build]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Jack Steiner <steiner@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Kanoj Sarcar <kanojsarcar@yahoo.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Avi Kivity <avi@qumranet.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Chris Wright <chrisw@redhat.com>
Cc: Marcelo Tosatti <marcelo@kvack.org>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: Izik Eidus <izike@qumranet.com>
Cc: Anthony Liguori <aliguori@us.ibm.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>
2008-07-28 22:46:29 +00:00
|
|
|
|
2020-12-15 03:05:44 +00:00
|
|
|
if (is_cow) {
|
|
|
|
raw_write_seqcount_end(&src_mm->write_protect_seq);
|
2018-12-28 08:38:09 +00:00
|
|
|
mmu_notifier_invalidate_range_end(&range);
|
2020-12-15 03:05:44 +00:00
|
|
|
}
|
mmu-notifiers: core
With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages.
There are secondary MMUs (with secondary sptes and secondary tlbs) too.
sptes in the kvm case are shadow pagetables, but when I say spte in
mmu-notifier context, I mean "secondary pte". In GRU case there's no
actual secondary pte and there's only a secondary tlb because the GRU
secondary MMU has no knowledge about sptes and every secondary tlb miss
event in the MMU always generates a page fault that has to be resolved by
the CPU (this is not the case of KVM where the a secondary tlb miss will
walk sptes in hardware and it will refill the secondary tlb transparently
to software if the corresponding spte is present). The same way
zap_page_range has to invalidate the pte before freeing the page, the spte
(and secondary tlb) must also be invalidated before any page is freed and
reused.
Currently we take a page_count pin on every page mapped by sptes, but that
means the pages can't be swapped whenever they're mapped by any spte
because they're part of the guest working set. Furthermore a spte unmap
event can immediately lead to a page to be freed when the pin is released
(so requiring the same complex and relatively slow tlb_gather smp safe
logic we have in zap_page_range and that can be avoided completely if the
spte unmap event doesn't require an unpin of the page previously mapped in
the secondary MMU).
The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know
when the VM is swapping or freeing or doing anything on the primary MMU so
that the secondary MMU code can drop sptes before the pages are freed,
avoiding all page pinning and allowing 100% reliable swapping of guest
physical address space. Furthermore it avoids the code that teardown the
mappings of the secondary MMU, to implement a logic like tlb_gather in
zap_page_range that would require many IPI to flush other cpu tlbs, for
each fixed number of spte unmapped.
To make an example: if what happens on the primary MMU is a protection
downgrade (from writeable to wrprotect) the secondary MMU mappings will be
invalidated, and the next secondary-mmu-page-fault will call
get_user_pages and trigger a do_wp_page through get_user_pages if it
called get_user_pages with write=1, and it'll re-establishing an updated
spte or secondary-tlb-mapping on the copied page. Or it will setup a
readonly spte or readonly tlb mapping if it's a guest-read, if it calls
get_user_pages with write=0. This is just an example.
This allows to map any page pointed by any pte (and in turn visible in the
primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an
full MMU with both sptes and secondary-tlb like the shadow-pagetable layer
with kvm), or a remote DMA in software like XPMEM (hence needing of
schedule in XPMEM code to send the invalidate to the remote node, while no
need to schedule in kvm/gru as it's an immediate event like invalidating
primary-mmu pte).
At least for KVM without this patch it's impossible to swap guests
reliably. And having this feature and removing the page pin allows
several other optimizations that simplify life considerably.
Dependencies:
1) mm_take_all_locks() to register the mmu notifier when the whole VM
isn't doing anything with "mm". This allows mmu notifier users to keep
track if the VM is in the middle of the invalidate_range_begin/end
critical section with an atomic counter incraese in range_begin and
decreased in range_end. No secondary MMU page fault is allowed to map
any spte or secondary tlb reference, while the VM is in the middle of
range_begin/end as any page returned by get_user_pages in that critical
section could later immediately be freed without any further
->invalidate_page notification (invalidate_range_begin/end works on
ranges and ->invalidate_page isn't called immediately before freeing
the page). To stop all page freeing and pagetable overwrites the
mmap_sem must be taken in write mode and all other anon_vma/i_mmap
locks must be taken too.
2) It'd be a waste to add branches in the VM if nobody could possibly
run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if
CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of
mmu notifiers, but this already allows to compile a KVM external module
against a kernel with mmu notifiers enabled and from the next pull from
kvm.git we'll start using them. And GRU/XPMEM will also be able to
continue the development by enabling KVM=m in their config, until they
submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can
also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n).
This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM
are all =n.
The mmu_notifier_register call can fail because mm_take_all_locks may be
interrupted by a signal and return -EINTR. Because mmu_notifier_reigster
is used when a driver startup, a failure can be gracefully handled. Here
an example of the change applied to kvm to register the mmu notifiers.
Usually when a driver startups other allocations are required anyway and
-ENOMEM failure paths exists already.
struct kvm *kvm_arch_create_vm(void)
{
struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ int err;
if (!kvm)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
+ kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops;
+ err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm);
+ if (err) {
+ kfree(kvm);
+ return ERR_PTR(err);
+ }
+
return kvm;
}
mmu_notifier_unregister returns void and it's reliable.
The patch also adds a few needed but missing includes that would prevent
kernel to compile after these changes on non-x86 archs (x86 didn't need
them by luck).
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix mm/filemap_xip.c build]
[akpm@linux-foundation.org: fix mm/mmu_notifier.c build]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Jack Steiner <steiner@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Kanoj Sarcar <kanojsarcar@yahoo.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Avi Kivity <avi@qumranet.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Chris Wright <chrisw@redhat.com>
Cc: Marcelo Tosatti <marcelo@kvack.org>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: Izik Eidus <izike@qumranet.com>
Cc: Anthony Liguori <aliguori@us.ibm.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>
2008-07-28 22:46:29 +00:00
|
|
|
return ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
mm: don't skip swap entry even if zap_details specified
Patch series "mm: Rework zap ptes on swap entries", v5.
Patch 1 should fix a long standing bug for zap_pte_range() on
zap_details usage. The risk is we could have some swap entries skipped
while we should have zapped them.
Migration entries are not the major concern because file backed memory
always zap in the pattern that "first time without page lock, then
re-zap with page lock" hence the 2nd zap will always make sure all
migration entries are already recovered.
However there can be issues with real swap entries got skipped
errornoously. There's a reproducer provided in commit message of patch
1 for that.
Patch 2-4 are cleanups that are based on patch 1. After the whole
patchset applied, we should have a very clean view of zap_pte_range().
Only patch 1 needs to be backported to stable if necessary.
This patch (of 4):
The "details" pointer shouldn't be the token to decide whether we should
skip swap entries.
For example, when the callers specified details->zap_mapping==NULL, it
means the user wants to zap all the pages (including COWed pages), then
we need to look into swap entries because there can be private COWed
pages that was swapped out.
Skipping some swap entries when details is non-NULL may lead to wrongly
leaving some of the swap entries while we should have zapped them.
A reproducer of the problem:
===8<===
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <stdio.h>
#include <assert.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/types.h>
int page_size;
int shmem_fd;
char *buffer;
void main(void)
{
int ret;
char val;
page_size = getpagesize();
shmem_fd = memfd_create("test", 0);
assert(shmem_fd >= 0);
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
buffer = mmap(NULL, page_size * 2, PROT_READ | PROT_WRITE,
MAP_PRIVATE, shmem_fd, 0);
assert(buffer != MAP_FAILED);
/* Write private page, swap it out */
buffer[page_size] = 1;
madvise(buffer, page_size * 2, MADV_PAGEOUT);
/* This should drop private buffer[page_size] already */
ret = ftruncate(shmem_fd, page_size);
assert(ret == 0);
/* Recover the size */
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
/* Re-read the data, it should be all zero */
val = buffer[page_size];
if (val == 0)
printf("Good\n");
else
printf("BUG\n");
}
===8<===
We don't need to touch up the pmd path, because pmd never had a issue with
swap entries. For example, shmem pmd migration will always be split into
pte level, and same to swapping on anonymous.
Add another helper should_zap_cows() so that we can also check whether we
should zap private mappings when there's no page pointer specified.
This patch drops that trick, so we handle swap ptes coherently. Meanwhile
we should do the same check upon migration entry, hwpoison entry and
genuine swap entries too.
To be explicit, we should still remember to keep the private entries if
even_cows==false, and always zap them when even_cows==true.
The issue seems to exist starting from the initial commit of git.
[peterx@redhat.com: comment tweaks]
Link: https://lkml.kernel.org/r/20220217060746.71256-2-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220217060746.71256-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-2-peterx@redhat.com
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:15 +00:00
|
|
|
/* Whether we should zap all COWed (private) pages too */
|
|
|
|
static inline bool should_zap_cows(struct zap_details *details)
|
|
|
|
{
|
|
|
|
/* By default, zap all pages */
|
|
|
|
if (!details)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
/* Or, we zap COWed pages only if the caller wants to */
|
mm: change zap_details.zap_mapping into even_cows
Currently we have a zap_mapping pointer maintained in zap_details, when
it is specified we only want to zap the pages that has the same mapping
with what the caller has specified.
But what we want to do is actually simpler: we want to skip zapping
private (COW-ed) pages in some cases. We can refer to
unmap_mapping_pages() callers where we could have passed in different
even_cows values. The other user is unmap_mapping_folio() where we
always want to skip private pages.
According to Hugh, we used a mapping pointer for historical reason, as
explained here:
https://lore.kernel.org/lkml/391aa58d-ce84-9d4-d68d-d98a9c533255@google.com/
Quoting partly from Hugh:
Which raises the question again of why I did not just use a boolean flag
there originally: aah, I think I've found why. In those days there was a
horrible "optimization", for better performance on some benchmark I guess,
which when you read from /dev/zero into a private mapping, would map the zero
page there (look up read_zero_pagealigned() and zeromap_page_range() if you
dare). So there was another category of page to be skipped along with the
anon COWs, and I didn't want multiple tests in the zap loop, so checking
check_mapping against page->mapping did both. I think nowadays you could do
it by checking for PageAnon page (or genuine swap entry) instead.
This patch replaces the zap_details.zap_mapping pointer into the even_cows
boolean, then we check it against PageAnon.
Link: https://lkml.kernel.org/r/20220216094810.60572-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:21 +00:00
|
|
|
return details->even_cows;
|
mm: don't skip swap entry even if zap_details specified
Patch series "mm: Rework zap ptes on swap entries", v5.
Patch 1 should fix a long standing bug for zap_pte_range() on
zap_details usage. The risk is we could have some swap entries skipped
while we should have zapped them.
Migration entries are not the major concern because file backed memory
always zap in the pattern that "first time without page lock, then
re-zap with page lock" hence the 2nd zap will always make sure all
migration entries are already recovered.
However there can be issues with real swap entries got skipped
errornoously. There's a reproducer provided in commit message of patch
1 for that.
Patch 2-4 are cleanups that are based on patch 1. After the whole
patchset applied, we should have a very clean view of zap_pte_range().
Only patch 1 needs to be backported to stable if necessary.
This patch (of 4):
The "details" pointer shouldn't be the token to decide whether we should
skip swap entries.
For example, when the callers specified details->zap_mapping==NULL, it
means the user wants to zap all the pages (including COWed pages), then
we need to look into swap entries because there can be private COWed
pages that was swapped out.
Skipping some swap entries when details is non-NULL may lead to wrongly
leaving some of the swap entries while we should have zapped them.
A reproducer of the problem:
===8<===
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <stdio.h>
#include <assert.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/types.h>
int page_size;
int shmem_fd;
char *buffer;
void main(void)
{
int ret;
char val;
page_size = getpagesize();
shmem_fd = memfd_create("test", 0);
assert(shmem_fd >= 0);
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
buffer = mmap(NULL, page_size * 2, PROT_READ | PROT_WRITE,
MAP_PRIVATE, shmem_fd, 0);
assert(buffer != MAP_FAILED);
/* Write private page, swap it out */
buffer[page_size] = 1;
madvise(buffer, page_size * 2, MADV_PAGEOUT);
/* This should drop private buffer[page_size] already */
ret = ftruncate(shmem_fd, page_size);
assert(ret == 0);
/* Recover the size */
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
/* Re-read the data, it should be all zero */
val = buffer[page_size];
if (val == 0)
printf("Good\n");
else
printf("BUG\n");
}
===8<===
We don't need to touch up the pmd path, because pmd never had a issue with
swap entries. For example, shmem pmd migration will always be split into
pte level, and same to swapping on anonymous.
Add another helper should_zap_cows() so that we can also check whether we
should zap private mappings when there's no page pointer specified.
This patch drops that trick, so we handle swap ptes coherently. Meanwhile
we should do the same check upon migration entry, hwpoison entry and
genuine swap entries too.
To be explicit, we should still remember to keep the private entries if
even_cows==false, and always zap them when even_cows==true.
The issue seems to exist starting from the initial commit of git.
[peterx@redhat.com: comment tweaks]
Link: https://lkml.kernel.org/r/20220217060746.71256-2-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220217060746.71256-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-2-peterx@redhat.com
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:15 +00:00
|
|
|
}
|
|
|
|
|
mm: change zap_details.zap_mapping into even_cows
Currently we have a zap_mapping pointer maintained in zap_details, when
it is specified we only want to zap the pages that has the same mapping
with what the caller has specified.
But what we want to do is actually simpler: we want to skip zapping
private (COW-ed) pages in some cases. We can refer to
unmap_mapping_pages() callers where we could have passed in different
even_cows values. The other user is unmap_mapping_folio() where we
always want to skip private pages.
According to Hugh, we used a mapping pointer for historical reason, as
explained here:
https://lore.kernel.org/lkml/391aa58d-ce84-9d4-d68d-d98a9c533255@google.com/
Quoting partly from Hugh:
Which raises the question again of why I did not just use a boolean flag
there originally: aah, I think I've found why. In those days there was a
horrible "optimization", for better performance on some benchmark I guess,
which when you read from /dev/zero into a private mapping, would map the zero
page there (look up read_zero_pagealigned() and zeromap_page_range() if you
dare). So there was another category of page to be skipped along with the
anon COWs, and I didn't want multiple tests in the zap loop, so checking
check_mapping against page->mapping did both. I think nowadays you could do
it by checking for PageAnon page (or genuine swap entry) instead.
This patch replaces the zap_details.zap_mapping pointer into the even_cows
boolean, then we check it against PageAnon.
Link: https://lkml.kernel.org/r/20220216094810.60572-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:21 +00:00
|
|
|
/* Decides whether we should zap this page with the page pointer specified */
|
2022-03-22 21:42:18 +00:00
|
|
|
static inline bool should_zap_page(struct zap_details *details, struct page *page)
|
2021-11-28 19:53:35 +00:00
|
|
|
{
|
mm: don't skip swap entry even if zap_details specified
Patch series "mm: Rework zap ptes on swap entries", v5.
Patch 1 should fix a long standing bug for zap_pte_range() on
zap_details usage. The risk is we could have some swap entries skipped
while we should have zapped them.
Migration entries are not the major concern because file backed memory
always zap in the pattern that "first time without page lock, then
re-zap with page lock" hence the 2nd zap will always make sure all
migration entries are already recovered.
However there can be issues with real swap entries got skipped
errornoously. There's a reproducer provided in commit message of patch
1 for that.
Patch 2-4 are cleanups that are based on patch 1. After the whole
patchset applied, we should have a very clean view of zap_pte_range().
Only patch 1 needs to be backported to stable if necessary.
This patch (of 4):
The "details" pointer shouldn't be the token to decide whether we should
skip swap entries.
For example, when the callers specified details->zap_mapping==NULL, it
means the user wants to zap all the pages (including COWed pages), then
we need to look into swap entries because there can be private COWed
pages that was swapped out.
Skipping some swap entries when details is non-NULL may lead to wrongly
leaving some of the swap entries while we should have zapped them.
A reproducer of the problem:
===8<===
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <stdio.h>
#include <assert.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/types.h>
int page_size;
int shmem_fd;
char *buffer;
void main(void)
{
int ret;
char val;
page_size = getpagesize();
shmem_fd = memfd_create("test", 0);
assert(shmem_fd >= 0);
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
buffer = mmap(NULL, page_size * 2, PROT_READ | PROT_WRITE,
MAP_PRIVATE, shmem_fd, 0);
assert(buffer != MAP_FAILED);
/* Write private page, swap it out */
buffer[page_size] = 1;
madvise(buffer, page_size * 2, MADV_PAGEOUT);
/* This should drop private buffer[page_size] already */
ret = ftruncate(shmem_fd, page_size);
assert(ret == 0);
/* Recover the size */
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
/* Re-read the data, it should be all zero */
val = buffer[page_size];
if (val == 0)
printf("Good\n");
else
printf("BUG\n");
}
===8<===
We don't need to touch up the pmd path, because pmd never had a issue with
swap entries. For example, shmem pmd migration will always be split into
pte level, and same to swapping on anonymous.
Add another helper should_zap_cows() so that we can also check whether we
should zap private mappings when there's no page pointer specified.
This patch drops that trick, so we handle swap ptes coherently. Meanwhile
we should do the same check upon migration entry, hwpoison entry and
genuine swap entries too.
To be explicit, we should still remember to keep the private entries if
even_cows==false, and always zap them when even_cows==true.
The issue seems to exist starting from the initial commit of git.
[peterx@redhat.com: comment tweaks]
Link: https://lkml.kernel.org/r/20220217060746.71256-2-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220217060746.71256-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-2-peterx@redhat.com
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:15 +00:00
|
|
|
/* If we can make a decision without *page.. */
|
|
|
|
if (should_zap_cows(details))
|
2022-03-22 21:42:18 +00:00
|
|
|
return true;
|
mm: don't skip swap entry even if zap_details specified
Patch series "mm: Rework zap ptes on swap entries", v5.
Patch 1 should fix a long standing bug for zap_pte_range() on
zap_details usage. The risk is we could have some swap entries skipped
while we should have zapped them.
Migration entries are not the major concern because file backed memory
always zap in the pattern that "first time without page lock, then
re-zap with page lock" hence the 2nd zap will always make sure all
migration entries are already recovered.
However there can be issues with real swap entries got skipped
errornoously. There's a reproducer provided in commit message of patch
1 for that.
Patch 2-4 are cleanups that are based on patch 1. After the whole
patchset applied, we should have a very clean view of zap_pte_range().
Only patch 1 needs to be backported to stable if necessary.
This patch (of 4):
The "details" pointer shouldn't be the token to decide whether we should
skip swap entries.
For example, when the callers specified details->zap_mapping==NULL, it
means the user wants to zap all the pages (including COWed pages), then
we need to look into swap entries because there can be private COWed
pages that was swapped out.
Skipping some swap entries when details is non-NULL may lead to wrongly
leaving some of the swap entries while we should have zapped them.
A reproducer of the problem:
===8<===
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <stdio.h>
#include <assert.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/types.h>
int page_size;
int shmem_fd;
char *buffer;
void main(void)
{
int ret;
char val;
page_size = getpagesize();
shmem_fd = memfd_create("test", 0);
assert(shmem_fd >= 0);
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
buffer = mmap(NULL, page_size * 2, PROT_READ | PROT_WRITE,
MAP_PRIVATE, shmem_fd, 0);
assert(buffer != MAP_FAILED);
/* Write private page, swap it out */
buffer[page_size] = 1;
madvise(buffer, page_size * 2, MADV_PAGEOUT);
/* This should drop private buffer[page_size] already */
ret = ftruncate(shmem_fd, page_size);
assert(ret == 0);
/* Recover the size */
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
/* Re-read the data, it should be all zero */
val = buffer[page_size];
if (val == 0)
printf("Good\n");
else
printf("BUG\n");
}
===8<===
We don't need to touch up the pmd path, because pmd never had a issue with
swap entries. For example, shmem pmd migration will always be split into
pte level, and same to swapping on anonymous.
Add another helper should_zap_cows() so that we can also check whether we
should zap private mappings when there's no page pointer specified.
This patch drops that trick, so we handle swap ptes coherently. Meanwhile
we should do the same check upon migration entry, hwpoison entry and
genuine swap entries too.
To be explicit, we should still remember to keep the private entries if
even_cows==false, and always zap them when even_cows==true.
The issue seems to exist starting from the initial commit of git.
[peterx@redhat.com: comment tweaks]
Link: https://lkml.kernel.org/r/20220217060746.71256-2-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220217060746.71256-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-2-peterx@redhat.com
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:15 +00:00
|
|
|
|
|
|
|
/* E.g. the caller passes NULL for the case of a zero page */
|
|
|
|
if (!page)
|
2022-03-22 21:42:18 +00:00
|
|
|
return true;
|
2021-11-28 19:53:35 +00:00
|
|
|
|
mm: change zap_details.zap_mapping into even_cows
Currently we have a zap_mapping pointer maintained in zap_details, when
it is specified we only want to zap the pages that has the same mapping
with what the caller has specified.
But what we want to do is actually simpler: we want to skip zapping
private (COW-ed) pages in some cases. We can refer to
unmap_mapping_pages() callers where we could have passed in different
even_cows values. The other user is unmap_mapping_folio() where we
always want to skip private pages.
According to Hugh, we used a mapping pointer for historical reason, as
explained here:
https://lore.kernel.org/lkml/391aa58d-ce84-9d4-d68d-d98a9c533255@google.com/
Quoting partly from Hugh:
Which raises the question again of why I did not just use a boolean flag
there originally: aah, I think I've found why. In those days there was a
horrible "optimization", for better performance on some benchmark I guess,
which when you read from /dev/zero into a private mapping, would map the zero
page there (look up read_zero_pagealigned() and zeromap_page_range() if you
dare). So there was another category of page to be skipped along with the
anon COWs, and I didn't want multiple tests in the zap loop, so checking
check_mapping against page->mapping did both. I think nowadays you could do
it by checking for PageAnon page (or genuine swap entry) instead.
This patch replaces the zap_details.zap_mapping pointer into the even_cows
boolean, then we check it against PageAnon.
Link: https://lkml.kernel.org/r/20220216094810.60572-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:21 +00:00
|
|
|
/* Otherwise we should only zap non-anon pages */
|
|
|
|
return !PageAnon(page);
|
2021-11-28 19:53:35 +00:00
|
|
|
}
|
|
|
|
|
2022-05-13 03:22:53 +00:00
|
|
|
static inline bool zap_drop_file_uffd_wp(struct zap_details *details)
|
|
|
|
{
|
|
|
|
if (!details)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return details->zap_flags & ZAP_FLAG_DROP_MARKER;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This function makes sure that we'll replace the none pte with an uffd-wp
|
|
|
|
* swap special pte marker when necessary. Must be with the pgtable lock held.
|
|
|
|
*/
|
|
|
|
static inline void
|
|
|
|
zap_install_uffd_wp_if_needed(struct vm_area_struct *vma,
|
|
|
|
unsigned long addr, pte_t *pte,
|
|
|
|
struct zap_details *details, pte_t pteval)
|
|
|
|
{
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
/* Zap on anonymous always means dropping everything */
|
|
|
|
if (vma_is_anonymous(vma))
|
|
|
|
return;
|
|
|
|
|
2022-05-13 03:22:53 +00:00
|
|
|
if (zap_drop_file_uffd_wp(details))
|
|
|
|
return;
|
|
|
|
|
|
|
|
pte_install_uffd_wp_if_needed(vma, addr, pte, pteval);
|
|
|
|
}
|
|
|
|
|
2005-11-14 00:06:42 +00:00
|
|
|
static unsigned long zap_pte_range(struct mmu_gather *tlb,
|
2005-10-30 01:16:12 +00:00
|
|
|
struct vm_area_struct *vma, pmd_t *pmd,
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long addr, unsigned long end,
|
2011-05-25 00:12:04 +00:00
|
|
|
struct zap_details *details)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2005-10-30 01:16:12 +00:00
|
|
|
struct mm_struct *mm = tlb->mm;
|
2011-05-25 00:11:45 +00:00
|
|
|
int force_flush = 0;
|
2010-03-05 21:41:39 +00:00
|
|
|
int rss[NR_MM_COUNTERS];
|
2011-05-25 00:12:04 +00:00
|
|
|
spinlock_t *ptl;
|
mm: fix wrong kunmap_atomic() pointer
Running a ktest.pl test, I hit the following bug on x86_32:
------------[ cut here ]------------
WARNING: at arch/x86/mm/highmem_32.c:81 __kunmap_atomic+0x64/0xc1()
Hardware name:
Modules linked in:
Pid: 93, comm: sh Not tainted 2.6.39-test+ #1
Call Trace:
[<c04450da>] warn_slowpath_common+0x7c/0x91
[<c042f5df>] ? __kunmap_atomic+0x64/0xc1
[<c042f5df>] ? __kunmap_atomic+0x64/0xc1^M
[<c0445111>] warn_slowpath_null+0x22/0x24
[<c042f5df>] __kunmap_atomic+0x64/0xc1
[<c04d4a22>] unmap_vmas+0x43a/0x4e0
[<c04d9065>] exit_mmap+0x91/0xd2
[<c0443057>] mmput+0x43/0xad
[<c0448358>] exit_mm+0x111/0x119
[<c044855f>] do_exit+0x1ff/0x5fa
[<c0454ea2>] ? set_current_blocked+0x3c/0x40
[<c0454f24>] ? sigprocmask+0x7e/0x8e
[<c0448b55>] do_group_exit+0x65/0x88
[<c0448b90>] sys_exit_group+0x18/0x1c
[<c0c3915f>] sysenter_do_call+0x12/0x38
---[ end trace 8055f74ea3c0eb62 ]---
Running a ktest.pl git bisect, found the culprit: commit e303297e6c3a
("mm: extended batches for generic mmu_gather")
But although this was the commit triggering the bug, it was not the one
originally responsible for the bug. That was commit d16dfc550f53 ("mm:
mmu_gather rework").
The code in zap_pte_range() has something that looks like the following:
pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
do {
[...]
} while (pte++, addr += PAGE_SIZE, addr != end);
pte_unmap_unlock(pte - 1, ptl);
The pte starts off pointing at the first element in the page table
directory that was returned by the pte_offset_map_lock(). When it's done
with the page, pte will be pointing to anything between the next entry and
the first entry of the next page inclusive. By doing a pte - 1, this puts
the pte back onto the original page, which is all that pte_unmap_unlock()
needs.
In most archs (64 bit), this is not an issue as the pte is ignored in the
pte_unmap_unlock(). But on 32 bit archs, where things may be kmapped, it
is essential that the pte passed to pte_unmap_unlock() resides on the same
page that was given by pte_offest_map_lock().
The problem came in d16dfc55 ("mm: mmu_gather rework") where it introduced
a "break;" from the while loop. This alone did not seem to easily trigger
the bug. But the modifications made by e303297e6 caused that "break;" to
be hit on the first iteration, before the pte++.
The pte not being incremented will now cause pte_unmap_unlock(pte - 1) to
be pointing to the previous page. This will cause the wrong page to be
unmapped, and also trigger the warning above.
The simple solution is to just save the pointer given by
pte_offset_map_lock() and use it in the unlock.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-06-15 22:08:23 +00:00
|
|
|
pte_t *start_pte;
|
2011-05-25 00:12:04 +00:00
|
|
|
pte_t *pte;
|
2015-02-10 22:09:49 +00:00
|
|
|
swp_entry_t entry;
|
2010-03-05 21:41:39 +00:00
|
|
|
|
2018-08-31 12:46:08 +00:00
|
|
|
tlb_change_page_size(tlb, PAGE_SIZE);
|
2011-05-25 00:11:45 +00:00
|
|
|
again:
|
2011-05-25 00:12:01 +00:00
|
|
|
init_rss_vec(rss);
|
mm: fix wrong kunmap_atomic() pointer
Running a ktest.pl test, I hit the following bug on x86_32:
------------[ cut here ]------------
WARNING: at arch/x86/mm/highmem_32.c:81 __kunmap_atomic+0x64/0xc1()
Hardware name:
Modules linked in:
Pid: 93, comm: sh Not tainted 2.6.39-test+ #1
Call Trace:
[<c04450da>] warn_slowpath_common+0x7c/0x91
[<c042f5df>] ? __kunmap_atomic+0x64/0xc1
[<c042f5df>] ? __kunmap_atomic+0x64/0xc1^M
[<c0445111>] warn_slowpath_null+0x22/0x24
[<c042f5df>] __kunmap_atomic+0x64/0xc1
[<c04d4a22>] unmap_vmas+0x43a/0x4e0
[<c04d9065>] exit_mmap+0x91/0xd2
[<c0443057>] mmput+0x43/0xad
[<c0448358>] exit_mm+0x111/0x119
[<c044855f>] do_exit+0x1ff/0x5fa
[<c0454ea2>] ? set_current_blocked+0x3c/0x40
[<c0454f24>] ? sigprocmask+0x7e/0x8e
[<c0448b55>] do_group_exit+0x65/0x88
[<c0448b90>] sys_exit_group+0x18/0x1c
[<c0c3915f>] sysenter_do_call+0x12/0x38
---[ end trace 8055f74ea3c0eb62 ]---
Running a ktest.pl git bisect, found the culprit: commit e303297e6c3a
("mm: extended batches for generic mmu_gather")
But although this was the commit triggering the bug, it was not the one
originally responsible for the bug. That was commit d16dfc550f53 ("mm:
mmu_gather rework").
The code in zap_pte_range() has something that looks like the following:
pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
do {
[...]
} while (pte++, addr += PAGE_SIZE, addr != end);
pte_unmap_unlock(pte - 1, ptl);
The pte starts off pointing at the first element in the page table
directory that was returned by the pte_offset_map_lock(). When it's done
with the page, pte will be pointing to anything between the next entry and
the first entry of the next page inclusive. By doing a pte - 1, this puts
the pte back onto the original page, which is all that pte_unmap_unlock()
needs.
In most archs (64 bit), this is not an issue as the pte is ignored in the
pte_unmap_unlock(). But on 32 bit archs, where things may be kmapped, it
is essential that the pte passed to pte_unmap_unlock() resides on the same
page that was given by pte_offest_map_lock().
The problem came in d16dfc55 ("mm: mmu_gather rework") where it introduced
a "break;" from the while loop. This alone did not seem to easily trigger
the bug. But the modifications made by e303297e6 caused that "break;" to
be hit on the first iteration, before the pte++.
The pte not being incremented will now cause pte_unmap_unlock(pte - 1) to
be pointing to the previous page. This will cause the wrong page to be
unmapped, and also trigger the warning above.
The simple solution is to just save the pointer given by
pte_offset_map_lock() and use it in the unlock.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-06-15 22:08:23 +00:00
|
|
|
start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
|
|
|
|
pte = start_pte;
|
2017-08-02 20:31:52 +00:00
|
|
|
flush_tlb_batched_pending(mm);
|
2006-10-01 06:29:33 +00:00
|
|
|
arch_enter_lazy_mmu_mode();
|
2005-04-16 22:20:36 +00:00
|
|
|
do {
|
|
|
|
pte_t ptent = *pte;
|
2022-03-22 21:42:24 +00:00
|
|
|
struct page *page;
|
|
|
|
|
2017-02-24 22:59:01 +00:00
|
|
|
if (pte_none(ptent))
|
2005-04-16 22:20:36 +00:00
|
|
|
continue;
|
2006-03-17 07:04:09 +00:00
|
|
|
|
2019-09-24 00:02:24 +00:00
|
|
|
if (need_resched())
|
|
|
|
break;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (pte_present(ptent)) {
|
mm: delay page_remove_rmap() until after the TLB has been flushed
When we remove a page table entry, we are very careful to only free the
page after we have flushed the TLB, because other CPUs could still be
using the page through stale TLB entries until after the flush.
However, we have removed the rmap entry for that page early, which means
that functions like folio_mkclean() would end up not serializing with the
page table lock because the page had already been made invisible to rmap.
And that is a problem, because while the TLB entry exists, we could end up
with the following situation:
(a) one CPU could come in and clean it, never seeing our mapping of the
page
(b) another CPU could continue to use the stale and dirty TLB entry and
continue to write to said page
resulting in a page that has been dirtied, but then marked clean again,
all while another CPU might have dirtied it some more.
End result: possibly lost dirty data.
This extends our current TLB gather infrastructure to optionally track a
"should I do a delayed page_remove_rmap() for this page after flushing the
TLB". It uses the newly introduced 'encoded page pointer' to do that
without having to keep separate data around.
Note, this is complicated by a couple of issues:
- we want to delay the rmap removal, but not past the page table lock,
because that simplifies the memcg accounting
- only SMP configurations want to delay TLB flushing, since on UP
there are obviously no remote TLBs to worry about, and the page
table lock means there are no preemption issues either
- s390 has its own mmu_gather model that doesn't delay TLB flushing,
and as a result also does not want the delayed rmap. As such, we can
treat S390 like the UP case and use a common fallback for the "no
delays" case.
- we can track an enormous number of pages in our mmu_gather structure,
with MAX_GATHER_BATCH_COUNT batches of MAX_TABLE_BATCH pages each,
all set up to be approximately 10k pending pages.
We do not want to have a huge number of batched pages that we then
need to check for delayed rmap handling inside the page table lock.
Particularly that last point results in a noteworthy detail, where the
normal page batch gathering is limited once we have delayed rmaps pending,
in such a way that only the last batch (the so-called "active batch") in
the mmu_gather structure can have any delayed entries.
NOTE! While the "possibly lost dirty data" sounds catastrophic, for this
all to happen you need to have a user thread doing either madvise() with
MADV_DONTNEED or a full re-mmap() of the area concurrently with another
thread continuing to use said mapping.
So arguably this is about user space doing crazy things, but from a VM
consistency standpoint it's better if we track the dirty bit properly even
when user space goes off the rails.
[akpm@linux-foundation.org: fix UP build, per Linus]
Link: https://lore.kernel.org/all/B88D3073-440A-41C7-95F4-895D3F657EF2@gmail.com/
Link: https://lkml.kernel.org/r/20221109203051.1835763-4-torvalds@linux-foundation.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Tested-by: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-09 20:30:51 +00:00
|
|
|
unsigned int delay_rmap;
|
|
|
|
|
2019-06-13 20:50:49 +00:00
|
|
|
page = vm_normal_page(vma, addr, ptent);
|
2022-03-22 21:42:18 +00:00
|
|
|
if (unlikely(!should_zap_page(details, page)))
|
2021-11-05 20:38:34 +00:00
|
|
|
continue;
|
2005-10-30 01:16:12 +00:00
|
|
|
ptent = ptep_get_and_clear_full(mm, addr, pte,
|
[PATCH] x86: ptep_clear optimization
Add a new accessor for PTEs, which passes the full hint from the mmu_gather
struct; this allows architectures with hardware pagetables to optimize away
atomic PTE operations when destroying an address space. Removing the
locked operation should allow better pipelining of memory access in this
loop. I measured an average savings of 30-35 cycles per zap_pte_range on
the first 500 destructions on Pentium-M, but I believe the optimization
would win more on older processors which still assert the bus lock on xchg
for an exclusive cacheline.
Update: I made some new measurements, and this saves exactly 26 cycles over
ptep_get_and_clear on Pentium M. On P4, with a PAE kernel, this saves 180
cycles per ptep_get_and_clear, for a whopping 92160 cycles savings for a
full address space destruction.
pte_clear_full is not yet used, but is provided for future optimizations
(in particular, when running inside of a hypervisor that queues page table
updates, the full hint allows us to avoid queueing unnecessary page table
update for an address space in the process of being destroyed.
This is not a huge win, but it does help a bit, and sets the stage for
further hypervisor optimization of the mm layer on all architectures.
Signed-off-by: Zachary Amsden <zach@vmware.com>
Cc: Christoph Lameter <christoph@lameter.com>
Cc: <linux-mm@kvack.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-09-03 22:55:04 +00:00
|
|
|
tlb->fullmm);
|
2005-04-16 22:20:36 +00:00
|
|
|
tlb_remove_tlb_entry(tlb, pte, addr);
|
2022-05-13 03:22:53 +00:00
|
|
|
zap_install_uffd_wp_if_needed(vma, addr, pte, details,
|
|
|
|
ptent);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (unlikely(!page))
|
|
|
|
continue;
|
2016-01-14 23:19:26 +00:00
|
|
|
|
mm: delay page_remove_rmap() until after the TLB has been flushed
When we remove a page table entry, we are very careful to only free the
page after we have flushed the TLB, because other CPUs could still be
using the page through stale TLB entries until after the flush.
However, we have removed the rmap entry for that page early, which means
that functions like folio_mkclean() would end up not serializing with the
page table lock because the page had already been made invisible to rmap.
And that is a problem, because while the TLB entry exists, we could end up
with the following situation:
(a) one CPU could come in and clean it, never seeing our mapping of the
page
(b) another CPU could continue to use the stale and dirty TLB entry and
continue to write to said page
resulting in a page that has been dirtied, but then marked clean again,
all while another CPU might have dirtied it some more.
End result: possibly lost dirty data.
This extends our current TLB gather infrastructure to optionally track a
"should I do a delayed page_remove_rmap() for this page after flushing the
TLB". It uses the newly introduced 'encoded page pointer' to do that
without having to keep separate data around.
Note, this is complicated by a couple of issues:
- we want to delay the rmap removal, but not past the page table lock,
because that simplifies the memcg accounting
- only SMP configurations want to delay TLB flushing, since on UP
there are obviously no remote TLBs to worry about, and the page
table lock means there are no preemption issues either
- s390 has its own mmu_gather model that doesn't delay TLB flushing,
and as a result also does not want the delayed rmap. As such, we can
treat S390 like the UP case and use a common fallback for the "no
delays" case.
- we can track an enormous number of pages in our mmu_gather structure,
with MAX_GATHER_BATCH_COUNT batches of MAX_TABLE_BATCH pages each,
all set up to be approximately 10k pending pages.
We do not want to have a huge number of batched pages that we then
need to check for delayed rmap handling inside the page table lock.
Particularly that last point results in a noteworthy detail, where the
normal page batch gathering is limited once we have delayed rmaps pending,
in such a way that only the last batch (the so-called "active batch") in
the mmu_gather structure can have any delayed entries.
NOTE! While the "possibly lost dirty data" sounds catastrophic, for this
all to happen you need to have a user thread doing either madvise() with
MADV_DONTNEED or a full re-mmap() of the area concurrently with another
thread continuing to use said mapping.
So arguably this is about user space doing crazy things, but from a VM
consistency standpoint it's better if we track the dirty bit properly even
when user space goes off the rails.
[akpm@linux-foundation.org: fix UP build, per Linus]
Link: https://lore.kernel.org/all/B88D3073-440A-41C7-95F4-895D3F657EF2@gmail.com/
Link: https://lkml.kernel.org/r/20221109203051.1835763-4-torvalds@linux-foundation.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Tested-by: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-09 20:30:51 +00:00
|
|
|
delay_rmap = 0;
|
2016-01-14 23:19:26 +00:00
|
|
|
if (!PageAnon(page)) {
|
2014-04-25 23:05:40 +00:00
|
|
|
if (pte_dirty(ptent)) {
|
2005-10-30 01:15:54 +00:00
|
|
|
set_page_dirty(page);
|
mm: delay page_remove_rmap() until after the TLB has been flushed
When we remove a page table entry, we are very careful to only free the
page after we have flushed the TLB, because other CPUs could still be
using the page through stale TLB entries until after the flush.
However, we have removed the rmap entry for that page early, which means
that functions like folio_mkclean() would end up not serializing with the
page table lock because the page had already been made invisible to rmap.
And that is a problem, because while the TLB entry exists, we could end up
with the following situation:
(a) one CPU could come in and clean it, never seeing our mapping of the
page
(b) another CPU could continue to use the stale and dirty TLB entry and
continue to write to said page
resulting in a page that has been dirtied, but then marked clean again,
all while another CPU might have dirtied it some more.
End result: possibly lost dirty data.
This extends our current TLB gather infrastructure to optionally track a
"should I do a delayed page_remove_rmap() for this page after flushing the
TLB". It uses the newly introduced 'encoded page pointer' to do that
without having to keep separate data around.
Note, this is complicated by a couple of issues:
- we want to delay the rmap removal, but not past the page table lock,
because that simplifies the memcg accounting
- only SMP configurations want to delay TLB flushing, since on UP
there are obviously no remote TLBs to worry about, and the page
table lock means there are no preemption issues either
- s390 has its own mmu_gather model that doesn't delay TLB flushing,
and as a result also does not want the delayed rmap. As such, we can
treat S390 like the UP case and use a common fallback for the "no
delays" case.
- we can track an enormous number of pages in our mmu_gather structure,
with MAX_GATHER_BATCH_COUNT batches of MAX_TABLE_BATCH pages each,
all set up to be approximately 10k pending pages.
We do not want to have a huge number of batched pages that we then
need to check for delayed rmap handling inside the page table lock.
Particularly that last point results in a noteworthy detail, where the
normal page batch gathering is limited once we have delayed rmaps pending,
in such a way that only the last batch (the so-called "active batch") in
the mmu_gather structure can have any delayed entries.
NOTE! While the "possibly lost dirty data" sounds catastrophic, for this
all to happen you need to have a user thread doing either madvise() with
MADV_DONTNEED or a full re-mmap() of the area concurrently with another
thread continuing to use said mapping.
So arguably this is about user space doing crazy things, but from a VM
consistency standpoint it's better if we track the dirty bit properly even
when user space goes off the rails.
[akpm@linux-foundation.org: fix UP build, per Linus]
Link: https://lore.kernel.org/all/B88D3073-440A-41C7-95F4-895D3F657EF2@gmail.com/
Link: https://lkml.kernel.org/r/20221109203051.1835763-4-torvalds@linux-foundation.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Tested-by: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-09 20:30:51 +00:00
|
|
|
if (tlb_delay_rmap(tlb)) {
|
|
|
|
delay_rmap = 1;
|
|
|
|
force_flush = 1;
|
|
|
|
}
|
2014-04-25 23:05:40 +00:00
|
|
|
}
|
2022-12-30 21:52:51 +00:00
|
|
|
if (pte_young(ptent) && likely(vma_has_recency(vma)))
|
2009-01-06 22:38:55 +00:00
|
|
|
mark_page_accessed(page);
|
2005-10-30 01:15:54 +00:00
|
|
|
}
|
2016-01-14 23:19:26 +00:00
|
|
|
rss[mm_counter(page)]--;
|
mm: delay page_remove_rmap() until after the TLB has been flushed
When we remove a page table entry, we are very careful to only free the
page after we have flushed the TLB, because other CPUs could still be
using the page through stale TLB entries until after the flush.
However, we have removed the rmap entry for that page early, which means
that functions like folio_mkclean() would end up not serializing with the
page table lock because the page had already been made invisible to rmap.
And that is a problem, because while the TLB entry exists, we could end up
with the following situation:
(a) one CPU could come in and clean it, never seeing our mapping of the
page
(b) another CPU could continue to use the stale and dirty TLB entry and
continue to write to said page
resulting in a page that has been dirtied, but then marked clean again,
all while another CPU might have dirtied it some more.
End result: possibly lost dirty data.
This extends our current TLB gather infrastructure to optionally track a
"should I do a delayed page_remove_rmap() for this page after flushing the
TLB". It uses the newly introduced 'encoded page pointer' to do that
without having to keep separate data around.
Note, this is complicated by a couple of issues:
- we want to delay the rmap removal, but not past the page table lock,
because that simplifies the memcg accounting
- only SMP configurations want to delay TLB flushing, since on UP
there are obviously no remote TLBs to worry about, and the page
table lock means there are no preemption issues either
- s390 has its own mmu_gather model that doesn't delay TLB flushing,
and as a result also does not want the delayed rmap. As such, we can
treat S390 like the UP case and use a common fallback for the "no
delays" case.
- we can track an enormous number of pages in our mmu_gather structure,
with MAX_GATHER_BATCH_COUNT batches of MAX_TABLE_BATCH pages each,
all set up to be approximately 10k pending pages.
We do not want to have a huge number of batched pages that we then
need to check for delayed rmap handling inside the page table lock.
Particularly that last point results in a noteworthy detail, where the
normal page batch gathering is limited once we have delayed rmaps pending,
in such a way that only the last batch (the so-called "active batch") in
the mmu_gather structure can have any delayed entries.
NOTE! While the "possibly lost dirty data" sounds catastrophic, for this
all to happen you need to have a user thread doing either madvise() with
MADV_DONTNEED or a full re-mmap() of the area concurrently with another
thread continuing to use said mapping.
So arguably this is about user space doing crazy things, but from a VM
consistency standpoint it's better if we track the dirty bit properly even
when user space goes off the rails.
[akpm@linux-foundation.org: fix UP build, per Linus]
Link: https://lore.kernel.org/all/B88D3073-440A-41C7-95F4-895D3F657EF2@gmail.com/
Link: https://lkml.kernel.org/r/20221109203051.1835763-4-torvalds@linux-foundation.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Tested-by: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-09 20:30:51 +00:00
|
|
|
if (!delay_rmap) {
|
|
|
|
page_remove_rmap(page, vma, false);
|
|
|
|
if (unlikely(page_mapcount(page) < 0))
|
|
|
|
print_bad_pte(vma, addr, ptent, page);
|
|
|
|
}
|
|
|
|
if (unlikely(__tlb_remove_page(tlb, page, delay_rmap))) {
|
2014-04-25 23:05:40 +00:00
|
|
|
force_flush = 1;
|
2014-10-28 20:16:28 +00:00
|
|
|
addr += PAGE_SIZE;
|
2011-05-25 00:11:45 +00:00
|
|
|
break;
|
2014-04-25 23:05:40 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
continue;
|
|
|
|
}
|
2017-09-08 23:11:43 +00:00
|
|
|
|
|
|
|
entry = pte_to_swp_entry(ptent);
|
2021-07-01 01:54:25 +00:00
|
|
|
if (is_device_private_entry(entry) ||
|
|
|
|
is_device_exclusive_entry(entry)) {
|
2022-03-22 21:42:24 +00:00
|
|
|
page = pfn_swap_entry_to_page(entry);
|
2022-03-22 21:42:18 +00:00
|
|
|
if (unlikely(!should_zap_page(details, page)))
|
2021-11-05 20:38:34 +00:00
|
|
|
continue;
|
2022-05-13 03:22:53 +00:00
|
|
|
/*
|
|
|
|
* Both device private/exclusive mappings should only
|
|
|
|
* work with anonymous page so far, so we don't need to
|
|
|
|
* consider uffd-wp bit when zap. For more information,
|
|
|
|
* see zap_install_uffd_wp_if_needed().
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(!vma_is_anonymous(vma));
|
2017-09-08 23:11:43 +00:00
|
|
|
rss[mm_counter(page)]--;
|
2021-07-01 01:54:25 +00:00
|
|
|
if (is_device_private_entry(entry))
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
page_remove_rmap(page, vma, false);
|
2017-09-08 23:11:43 +00:00
|
|
|
put_page(page);
|
2022-03-22 21:42:24 +00:00
|
|
|
} else if (!non_swap_entry(entry)) {
|
mm: don't skip swap entry even if zap_details specified
Patch series "mm: Rework zap ptes on swap entries", v5.
Patch 1 should fix a long standing bug for zap_pte_range() on
zap_details usage. The risk is we could have some swap entries skipped
while we should have zapped them.
Migration entries are not the major concern because file backed memory
always zap in the pattern that "first time without page lock, then
re-zap with page lock" hence the 2nd zap will always make sure all
migration entries are already recovered.
However there can be issues with real swap entries got skipped
errornoously. There's a reproducer provided in commit message of patch
1 for that.
Patch 2-4 are cleanups that are based on patch 1. After the whole
patchset applied, we should have a very clean view of zap_pte_range().
Only patch 1 needs to be backported to stable if necessary.
This patch (of 4):
The "details" pointer shouldn't be the token to decide whether we should
skip swap entries.
For example, when the callers specified details->zap_mapping==NULL, it
means the user wants to zap all the pages (including COWed pages), then
we need to look into swap entries because there can be private COWed
pages that was swapped out.
Skipping some swap entries when details is non-NULL may lead to wrongly
leaving some of the swap entries while we should have zapped them.
A reproducer of the problem:
===8<===
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <stdio.h>
#include <assert.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/types.h>
int page_size;
int shmem_fd;
char *buffer;
void main(void)
{
int ret;
char val;
page_size = getpagesize();
shmem_fd = memfd_create("test", 0);
assert(shmem_fd >= 0);
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
buffer = mmap(NULL, page_size * 2, PROT_READ | PROT_WRITE,
MAP_PRIVATE, shmem_fd, 0);
assert(buffer != MAP_FAILED);
/* Write private page, swap it out */
buffer[page_size] = 1;
madvise(buffer, page_size * 2, MADV_PAGEOUT);
/* This should drop private buffer[page_size] already */
ret = ftruncate(shmem_fd, page_size);
assert(ret == 0);
/* Recover the size */
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
/* Re-read the data, it should be all zero */
val = buffer[page_size];
if (val == 0)
printf("Good\n");
else
printf("BUG\n");
}
===8<===
We don't need to touch up the pmd path, because pmd never had a issue with
swap entries. For example, shmem pmd migration will always be split into
pte level, and same to swapping on anonymous.
Add another helper should_zap_cows() so that we can also check whether we
should zap private mappings when there's no page pointer specified.
This patch drops that trick, so we handle swap ptes coherently. Meanwhile
we should do the same check upon migration entry, hwpoison entry and
genuine swap entries too.
To be explicit, we should still remember to keep the private entries if
even_cows==false, and always zap them when even_cows==true.
The issue seems to exist starting from the initial commit of git.
[peterx@redhat.com: comment tweaks]
Link: https://lkml.kernel.org/r/20220217060746.71256-2-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220217060746.71256-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-2-peterx@redhat.com
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:15 +00:00
|
|
|
/* Genuine swap entry, hence a private anon page */
|
|
|
|
if (!should_zap_cows(details))
|
|
|
|
continue;
|
2015-02-10 22:09:49 +00:00
|
|
|
rss[MM_SWAPENTS]--;
|
2022-03-22 21:42:24 +00:00
|
|
|
if (unlikely(!free_swap_and_cache(entry)))
|
|
|
|
print_bad_pte(vma, addr, ptent, NULL);
|
mm: don't skip swap entry even if zap_details specified
Patch series "mm: Rework zap ptes on swap entries", v5.
Patch 1 should fix a long standing bug for zap_pte_range() on
zap_details usage. The risk is we could have some swap entries skipped
while we should have zapped them.
Migration entries are not the major concern because file backed memory
always zap in the pattern that "first time without page lock, then
re-zap with page lock" hence the 2nd zap will always make sure all
migration entries are already recovered.
However there can be issues with real swap entries got skipped
errornoously. There's a reproducer provided in commit message of patch
1 for that.
Patch 2-4 are cleanups that are based on patch 1. After the whole
patchset applied, we should have a very clean view of zap_pte_range().
Only patch 1 needs to be backported to stable if necessary.
This patch (of 4):
The "details" pointer shouldn't be the token to decide whether we should
skip swap entries.
For example, when the callers specified details->zap_mapping==NULL, it
means the user wants to zap all the pages (including COWed pages), then
we need to look into swap entries because there can be private COWed
pages that was swapped out.
Skipping some swap entries when details is non-NULL may lead to wrongly
leaving some of the swap entries while we should have zapped them.
A reproducer of the problem:
===8<===
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <stdio.h>
#include <assert.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/types.h>
int page_size;
int shmem_fd;
char *buffer;
void main(void)
{
int ret;
char val;
page_size = getpagesize();
shmem_fd = memfd_create("test", 0);
assert(shmem_fd >= 0);
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
buffer = mmap(NULL, page_size * 2, PROT_READ | PROT_WRITE,
MAP_PRIVATE, shmem_fd, 0);
assert(buffer != MAP_FAILED);
/* Write private page, swap it out */
buffer[page_size] = 1;
madvise(buffer, page_size * 2, MADV_PAGEOUT);
/* This should drop private buffer[page_size] already */
ret = ftruncate(shmem_fd, page_size);
assert(ret == 0);
/* Recover the size */
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
/* Re-read the data, it should be all zero */
val = buffer[page_size];
if (val == 0)
printf("Good\n");
else
printf("BUG\n");
}
===8<===
We don't need to touch up the pmd path, because pmd never had a issue with
swap entries. For example, shmem pmd migration will always be split into
pte level, and same to swapping on anonymous.
Add another helper should_zap_cows() so that we can also check whether we
should zap private mappings when there's no page pointer specified.
This patch drops that trick, so we handle swap ptes coherently. Meanwhile
we should do the same check upon migration entry, hwpoison entry and
genuine swap entries too.
To be explicit, we should still remember to keep the private entries if
even_cows==false, and always zap them when even_cows==true.
The issue seems to exist starting from the initial commit of git.
[peterx@redhat.com: comment tweaks]
Link: https://lkml.kernel.org/r/20220217060746.71256-2-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220217060746.71256-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-2-peterx@redhat.com
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:15 +00:00
|
|
|
} else if (is_migration_entry(entry)) {
|
2021-07-01 01:54:06 +00:00
|
|
|
page = pfn_swap_entry_to_page(entry);
|
2022-03-22 21:42:18 +00:00
|
|
|
if (!should_zap_page(details, page))
|
mm: don't skip swap entry even if zap_details specified
Patch series "mm: Rework zap ptes on swap entries", v5.
Patch 1 should fix a long standing bug for zap_pte_range() on
zap_details usage. The risk is we could have some swap entries skipped
while we should have zapped them.
Migration entries are not the major concern because file backed memory
always zap in the pattern that "first time without page lock, then
re-zap with page lock" hence the 2nd zap will always make sure all
migration entries are already recovered.
However there can be issues with real swap entries got skipped
errornoously. There's a reproducer provided in commit message of patch
1 for that.
Patch 2-4 are cleanups that are based on patch 1. After the whole
patchset applied, we should have a very clean view of zap_pte_range().
Only patch 1 needs to be backported to stable if necessary.
This patch (of 4):
The "details" pointer shouldn't be the token to decide whether we should
skip swap entries.
For example, when the callers specified details->zap_mapping==NULL, it
means the user wants to zap all the pages (including COWed pages), then
we need to look into swap entries because there can be private COWed
pages that was swapped out.
Skipping some swap entries when details is non-NULL may lead to wrongly
leaving some of the swap entries while we should have zapped them.
A reproducer of the problem:
===8<===
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <stdio.h>
#include <assert.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/types.h>
int page_size;
int shmem_fd;
char *buffer;
void main(void)
{
int ret;
char val;
page_size = getpagesize();
shmem_fd = memfd_create("test", 0);
assert(shmem_fd >= 0);
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
buffer = mmap(NULL, page_size * 2, PROT_READ | PROT_WRITE,
MAP_PRIVATE, shmem_fd, 0);
assert(buffer != MAP_FAILED);
/* Write private page, swap it out */
buffer[page_size] = 1;
madvise(buffer, page_size * 2, MADV_PAGEOUT);
/* This should drop private buffer[page_size] already */
ret = ftruncate(shmem_fd, page_size);
assert(ret == 0);
/* Recover the size */
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
/* Re-read the data, it should be all zero */
val = buffer[page_size];
if (val == 0)
printf("Good\n");
else
printf("BUG\n");
}
===8<===
We don't need to touch up the pmd path, because pmd never had a issue with
swap entries. For example, shmem pmd migration will always be split into
pte level, and same to swapping on anonymous.
Add another helper should_zap_cows() so that we can also check whether we
should zap private mappings when there's no page pointer specified.
This patch drops that trick, so we handle swap ptes coherently. Meanwhile
we should do the same check upon migration entry, hwpoison entry and
genuine swap entries too.
To be explicit, we should still remember to keep the private entries if
even_cows==false, and always zap them when even_cows==true.
The issue seems to exist starting from the initial commit of git.
[peterx@redhat.com: comment tweaks]
Link: https://lkml.kernel.org/r/20220217060746.71256-2-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220217060746.71256-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-2-peterx@redhat.com
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:15 +00:00
|
|
|
continue;
|
2016-01-14 23:19:26 +00:00
|
|
|
rss[mm_counter(page)]--;
|
2022-05-13 03:22:53 +00:00
|
|
|
} else if (pte_marker_entry_uffd_wp(entry)) {
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
/*
|
|
|
|
* For anon: always drop the marker; for file: only
|
|
|
|
* drop the marker if explicitly requested.
|
|
|
|
*/
|
|
|
|
if (!vma_is_anonymous(vma) &&
|
|
|
|
!zap_drop_file_uffd_wp(details))
|
2022-05-13 03:22:53 +00:00
|
|
|
continue;
|
2022-05-19 12:50:26 +00:00
|
|
|
} else if (is_hwpoison_entry(entry) ||
|
|
|
|
is_swapin_error_entry(entry)) {
|
mm: don't skip swap entry even if zap_details specified
Patch series "mm: Rework zap ptes on swap entries", v5.
Patch 1 should fix a long standing bug for zap_pte_range() on
zap_details usage. The risk is we could have some swap entries skipped
while we should have zapped them.
Migration entries are not the major concern because file backed memory
always zap in the pattern that "first time without page lock, then
re-zap with page lock" hence the 2nd zap will always make sure all
migration entries are already recovered.
However there can be issues with real swap entries got skipped
errornoously. There's a reproducer provided in commit message of patch
1 for that.
Patch 2-4 are cleanups that are based on patch 1. After the whole
patchset applied, we should have a very clean view of zap_pte_range().
Only patch 1 needs to be backported to stable if necessary.
This patch (of 4):
The "details" pointer shouldn't be the token to decide whether we should
skip swap entries.
For example, when the callers specified details->zap_mapping==NULL, it
means the user wants to zap all the pages (including COWed pages), then
we need to look into swap entries because there can be private COWed
pages that was swapped out.
Skipping some swap entries when details is non-NULL may lead to wrongly
leaving some of the swap entries while we should have zapped them.
A reproducer of the problem:
===8<===
#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <stdio.h>
#include <assert.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/types.h>
int page_size;
int shmem_fd;
char *buffer;
void main(void)
{
int ret;
char val;
page_size = getpagesize();
shmem_fd = memfd_create("test", 0);
assert(shmem_fd >= 0);
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
buffer = mmap(NULL, page_size * 2, PROT_READ | PROT_WRITE,
MAP_PRIVATE, shmem_fd, 0);
assert(buffer != MAP_FAILED);
/* Write private page, swap it out */
buffer[page_size] = 1;
madvise(buffer, page_size * 2, MADV_PAGEOUT);
/* This should drop private buffer[page_size] already */
ret = ftruncate(shmem_fd, page_size);
assert(ret == 0);
/* Recover the size */
ret = ftruncate(shmem_fd, page_size * 2);
assert(ret == 0);
/* Re-read the data, it should be all zero */
val = buffer[page_size];
if (val == 0)
printf("Good\n");
else
printf("BUG\n");
}
===8<===
We don't need to touch up the pmd path, because pmd never had a issue with
swap entries. For example, shmem pmd migration will always be split into
pte level, and same to swapping on anonymous.
Add another helper should_zap_cows() so that we can also check whether we
should zap private mappings when there's no page pointer specified.
This patch drops that trick, so we handle swap ptes coherently. Meanwhile
we should do the same check upon migration entry, hwpoison entry and
genuine swap entries too.
To be explicit, we should still remember to keep the private entries if
even_cows==false, and always zap them when even_cows==true.
The issue seems to exist starting from the initial commit of git.
[peterx@redhat.com: comment tweaks]
Link: https://lkml.kernel.org/r/20220217060746.71256-2-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220217060746.71256-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220216094810.60572-2-peterx@redhat.com
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:15 +00:00
|
|
|
if (!should_zap_cows(details))
|
|
|
|
continue;
|
|
|
|
} else {
|
|
|
|
/* We should have covered all the swap entry types */
|
|
|
|
WARN_ON_ONCE(1);
|
2010-03-05 21:41:42 +00:00
|
|
|
}
|
2006-10-01 06:29:31 +00:00
|
|
|
pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
|
2022-05-13 03:22:53 +00:00
|
|
|
zap_install_uffd_wp_if_needed(vma, addr, pte, details, ptent);
|
2011-05-25 00:12:04 +00:00
|
|
|
} while (pte++, addr += PAGE_SIZE, addr != end);
|
2005-10-30 01:16:05 +00:00
|
|
|
|
2010-03-05 21:41:39 +00:00
|
|
|
add_mm_rss_vec(mm, rss);
|
2006-10-01 06:29:33 +00:00
|
|
|
arch_leave_lazy_mmu_mode();
|
2005-11-14 00:06:42 +00:00
|
|
|
|
2014-04-25 23:05:40 +00:00
|
|
|
/* Do the actual TLB flush before dropping ptl */
|
mm: delay page_remove_rmap() until after the TLB has been flushed
When we remove a page table entry, we are very careful to only free the
page after we have flushed the TLB, because other CPUs could still be
using the page through stale TLB entries until after the flush.
However, we have removed the rmap entry for that page early, which means
that functions like folio_mkclean() would end up not serializing with the
page table lock because the page had already been made invisible to rmap.
And that is a problem, because while the TLB entry exists, we could end up
with the following situation:
(a) one CPU could come in and clean it, never seeing our mapping of the
page
(b) another CPU could continue to use the stale and dirty TLB entry and
continue to write to said page
resulting in a page that has been dirtied, but then marked clean again,
all while another CPU might have dirtied it some more.
End result: possibly lost dirty data.
This extends our current TLB gather infrastructure to optionally track a
"should I do a delayed page_remove_rmap() for this page after flushing the
TLB". It uses the newly introduced 'encoded page pointer' to do that
without having to keep separate data around.
Note, this is complicated by a couple of issues:
- we want to delay the rmap removal, but not past the page table lock,
because that simplifies the memcg accounting
- only SMP configurations want to delay TLB flushing, since on UP
there are obviously no remote TLBs to worry about, and the page
table lock means there are no preemption issues either
- s390 has its own mmu_gather model that doesn't delay TLB flushing,
and as a result also does not want the delayed rmap. As such, we can
treat S390 like the UP case and use a common fallback for the "no
delays" case.
- we can track an enormous number of pages in our mmu_gather structure,
with MAX_GATHER_BATCH_COUNT batches of MAX_TABLE_BATCH pages each,
all set up to be approximately 10k pending pages.
We do not want to have a huge number of batched pages that we then
need to check for delayed rmap handling inside the page table lock.
Particularly that last point results in a noteworthy detail, where the
normal page batch gathering is limited once we have delayed rmaps pending,
in such a way that only the last batch (the so-called "active batch") in
the mmu_gather structure can have any delayed entries.
NOTE! While the "possibly lost dirty data" sounds catastrophic, for this
all to happen you need to have a user thread doing either madvise() with
MADV_DONTNEED or a full re-mmap() of the area concurrently with another
thread continuing to use said mapping.
So arguably this is about user space doing crazy things, but from a VM
consistency standpoint it's better if we track the dirty bit properly even
when user space goes off the rails.
[akpm@linux-foundation.org: fix UP build, per Linus]
Link: https://lore.kernel.org/all/B88D3073-440A-41C7-95F4-895D3F657EF2@gmail.com/
Link: https://lkml.kernel.org/r/20221109203051.1835763-4-torvalds@linux-foundation.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Tested-by: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-09 20:30:51 +00:00
|
|
|
if (force_flush) {
|
2014-04-25 23:05:40 +00:00
|
|
|
tlb_flush_mmu_tlbonly(tlb);
|
2022-11-16 07:49:30 +00:00
|
|
|
tlb_flush_rmaps(tlb, vma);
|
mm: delay page_remove_rmap() until after the TLB has been flushed
When we remove a page table entry, we are very careful to only free the
page after we have flushed the TLB, because other CPUs could still be
using the page through stale TLB entries until after the flush.
However, we have removed the rmap entry for that page early, which means
that functions like folio_mkclean() would end up not serializing with the
page table lock because the page had already been made invisible to rmap.
And that is a problem, because while the TLB entry exists, we could end up
with the following situation:
(a) one CPU could come in and clean it, never seeing our mapping of the
page
(b) another CPU could continue to use the stale and dirty TLB entry and
continue to write to said page
resulting in a page that has been dirtied, but then marked clean again,
all while another CPU might have dirtied it some more.
End result: possibly lost dirty data.
This extends our current TLB gather infrastructure to optionally track a
"should I do a delayed page_remove_rmap() for this page after flushing the
TLB". It uses the newly introduced 'encoded page pointer' to do that
without having to keep separate data around.
Note, this is complicated by a couple of issues:
- we want to delay the rmap removal, but not past the page table lock,
because that simplifies the memcg accounting
- only SMP configurations want to delay TLB flushing, since on UP
there are obviously no remote TLBs to worry about, and the page
table lock means there are no preemption issues either
- s390 has its own mmu_gather model that doesn't delay TLB flushing,
and as a result also does not want the delayed rmap. As such, we can
treat S390 like the UP case and use a common fallback for the "no
delays" case.
- we can track an enormous number of pages in our mmu_gather structure,
with MAX_GATHER_BATCH_COUNT batches of MAX_TABLE_BATCH pages each,
all set up to be approximately 10k pending pages.
We do not want to have a huge number of batched pages that we then
need to check for delayed rmap handling inside the page table lock.
Particularly that last point results in a noteworthy detail, where the
normal page batch gathering is limited once we have delayed rmaps pending,
in such a way that only the last batch (the so-called "active batch") in
the mmu_gather structure can have any delayed entries.
NOTE! While the "possibly lost dirty data" sounds catastrophic, for this
all to happen you need to have a user thread doing either madvise() with
MADV_DONTNEED or a full re-mmap() of the area concurrently with another
thread continuing to use said mapping.
So arguably this is about user space doing crazy things, but from a VM
consistency standpoint it's better if we track the dirty bit properly even
when user space goes off the rails.
[akpm@linux-foundation.org: fix UP build, per Linus]
Link: https://lore.kernel.org/all/B88D3073-440A-41C7-95F4-895D3F657EF2@gmail.com/
Link: https://lkml.kernel.org/r/20221109203051.1835763-4-torvalds@linux-foundation.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Tested-by: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-09 20:30:51 +00:00
|
|
|
}
|
2014-04-25 23:05:40 +00:00
|
|
|
pte_unmap_unlock(start_pte, ptl);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we forced a TLB flush (either due to running out of
|
|
|
|
* batch buffers or because we needed to flush dirty TLB
|
|
|
|
* entries before releasing the ptl), free the batched
|
|
|
|
* memory too. Restart if we didn't do everything.
|
|
|
|
*/
|
|
|
|
if (force_flush) {
|
|
|
|
force_flush = 0;
|
2018-09-20 08:54:04 +00:00
|
|
|
tlb_flush_mmu(tlb);
|
2019-09-24 00:02:24 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (addr != end) {
|
|
|
|
cond_resched();
|
|
|
|
goto again;
|
2011-05-25 00:11:45 +00:00
|
|
|
}
|
|
|
|
|
2005-11-14 00:06:42 +00:00
|
|
|
return addr;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2005-11-14 00:06:42 +00:00
|
|
|
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
|
2005-10-30 01:16:12 +00:00
|
|
|
struct vm_area_struct *vma, pud_t *pud,
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long addr, unsigned long end,
|
2011-05-25 00:12:04 +00:00
|
|
|
struct zap_details *details)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pmd_t *pmd;
|
|
|
|
unsigned long next;
|
|
|
|
|
|
|
|
pmd = pmd_offset(pud, addr);
|
|
|
|
do {
|
|
|
|
next = pmd_addr_end(addr, end);
|
mm: thp: check pmd migration entry in common path
When THP migration is being used, memory management code needs to handle
pmd migration entries properly. This patch uses !pmd_present() or
is_swap_pmd() (depending on whether pmd_none() needs separate code or
not) to check pmd migration entries at the places where a pmd entry is
present.
Since pmd-related code uses split_huge_page(), split_huge_pmd(),
pmd_trans_huge(), pmd_trans_unstable(), or
pmd_none_or_trans_huge_or_clear_bad(), this patch:
1. adds pmd migration entry split code in split_huge_pmd(),
2. takes care of pmd migration entries whenever pmd_trans_huge() is present,
3. makes pmd_none_or_trans_huge_or_clear_bad() pmd migration entry aware.
Since split_huge_page() uses split_huge_pmd() and pmd_trans_unstable()
is equivalent to pmd_none_or_trans_huge_or_clear_bad(), we do not change
them.
Until this commit, a pmd entry should be:
1. pointing to a pte page,
2. is_swap_pmd(),
3. pmd_trans_huge(),
4. pmd_devmap(), or
5. pmd_none().
Signed-off-by: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Nellans <dnellans@nvidia.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:11:01 +00:00
|
|
|
if (is_swap_pmd(*pmd) || pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
|
mm: delete historical BUG from zap_pmd_range()
Delete the old VM_BUG_ON_VMA() from zap_pmd_range(), which asserted
that mmap_sem must be held when splitting an "anonymous" vma there.
Whether that's still strictly true nowadays is not entirely clear,
but the danger of sometimes crashing on the BUG is now fairly clear.
Even with the new stricter rules for anonymous vma marking, the
condition it checks for can possible trigger. Commit 44960f2a7b63
("staging: ashmem: Fix SIGBUS crash when traversing mmaped ashmem
pages") is good, and originally I thought it was safe from that
VM_BUG_ON_VMA(), because the /dev/ashmem fd exposed to the user is
disconnected from the vm_file in the vma, and madvise(,,MADV_REMOVE)
insists on VM_SHARED.
But after I read John's earlier mail, drawing attention to the
vfs_fallocate() in there: I may be wrong, and I don't know if Android
has THP in the config anyway, but it looks to me like an
unmap_mapping_range() from ashmem's vfs_fallocate() could hit precisely
the VM_BUG_ON_VMA(), once it's vma_is_anonymous().
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Kirill Shutemov <kirill.shutemov@linux.intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-01 18:31:52 +00:00
|
|
|
if (next - addr != HPAGE_PMD_SIZE)
|
2016-12-13 00:42:20 +00:00
|
|
|
__split_huge_pmd(vma, pmd, addr, false, NULL);
|
mm: delete historical BUG from zap_pmd_range()
Delete the old VM_BUG_ON_VMA() from zap_pmd_range(), which asserted
that mmap_sem must be held when splitting an "anonymous" vma there.
Whether that's still strictly true nowadays is not entirely clear,
but the danger of sometimes crashing on the BUG is now fairly clear.
Even with the new stricter rules for anonymous vma marking, the
condition it checks for can possible trigger. Commit 44960f2a7b63
("staging: ashmem: Fix SIGBUS crash when traversing mmaped ashmem
pages") is good, and originally I thought it was safe from that
VM_BUG_ON_VMA(), because the /dev/ashmem fd exposed to the user is
disconnected from the vm_file in the vma, and madvise(,,MADV_REMOVE)
insists on VM_SHARED.
But after I read John's earlier mail, drawing attention to the
vfs_fallocate() in there: I may be wrong, and I don't know if Android
has THP in the config anyway, but it looks to me like an
unmap_mapping_range() from ashmem's vfs_fallocate() could hit precisely
the VM_BUG_ON_VMA(), once it's vma_is_anonymous().
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Kirill Shutemov <kirill.shutemov@linux.intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-01 18:31:52 +00:00
|
|
|
else if (zap_huge_pmd(tlb, vma, pmd, addr))
|
mm: thp: fix pmd_bad() triggering in code paths holding mmap_sem read mode
In some cases it may happen that pmd_none_or_clear_bad() is called with
the mmap_sem hold in read mode. In those cases the huge page faults can
allocate hugepmds under pmd_none_or_clear_bad() and that can trigger a
false positive from pmd_bad() that will not like to see a pmd
materializing as trans huge.
It's not khugepaged causing the problem, khugepaged holds the mmap_sem
in write mode (and all those sites must hold the mmap_sem in read mode
to prevent pagetables to go away from under them, during code review it
seems vm86 mode on 32bit kernels requires that too unless it's
restricted to 1 thread per process or UP builds). The race is only with
the huge pagefaults that can convert a pmd_none() into a
pmd_trans_huge().
Effectively all these pmd_none_or_clear_bad() sites running with
mmap_sem in read mode are somewhat speculative with the page faults, and
the result is always undefined when they run simultaneously. This is
probably why it wasn't common to run into this. For example if the
madvise(MADV_DONTNEED) runs zap_page_range() shortly before the page
fault, the hugepage will not be zapped, if the page fault runs first it
will be zapped.
Altering pmd_bad() not to error out if it finds hugepmds won't be enough
to fix this, because zap_pmd_range would then proceed to call
zap_pte_range (which would be incorrect if the pmd become a
pmd_trans_huge()).
The simplest way to fix this is to read the pmd in the local stack
(regardless of what we read, no need of actual CPU barriers, only
compiler barrier needed), and be sure it is not changing under the code
that computes its value. Even if the real pmd is changing under the
value we hold on the stack, we don't care. If we actually end up in
zap_pte_range it means the pmd was not none already and it was not huge,
and it can't become huge from under us (khugepaged locking explained
above).
All we need is to enforce that there is no way anymore that in a code
path like below, pmd_trans_huge can be false, but pmd_none_or_clear_bad
can run into a hugepmd. The overhead of a barrier() is just a compiler
tweak and should not be measurable (I only added it for THP builds). I
don't exclude different compiler versions may have prevented the race
too by caching the value of *pmd on the stack (that hasn't been
verified, but it wouldn't be impossible considering
pmd_none_or_clear_bad, pmd_bad, pmd_trans_huge, pmd_none are all inlines
and there's no external function called in between pmd_trans_huge and
pmd_none_or_clear_bad).
if (pmd_trans_huge(*pmd)) {
if (next-addr != HPAGE_PMD_SIZE) {
VM_BUG_ON(!rwsem_is_locked(&tlb->mm->mmap_sem));
split_huge_page_pmd(vma->vm_mm, pmd);
} else if (zap_huge_pmd(tlb, vma, pmd, addr))
continue;
/* fall through */
}
if (pmd_none_or_clear_bad(pmd))
Because this race condition could be exercised without special
privileges this was reported in CVE-2012-1179.
The race was identified and fully explained by Ulrich who debugged it.
I'm quoting his accurate explanation below, for reference.
====== start quote =======
mapcount 0 page_mapcount 1
kernel BUG at mm/huge_memory.c:1384!
At some point prior to the panic, a "bad pmd ..." message similar to the
following is logged on the console:
mm/memory.c:145: bad pmd ffff8800376e1f98(80000000314000e7).
The "bad pmd ..." message is logged by pmd_clear_bad() before it clears
the page's PMD table entry.
143 void pmd_clear_bad(pmd_t *pmd)
144 {
-> 145 pmd_ERROR(*pmd);
146 pmd_clear(pmd);
147 }
After the PMD table entry has been cleared, there is an inconsistency
between the actual number of PMD table entries that are mapping the page
and the page's map count (_mapcount field in struct page). When the page
is subsequently reclaimed, __split_huge_page() detects this inconsistency.
1381 if (mapcount != page_mapcount(page))
1382 printk(KERN_ERR "mapcount %d page_mapcount %d\n",
1383 mapcount, page_mapcount(page));
-> 1384 BUG_ON(mapcount != page_mapcount(page));
The root cause of the problem is a race of two threads in a multithreaded
process. Thread B incurs a page fault on a virtual address that has never
been accessed (PMD entry is zero) while Thread A is executing an madvise()
system call on a virtual address within the same 2 MB (huge page) range.
virtual address space
.---------------------.
| |
| |
.-|---------------------|
| | |
| | |<-- B(fault)
| | |
2 MB | |/////////////////////|-.
huge < |/////////////////////| > A(range)
page | |/////////////////////|-'
| | |
| | |
'-|---------------------|
| |
| |
'---------------------'
- Thread A is executing an madvise(..., MADV_DONTNEED) system call
on the virtual address range "A(range)" shown in the picture.
sys_madvise
// Acquire the semaphore in shared mode.
down_read(¤t->mm->mmap_sem)
...
madvise_vma
switch (behavior)
case MADV_DONTNEED:
madvise_dontneed
zap_page_range
unmap_vmas
unmap_page_range
zap_pud_range
zap_pmd_range
//
// Assume that this huge page has never been accessed.
// I.e. content of the PMD entry is zero (not mapped).
//
if (pmd_trans_huge(*pmd)) {
// We don't get here due to the above assumption.
}
//
// Assume that Thread B incurred a page fault and
.---------> // sneaks in here as shown below.
| //
| if (pmd_none_or_clear_bad(pmd))
| {
| if (unlikely(pmd_bad(*pmd)))
| pmd_clear_bad
| {
| pmd_ERROR
| // Log "bad pmd ..." message here.
| pmd_clear
| // Clear the page's PMD entry.
| // Thread B incremented the map count
| // in page_add_new_anon_rmap(), but
| // now the page is no longer mapped
| // by a PMD entry (-> inconsistency).
| }
| }
|
v
- Thread B is handling a page fault on virtual address "B(fault)" shown
in the picture.
...
do_page_fault
__do_page_fault
// Acquire the semaphore in shared mode.
down_read_trylock(&mm->mmap_sem)
...
handle_mm_fault
if (pmd_none(*pmd) && transparent_hugepage_enabled(vma))
// We get here due to the above assumption (PMD entry is zero).
do_huge_pmd_anonymous_page
alloc_hugepage_vma
// Allocate a new transparent huge page here.
...
__do_huge_pmd_anonymous_page
...
spin_lock(&mm->page_table_lock)
...
page_add_new_anon_rmap
// Here we increment the page's map count (starts at -1).
atomic_set(&page->_mapcount, 0)
set_pmd_at
// Here we set the page's PMD entry which will be cleared
// when Thread A calls pmd_clear_bad().
...
spin_unlock(&mm->page_table_lock)
The mmap_sem does not prevent the race because both threads are acquiring
it in shared mode (down_read). Thread B holds the page_table_lock while
the page's map count and PMD table entry are updated. However, Thread A
does not synchronize on that lock.
====== end quote =======
[akpm@linux-foundation.org: checkpatch fixes]
Reported-by: Ulrich Obergfell <uobergfe@redhat.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Jones <davej@redhat.com>
Acked-by: Larry Woodman <lwoodman@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org> [2.6.38+]
Cc: Mark Salter <msalter@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-03-21 23:33:42 +00:00
|
|
|
goto next;
|
thp: transparent hugepage core
Lately I've been working to make KVM use hugepages transparently without
the usual restrictions of hugetlbfs. Some of the restrictions I'd like to
see removed:
1) hugepages have to be swappable or the guest physical memory remains
locked in RAM and can't be paged out to swap
2) if a hugepage allocation fails, regular pages should be allocated
instead and mixed in the same vma without any failure and without
userland noticing
3) if some task quits and more hugepages become available in the
buddy, guest physical memory backed by regular pages should be
relocated on hugepages automatically in regions under
madvise(MADV_HUGEPAGE) (ideally event driven by waking up the
kernel deamon if the order=HPAGE_PMD_SHIFT-PAGE_SHIFT list becomes
not null)
4) avoidance of reservation and maximization of use of hugepages whenever
possible. Reservation (needed to avoid runtime fatal faliures) may be ok for
1 machine with 1 database with 1 database cache with 1 database cache size
known at boot time. It's definitely not feasible with a virtualization
hypervisor usage like RHEV-H that runs an unknown number of virtual machines
with an unknown size of each virtual machine with an unknown amount of
pagecache that could be potentially useful in the host for guest not using
O_DIRECT (aka cache=off).
hugepages in the virtualization hypervisor (and also in the guest!) are
much more important than in a regular host not using virtualization,
becasue with NPT/EPT they decrease the tlb-miss cacheline accesses from 24
to 19 in case only the hypervisor uses transparent hugepages, and they
decrease the tlb-miss cacheline accesses from 19 to 15 in case both the
linux hypervisor and the linux guest both uses this patch (though the
guest will limit the addition speedup to anonymous regions only for
now...). Even more important is that the tlb miss handler is much slower
on a NPT/EPT guest than for a regular shadow paging or no-virtualization
scenario. So maximizing the amount of virtual memory cached by the TLB
pays off significantly more with NPT/EPT than without (even if there would
be no significant speedup in the tlb-miss runtime).
The first (and more tedious) part of this work requires allowing the VM to
handle anonymous hugepages mixed with regular pages transparently on
regular anonymous vmas. This is what this patch tries to achieve in the
least intrusive possible way. We want hugepages and hugetlb to be used in
a way so that all applications can benefit without changes (as usual we
leverage the KVM virtualization design: by improving the Linux VM at
large, KVM gets the performance boost too).
The most important design choice is: always fallback to 4k allocation if
the hugepage allocation fails! This is the _very_ opposite of some large
pagecache patches that failed with -EIO back then if a 64k (or similar)
allocation failed...
Second important decision (to reduce the impact of the feature on the
existing pagetable handling code) is that at any time we can split an
hugepage into 512 regular pages and it has to be done with an operation
that can't fail. This way the reliability of the swapping isn't decreased
(no need to allocate memory when we are short on memory to swap) and it's
trivial to plug a split_huge_page* one-liner where needed without
polluting the VM. Over time we can teach mprotect, mremap and friends to
handle pmd_trans_huge natively without calling split_huge_page*. The fact
it can't fail isn't just for swap: if split_huge_page would return -ENOMEM
(instead of the current void) we'd need to rollback the mprotect from the
middle of it (ideally including undoing the split_vma) which would be a
big change and in the very wrong direction (it'd likely be simpler not to
call split_huge_page at all and to teach mprotect and friends to handle
hugepages instead of rolling them back from the middle). In short the
very value of split_huge_page is that it can't fail.
The collapsing and madvise(MADV_HUGEPAGE) part will remain separated and
incremental and it'll just be an "harmless" addition later if this initial
part is agreed upon. It also should be noted that locking-wise replacing
regular pages with hugepages is going to be very easy if compared to what
I'm doing below in split_huge_page, as it will only happen when
page_count(page) matches page_mapcount(page) if we can take the PG_lock
and mmap_sem in write mode. collapse_huge_page will be a "best effort"
that (unlike split_huge_page) can fail at the minimal sign of trouble and
we can try again later. collapse_huge_page will be similar to how KSM
works and the madvise(MADV_HUGEPAGE) will work similar to
madvise(MADV_MERGEABLE).
The default I like is that transparent hugepages are used at page fault
time. This can be changed with
/sys/kernel/mm/transparent_hugepage/enabled. The control knob can be set
to three values "always", "madvise", "never" which mean respectively that
hugepages are always used, or only inside madvise(MADV_HUGEPAGE) regions,
or never used. /sys/kernel/mm/transparent_hugepage/defrag instead
controls if the hugepage allocation should defrag memory aggressively
"always", only inside "madvise" regions, or "never".
The pmd_trans_splitting/pmd_trans_huge locking is very solid. The
put_page (from get_user_page users that can't use mmu notifier like
O_DIRECT) that runs against a __split_huge_page_refcount instead was a
pain to serialize in a way that would result always in a coherent page
count for both tail and head. I think my locking solution with a
compound_lock taken only after the page_first is valid and is still a
PageHead should be safe but it surely needs review from SMP race point of
view. In short there is no current existing way to serialize the O_DIRECT
final put_page against split_huge_page_refcount so I had to invent a new
one (O_DIRECT loses knowledge on the mapping status by the time gup_fast
returns so...). And I didn't want to impact all gup/gup_fast users for
now, maybe if we change the gup interface substantially we can avoid this
locking, I admit I didn't think too much about it because changing the gup
unpinning interface would be invasive.
If we ignored O_DIRECT we could stick to the existing compound refcounting
code, by simply adding a get_user_pages_fast_flags(foll_flags) where KVM
(and any other mmu notifier user) would call it without FOLL_GET (and if
FOLL_GET isn't set we'd just BUG_ON if nobody registered itself in the
current task mmu notifier list yet). But O_DIRECT is fundamental for
decent performance of virtualized I/O on fast storage so we can't avoid it
to solve the race of put_page against split_huge_page_refcount to achieve
a complete hugepage feature for KVM.
Swap and oom works fine (well just like with regular pages ;). MMU
notifier is handled transparently too, with the exception of the young bit
on the pmd, that didn't have a range check but I think KVM will be fine
because the whole point of hugepages is that EPT/NPT will also use a huge
pmd when they notice gup returns pages with PageCompound set, so they
won't care of a range and there's just the pmd young bit to check in that
case.
NOTE: in some cases if the L2 cache is small, this may slowdown and waste
memory during COWs because 4M of memory are accessed in a single fault
instead of 8k (the payoff is that after COW the program can run faster).
So we might want to switch the copy_huge_page (and clear_huge_page too) to
not temporal stores. I also extensively researched ways to avoid this
cache trashing with a full prefault logic that would cow in 8k/16k/32k/64k
up to 1M (I can send those patches that fully implemented prefault) but I
concluded they're not worth it and they add an huge additional complexity
and they remove all tlb benefits until the full hugepage has been faulted
in, to save a little bit of memory and some cache during app startup, but
they still don't improve substantially the cache-trashing during startup
if the prefault happens in >4k chunks. One reason is that those 4k pte
entries copied are still mapped on a perfectly cache-colored hugepage, so
the trashing is the worst one can generate in those copies (cow of 4k page
copies aren't so well colored so they trashes less, but again this results
in software running faster after the page fault). Those prefault patches
allowed things like a pte where post-cow pages were local 4k regular anon
pages and the not-yet-cowed pte entries were pointing in the middle of
some hugepage mapped read-only. If it doesn't payoff substantially with
todays hardware it will payoff even less in the future with larger l2
caches, and the prefault logic would blot the VM a lot. If one is
emebdded transparent_hugepage can be disabled during boot with sysfs or
with the boot commandline parameter transparent_hugepage=0 (or
transparent_hugepage=2 to restrict hugepages inside madvise regions) that
will ensure not a single hugepage is allocated at boot time. It is simple
enough to just disable transparent hugepage globally and let transparent
hugepages be allocated selectively by applications in the MADV_HUGEPAGE
region (both at page fault time, and if enabled with the
collapse_huge_page too through the kernel daemon).
This patch supports only hugepages mapped in the pmd, archs that have
smaller hugepages will not fit in this patch alone. Also some archs like
power have certain tlb limits that prevents mixing different page size in
the same regions so they will not fit in this framework that requires
"graceful fallback" to basic PAGE_SIZE in case of physical memory
fragmentation. hugetlbfs remains a perfect fit for those because its
software limits happen to match the hardware limits. hugetlbfs also
remains a perfect fit for hugepage sizes like 1GByte that cannot be hoped
to be found not fragmented after a certain system uptime and that would be
very expensive to defragment with relocation, so requiring reservation.
hugetlbfs is the "reservation way", the point of transparent hugepages is
not to have any reservation at all and maximizing the use of cache and
hugepages at all times automatically.
Some performance result:
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largep
ages3
memset page fault 1566023
memset tlb miss 453854
memset second tlb miss 453321
random access tlb miss 41635
random access second tlb miss 41658
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largepages3
memset page fault 1566471
memset tlb miss 453375
memset second tlb miss 453320
random access tlb miss 41636
random access second tlb miss 41637
vmx andrea # ./largepages3
memset page fault 1566642
memset tlb miss 453417
memset second tlb miss 453313
random access tlb miss 41630
random access second tlb miss 41647
vmx andrea # ./largepages3
memset page fault 1566872
memset tlb miss 453418
memset second tlb miss 453315
random access tlb miss 41618
random access second tlb miss 41659
vmx andrea # echo 0 > /proc/sys/vm/transparent_hugepage
vmx andrea # ./largepages3
memset page fault 2182476
memset tlb miss 460305
memset second tlb miss 460179
random access tlb miss 44483
random access second tlb miss 44186
vmx andrea # ./largepages3
memset page fault 2182791
memset tlb miss 460742
memset second tlb miss 459962
random access tlb miss 43981
random access second tlb miss 43988
============
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#define SIZE (3UL*1024*1024*1024)
int main()
{
char *p = malloc(SIZE), *p2;
struct timeval before, after;
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset page fault %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
return 0;
}
============
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:46:52 +00:00
|
|
|
/* fall through */
|
2021-11-28 19:53:35 +00:00
|
|
|
} else if (details && details->single_folio &&
|
|
|
|
folio_test_pmd_mappable(details->single_folio) &&
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
next - addr == HPAGE_PMD_SIZE && pmd_none(*pmd)) {
|
|
|
|
spinlock_t *ptl = pmd_lock(tlb->mm, pmd);
|
|
|
|
/*
|
|
|
|
* Take and drop THP pmd lock so that we cannot return
|
|
|
|
* prematurely, while zap_huge_pmd() has cleared *pmd,
|
|
|
|
* but not yet decremented compound_mapcount().
|
|
|
|
*/
|
|
|
|
spin_unlock(ptl);
|
thp: transparent hugepage core
Lately I've been working to make KVM use hugepages transparently without
the usual restrictions of hugetlbfs. Some of the restrictions I'd like to
see removed:
1) hugepages have to be swappable or the guest physical memory remains
locked in RAM and can't be paged out to swap
2) if a hugepage allocation fails, regular pages should be allocated
instead and mixed in the same vma without any failure and without
userland noticing
3) if some task quits and more hugepages become available in the
buddy, guest physical memory backed by regular pages should be
relocated on hugepages automatically in regions under
madvise(MADV_HUGEPAGE) (ideally event driven by waking up the
kernel deamon if the order=HPAGE_PMD_SHIFT-PAGE_SHIFT list becomes
not null)
4) avoidance of reservation and maximization of use of hugepages whenever
possible. Reservation (needed to avoid runtime fatal faliures) may be ok for
1 machine with 1 database with 1 database cache with 1 database cache size
known at boot time. It's definitely not feasible with a virtualization
hypervisor usage like RHEV-H that runs an unknown number of virtual machines
with an unknown size of each virtual machine with an unknown amount of
pagecache that could be potentially useful in the host for guest not using
O_DIRECT (aka cache=off).
hugepages in the virtualization hypervisor (and also in the guest!) are
much more important than in a regular host not using virtualization,
becasue with NPT/EPT they decrease the tlb-miss cacheline accesses from 24
to 19 in case only the hypervisor uses transparent hugepages, and they
decrease the tlb-miss cacheline accesses from 19 to 15 in case both the
linux hypervisor and the linux guest both uses this patch (though the
guest will limit the addition speedup to anonymous regions only for
now...). Even more important is that the tlb miss handler is much slower
on a NPT/EPT guest than for a regular shadow paging or no-virtualization
scenario. So maximizing the amount of virtual memory cached by the TLB
pays off significantly more with NPT/EPT than without (even if there would
be no significant speedup in the tlb-miss runtime).
The first (and more tedious) part of this work requires allowing the VM to
handle anonymous hugepages mixed with regular pages transparently on
regular anonymous vmas. This is what this patch tries to achieve in the
least intrusive possible way. We want hugepages and hugetlb to be used in
a way so that all applications can benefit without changes (as usual we
leverage the KVM virtualization design: by improving the Linux VM at
large, KVM gets the performance boost too).
The most important design choice is: always fallback to 4k allocation if
the hugepage allocation fails! This is the _very_ opposite of some large
pagecache patches that failed with -EIO back then if a 64k (or similar)
allocation failed...
Second important decision (to reduce the impact of the feature on the
existing pagetable handling code) is that at any time we can split an
hugepage into 512 regular pages and it has to be done with an operation
that can't fail. This way the reliability of the swapping isn't decreased
(no need to allocate memory when we are short on memory to swap) and it's
trivial to plug a split_huge_page* one-liner where needed without
polluting the VM. Over time we can teach mprotect, mremap and friends to
handle pmd_trans_huge natively without calling split_huge_page*. The fact
it can't fail isn't just for swap: if split_huge_page would return -ENOMEM
(instead of the current void) we'd need to rollback the mprotect from the
middle of it (ideally including undoing the split_vma) which would be a
big change and in the very wrong direction (it'd likely be simpler not to
call split_huge_page at all and to teach mprotect and friends to handle
hugepages instead of rolling them back from the middle). In short the
very value of split_huge_page is that it can't fail.
The collapsing and madvise(MADV_HUGEPAGE) part will remain separated and
incremental and it'll just be an "harmless" addition later if this initial
part is agreed upon. It also should be noted that locking-wise replacing
regular pages with hugepages is going to be very easy if compared to what
I'm doing below in split_huge_page, as it will only happen when
page_count(page) matches page_mapcount(page) if we can take the PG_lock
and mmap_sem in write mode. collapse_huge_page will be a "best effort"
that (unlike split_huge_page) can fail at the minimal sign of trouble and
we can try again later. collapse_huge_page will be similar to how KSM
works and the madvise(MADV_HUGEPAGE) will work similar to
madvise(MADV_MERGEABLE).
The default I like is that transparent hugepages are used at page fault
time. This can be changed with
/sys/kernel/mm/transparent_hugepage/enabled. The control knob can be set
to three values "always", "madvise", "never" which mean respectively that
hugepages are always used, or only inside madvise(MADV_HUGEPAGE) regions,
or never used. /sys/kernel/mm/transparent_hugepage/defrag instead
controls if the hugepage allocation should defrag memory aggressively
"always", only inside "madvise" regions, or "never".
The pmd_trans_splitting/pmd_trans_huge locking is very solid. The
put_page (from get_user_page users that can't use mmu notifier like
O_DIRECT) that runs against a __split_huge_page_refcount instead was a
pain to serialize in a way that would result always in a coherent page
count for both tail and head. I think my locking solution with a
compound_lock taken only after the page_first is valid and is still a
PageHead should be safe but it surely needs review from SMP race point of
view. In short there is no current existing way to serialize the O_DIRECT
final put_page against split_huge_page_refcount so I had to invent a new
one (O_DIRECT loses knowledge on the mapping status by the time gup_fast
returns so...). And I didn't want to impact all gup/gup_fast users for
now, maybe if we change the gup interface substantially we can avoid this
locking, I admit I didn't think too much about it because changing the gup
unpinning interface would be invasive.
If we ignored O_DIRECT we could stick to the existing compound refcounting
code, by simply adding a get_user_pages_fast_flags(foll_flags) where KVM
(and any other mmu notifier user) would call it without FOLL_GET (and if
FOLL_GET isn't set we'd just BUG_ON if nobody registered itself in the
current task mmu notifier list yet). But O_DIRECT is fundamental for
decent performance of virtualized I/O on fast storage so we can't avoid it
to solve the race of put_page against split_huge_page_refcount to achieve
a complete hugepage feature for KVM.
Swap and oom works fine (well just like with regular pages ;). MMU
notifier is handled transparently too, with the exception of the young bit
on the pmd, that didn't have a range check but I think KVM will be fine
because the whole point of hugepages is that EPT/NPT will also use a huge
pmd when they notice gup returns pages with PageCompound set, so they
won't care of a range and there's just the pmd young bit to check in that
case.
NOTE: in some cases if the L2 cache is small, this may slowdown and waste
memory during COWs because 4M of memory are accessed in a single fault
instead of 8k (the payoff is that after COW the program can run faster).
So we might want to switch the copy_huge_page (and clear_huge_page too) to
not temporal stores. I also extensively researched ways to avoid this
cache trashing with a full prefault logic that would cow in 8k/16k/32k/64k
up to 1M (I can send those patches that fully implemented prefault) but I
concluded they're not worth it and they add an huge additional complexity
and they remove all tlb benefits until the full hugepage has been faulted
in, to save a little bit of memory and some cache during app startup, but
they still don't improve substantially the cache-trashing during startup
if the prefault happens in >4k chunks. One reason is that those 4k pte
entries copied are still mapped on a perfectly cache-colored hugepage, so
the trashing is the worst one can generate in those copies (cow of 4k page
copies aren't so well colored so they trashes less, but again this results
in software running faster after the page fault). Those prefault patches
allowed things like a pte where post-cow pages were local 4k regular anon
pages and the not-yet-cowed pte entries were pointing in the middle of
some hugepage mapped read-only. If it doesn't payoff substantially with
todays hardware it will payoff even less in the future with larger l2
caches, and the prefault logic would blot the VM a lot. If one is
emebdded transparent_hugepage can be disabled during boot with sysfs or
with the boot commandline parameter transparent_hugepage=0 (or
transparent_hugepage=2 to restrict hugepages inside madvise regions) that
will ensure not a single hugepage is allocated at boot time. It is simple
enough to just disable transparent hugepage globally and let transparent
hugepages be allocated selectively by applications in the MADV_HUGEPAGE
region (both at page fault time, and if enabled with the
collapse_huge_page too through the kernel daemon).
This patch supports only hugepages mapped in the pmd, archs that have
smaller hugepages will not fit in this patch alone. Also some archs like
power have certain tlb limits that prevents mixing different page size in
the same regions so they will not fit in this framework that requires
"graceful fallback" to basic PAGE_SIZE in case of physical memory
fragmentation. hugetlbfs remains a perfect fit for those because its
software limits happen to match the hardware limits. hugetlbfs also
remains a perfect fit for hugepage sizes like 1GByte that cannot be hoped
to be found not fragmented after a certain system uptime and that would be
very expensive to defragment with relocation, so requiring reservation.
hugetlbfs is the "reservation way", the point of transparent hugepages is
not to have any reservation at all and maximizing the use of cache and
hugepages at all times automatically.
Some performance result:
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largep
ages3
memset page fault 1566023
memset tlb miss 453854
memset second tlb miss 453321
random access tlb miss 41635
random access second tlb miss 41658
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largepages3
memset page fault 1566471
memset tlb miss 453375
memset second tlb miss 453320
random access tlb miss 41636
random access second tlb miss 41637
vmx andrea # ./largepages3
memset page fault 1566642
memset tlb miss 453417
memset second tlb miss 453313
random access tlb miss 41630
random access second tlb miss 41647
vmx andrea # ./largepages3
memset page fault 1566872
memset tlb miss 453418
memset second tlb miss 453315
random access tlb miss 41618
random access second tlb miss 41659
vmx andrea # echo 0 > /proc/sys/vm/transparent_hugepage
vmx andrea # ./largepages3
memset page fault 2182476
memset tlb miss 460305
memset second tlb miss 460179
random access tlb miss 44483
random access second tlb miss 44186
vmx andrea # ./largepages3
memset page fault 2182791
memset tlb miss 460742
memset second tlb miss 459962
random access tlb miss 43981
random access second tlb miss 43988
============
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#define SIZE (3UL*1024*1024*1024)
int main()
{
char *p = malloc(SIZE), *p2;
struct timeval before, after;
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset page fault %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
return 0;
}
============
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:46:52 +00:00
|
|
|
}
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
|
mm: thp: fix pmd_bad() triggering in code paths holding mmap_sem read mode
In some cases it may happen that pmd_none_or_clear_bad() is called with
the mmap_sem hold in read mode. In those cases the huge page faults can
allocate hugepmds under pmd_none_or_clear_bad() and that can trigger a
false positive from pmd_bad() that will not like to see a pmd
materializing as trans huge.
It's not khugepaged causing the problem, khugepaged holds the mmap_sem
in write mode (and all those sites must hold the mmap_sem in read mode
to prevent pagetables to go away from under them, during code review it
seems vm86 mode on 32bit kernels requires that too unless it's
restricted to 1 thread per process or UP builds). The race is only with
the huge pagefaults that can convert a pmd_none() into a
pmd_trans_huge().
Effectively all these pmd_none_or_clear_bad() sites running with
mmap_sem in read mode are somewhat speculative with the page faults, and
the result is always undefined when they run simultaneously. This is
probably why it wasn't common to run into this. For example if the
madvise(MADV_DONTNEED) runs zap_page_range() shortly before the page
fault, the hugepage will not be zapped, if the page fault runs first it
will be zapped.
Altering pmd_bad() not to error out if it finds hugepmds won't be enough
to fix this, because zap_pmd_range would then proceed to call
zap_pte_range (which would be incorrect if the pmd become a
pmd_trans_huge()).
The simplest way to fix this is to read the pmd in the local stack
(regardless of what we read, no need of actual CPU barriers, only
compiler barrier needed), and be sure it is not changing under the code
that computes its value. Even if the real pmd is changing under the
value we hold on the stack, we don't care. If we actually end up in
zap_pte_range it means the pmd was not none already and it was not huge,
and it can't become huge from under us (khugepaged locking explained
above).
All we need is to enforce that there is no way anymore that in a code
path like below, pmd_trans_huge can be false, but pmd_none_or_clear_bad
can run into a hugepmd. The overhead of a barrier() is just a compiler
tweak and should not be measurable (I only added it for THP builds). I
don't exclude different compiler versions may have prevented the race
too by caching the value of *pmd on the stack (that hasn't been
verified, but it wouldn't be impossible considering
pmd_none_or_clear_bad, pmd_bad, pmd_trans_huge, pmd_none are all inlines
and there's no external function called in between pmd_trans_huge and
pmd_none_or_clear_bad).
if (pmd_trans_huge(*pmd)) {
if (next-addr != HPAGE_PMD_SIZE) {
VM_BUG_ON(!rwsem_is_locked(&tlb->mm->mmap_sem));
split_huge_page_pmd(vma->vm_mm, pmd);
} else if (zap_huge_pmd(tlb, vma, pmd, addr))
continue;
/* fall through */
}
if (pmd_none_or_clear_bad(pmd))
Because this race condition could be exercised without special
privileges this was reported in CVE-2012-1179.
The race was identified and fully explained by Ulrich who debugged it.
I'm quoting his accurate explanation below, for reference.
====== start quote =======
mapcount 0 page_mapcount 1
kernel BUG at mm/huge_memory.c:1384!
At some point prior to the panic, a "bad pmd ..." message similar to the
following is logged on the console:
mm/memory.c:145: bad pmd ffff8800376e1f98(80000000314000e7).
The "bad pmd ..." message is logged by pmd_clear_bad() before it clears
the page's PMD table entry.
143 void pmd_clear_bad(pmd_t *pmd)
144 {
-> 145 pmd_ERROR(*pmd);
146 pmd_clear(pmd);
147 }
After the PMD table entry has been cleared, there is an inconsistency
between the actual number of PMD table entries that are mapping the page
and the page's map count (_mapcount field in struct page). When the page
is subsequently reclaimed, __split_huge_page() detects this inconsistency.
1381 if (mapcount != page_mapcount(page))
1382 printk(KERN_ERR "mapcount %d page_mapcount %d\n",
1383 mapcount, page_mapcount(page));
-> 1384 BUG_ON(mapcount != page_mapcount(page));
The root cause of the problem is a race of two threads in a multithreaded
process. Thread B incurs a page fault on a virtual address that has never
been accessed (PMD entry is zero) while Thread A is executing an madvise()
system call on a virtual address within the same 2 MB (huge page) range.
virtual address space
.---------------------.
| |
| |
.-|---------------------|
| | |
| | |<-- B(fault)
| | |
2 MB | |/////////////////////|-.
huge < |/////////////////////| > A(range)
page | |/////////////////////|-'
| | |
| | |
'-|---------------------|
| |
| |
'---------------------'
- Thread A is executing an madvise(..., MADV_DONTNEED) system call
on the virtual address range "A(range)" shown in the picture.
sys_madvise
// Acquire the semaphore in shared mode.
down_read(¤t->mm->mmap_sem)
...
madvise_vma
switch (behavior)
case MADV_DONTNEED:
madvise_dontneed
zap_page_range
unmap_vmas
unmap_page_range
zap_pud_range
zap_pmd_range
//
// Assume that this huge page has never been accessed.
// I.e. content of the PMD entry is zero (not mapped).
//
if (pmd_trans_huge(*pmd)) {
// We don't get here due to the above assumption.
}
//
// Assume that Thread B incurred a page fault and
.---------> // sneaks in here as shown below.
| //
| if (pmd_none_or_clear_bad(pmd))
| {
| if (unlikely(pmd_bad(*pmd)))
| pmd_clear_bad
| {
| pmd_ERROR
| // Log "bad pmd ..." message here.
| pmd_clear
| // Clear the page's PMD entry.
| // Thread B incremented the map count
| // in page_add_new_anon_rmap(), but
| // now the page is no longer mapped
| // by a PMD entry (-> inconsistency).
| }
| }
|
v
- Thread B is handling a page fault on virtual address "B(fault)" shown
in the picture.
...
do_page_fault
__do_page_fault
// Acquire the semaphore in shared mode.
down_read_trylock(&mm->mmap_sem)
...
handle_mm_fault
if (pmd_none(*pmd) && transparent_hugepage_enabled(vma))
// We get here due to the above assumption (PMD entry is zero).
do_huge_pmd_anonymous_page
alloc_hugepage_vma
// Allocate a new transparent huge page here.
...
__do_huge_pmd_anonymous_page
...
spin_lock(&mm->page_table_lock)
...
page_add_new_anon_rmap
// Here we increment the page's map count (starts at -1).
atomic_set(&page->_mapcount, 0)
set_pmd_at
// Here we set the page's PMD entry which will be cleared
// when Thread A calls pmd_clear_bad().
...
spin_unlock(&mm->page_table_lock)
The mmap_sem does not prevent the race because both threads are acquiring
it in shared mode (down_read). Thread B holds the page_table_lock while
the page's map count and PMD table entry are updated. However, Thread A
does not synchronize on that lock.
====== end quote =======
[akpm@linux-foundation.org: checkpatch fixes]
Reported-by: Ulrich Obergfell <uobergfe@redhat.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Jones <davej@redhat.com>
Acked-by: Larry Woodman <lwoodman@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org> [2.6.38+]
Cc: Mark Salter <msalter@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-03-21 23:33:42 +00:00
|
|
|
/*
|
|
|
|
* Here there can be other concurrent MADV_DONTNEED or
|
|
|
|
* trans huge page faults running, and if the pmd is
|
|
|
|
* none or trans huge it can change under us. This is
|
2020-06-09 04:33:54 +00:00
|
|
|
* because MADV_DONTNEED holds the mmap_lock in read
|
mm: thp: fix pmd_bad() triggering in code paths holding mmap_sem read mode
In some cases it may happen that pmd_none_or_clear_bad() is called with
the mmap_sem hold in read mode. In those cases the huge page faults can
allocate hugepmds under pmd_none_or_clear_bad() and that can trigger a
false positive from pmd_bad() that will not like to see a pmd
materializing as trans huge.
It's not khugepaged causing the problem, khugepaged holds the mmap_sem
in write mode (and all those sites must hold the mmap_sem in read mode
to prevent pagetables to go away from under them, during code review it
seems vm86 mode on 32bit kernels requires that too unless it's
restricted to 1 thread per process or UP builds). The race is only with
the huge pagefaults that can convert a pmd_none() into a
pmd_trans_huge().
Effectively all these pmd_none_or_clear_bad() sites running with
mmap_sem in read mode are somewhat speculative with the page faults, and
the result is always undefined when they run simultaneously. This is
probably why it wasn't common to run into this. For example if the
madvise(MADV_DONTNEED) runs zap_page_range() shortly before the page
fault, the hugepage will not be zapped, if the page fault runs first it
will be zapped.
Altering pmd_bad() not to error out if it finds hugepmds won't be enough
to fix this, because zap_pmd_range would then proceed to call
zap_pte_range (which would be incorrect if the pmd become a
pmd_trans_huge()).
The simplest way to fix this is to read the pmd in the local stack
(regardless of what we read, no need of actual CPU barriers, only
compiler barrier needed), and be sure it is not changing under the code
that computes its value. Even if the real pmd is changing under the
value we hold on the stack, we don't care. If we actually end up in
zap_pte_range it means the pmd was not none already and it was not huge,
and it can't become huge from under us (khugepaged locking explained
above).
All we need is to enforce that there is no way anymore that in a code
path like below, pmd_trans_huge can be false, but pmd_none_or_clear_bad
can run into a hugepmd. The overhead of a barrier() is just a compiler
tweak and should not be measurable (I only added it for THP builds). I
don't exclude different compiler versions may have prevented the race
too by caching the value of *pmd on the stack (that hasn't been
verified, but it wouldn't be impossible considering
pmd_none_or_clear_bad, pmd_bad, pmd_trans_huge, pmd_none are all inlines
and there's no external function called in between pmd_trans_huge and
pmd_none_or_clear_bad).
if (pmd_trans_huge(*pmd)) {
if (next-addr != HPAGE_PMD_SIZE) {
VM_BUG_ON(!rwsem_is_locked(&tlb->mm->mmap_sem));
split_huge_page_pmd(vma->vm_mm, pmd);
} else if (zap_huge_pmd(tlb, vma, pmd, addr))
continue;
/* fall through */
}
if (pmd_none_or_clear_bad(pmd))
Because this race condition could be exercised without special
privileges this was reported in CVE-2012-1179.
The race was identified and fully explained by Ulrich who debugged it.
I'm quoting his accurate explanation below, for reference.
====== start quote =======
mapcount 0 page_mapcount 1
kernel BUG at mm/huge_memory.c:1384!
At some point prior to the panic, a "bad pmd ..." message similar to the
following is logged on the console:
mm/memory.c:145: bad pmd ffff8800376e1f98(80000000314000e7).
The "bad pmd ..." message is logged by pmd_clear_bad() before it clears
the page's PMD table entry.
143 void pmd_clear_bad(pmd_t *pmd)
144 {
-> 145 pmd_ERROR(*pmd);
146 pmd_clear(pmd);
147 }
After the PMD table entry has been cleared, there is an inconsistency
between the actual number of PMD table entries that are mapping the page
and the page's map count (_mapcount field in struct page). When the page
is subsequently reclaimed, __split_huge_page() detects this inconsistency.
1381 if (mapcount != page_mapcount(page))
1382 printk(KERN_ERR "mapcount %d page_mapcount %d\n",
1383 mapcount, page_mapcount(page));
-> 1384 BUG_ON(mapcount != page_mapcount(page));
The root cause of the problem is a race of two threads in a multithreaded
process. Thread B incurs a page fault on a virtual address that has never
been accessed (PMD entry is zero) while Thread A is executing an madvise()
system call on a virtual address within the same 2 MB (huge page) range.
virtual address space
.---------------------.
| |
| |
.-|---------------------|
| | |
| | |<-- B(fault)
| | |
2 MB | |/////////////////////|-.
huge < |/////////////////////| > A(range)
page | |/////////////////////|-'
| | |
| | |
'-|---------------------|
| |
| |
'---------------------'
- Thread A is executing an madvise(..., MADV_DONTNEED) system call
on the virtual address range "A(range)" shown in the picture.
sys_madvise
// Acquire the semaphore in shared mode.
down_read(¤t->mm->mmap_sem)
...
madvise_vma
switch (behavior)
case MADV_DONTNEED:
madvise_dontneed
zap_page_range
unmap_vmas
unmap_page_range
zap_pud_range
zap_pmd_range
//
// Assume that this huge page has never been accessed.
// I.e. content of the PMD entry is zero (not mapped).
//
if (pmd_trans_huge(*pmd)) {
// We don't get here due to the above assumption.
}
//
// Assume that Thread B incurred a page fault and
.---------> // sneaks in here as shown below.
| //
| if (pmd_none_or_clear_bad(pmd))
| {
| if (unlikely(pmd_bad(*pmd)))
| pmd_clear_bad
| {
| pmd_ERROR
| // Log "bad pmd ..." message here.
| pmd_clear
| // Clear the page's PMD entry.
| // Thread B incremented the map count
| // in page_add_new_anon_rmap(), but
| // now the page is no longer mapped
| // by a PMD entry (-> inconsistency).
| }
| }
|
v
- Thread B is handling a page fault on virtual address "B(fault)" shown
in the picture.
...
do_page_fault
__do_page_fault
// Acquire the semaphore in shared mode.
down_read_trylock(&mm->mmap_sem)
...
handle_mm_fault
if (pmd_none(*pmd) && transparent_hugepage_enabled(vma))
// We get here due to the above assumption (PMD entry is zero).
do_huge_pmd_anonymous_page
alloc_hugepage_vma
// Allocate a new transparent huge page here.
...
__do_huge_pmd_anonymous_page
...
spin_lock(&mm->page_table_lock)
...
page_add_new_anon_rmap
// Here we increment the page's map count (starts at -1).
atomic_set(&page->_mapcount, 0)
set_pmd_at
// Here we set the page's PMD entry which will be cleared
// when Thread A calls pmd_clear_bad().
...
spin_unlock(&mm->page_table_lock)
The mmap_sem does not prevent the race because both threads are acquiring
it in shared mode (down_read). Thread B holds the page_table_lock while
the page's map count and PMD table entry are updated. However, Thread A
does not synchronize on that lock.
====== end quote =======
[akpm@linux-foundation.org: checkpatch fixes]
Reported-by: Ulrich Obergfell <uobergfe@redhat.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Jones <davej@redhat.com>
Acked-by: Larry Woodman <lwoodman@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org> [2.6.38+]
Cc: Mark Salter <msalter@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-03-21 23:33:42 +00:00
|
|
|
* mode.
|
|
|
|
*/
|
|
|
|
if (pmd_none_or_trans_huge_or_clear_bad(pmd))
|
|
|
|
goto next;
|
2011-05-25 00:12:04 +00:00
|
|
|
next = zap_pte_range(tlb, vma, pmd, addr, next, details);
|
mm: thp: fix pmd_bad() triggering in code paths holding mmap_sem read mode
In some cases it may happen that pmd_none_or_clear_bad() is called with
the mmap_sem hold in read mode. In those cases the huge page faults can
allocate hugepmds under pmd_none_or_clear_bad() and that can trigger a
false positive from pmd_bad() that will not like to see a pmd
materializing as trans huge.
It's not khugepaged causing the problem, khugepaged holds the mmap_sem
in write mode (and all those sites must hold the mmap_sem in read mode
to prevent pagetables to go away from under them, during code review it
seems vm86 mode on 32bit kernels requires that too unless it's
restricted to 1 thread per process or UP builds). The race is only with
the huge pagefaults that can convert a pmd_none() into a
pmd_trans_huge().
Effectively all these pmd_none_or_clear_bad() sites running with
mmap_sem in read mode are somewhat speculative with the page faults, and
the result is always undefined when they run simultaneously. This is
probably why it wasn't common to run into this. For example if the
madvise(MADV_DONTNEED) runs zap_page_range() shortly before the page
fault, the hugepage will not be zapped, if the page fault runs first it
will be zapped.
Altering pmd_bad() not to error out if it finds hugepmds won't be enough
to fix this, because zap_pmd_range would then proceed to call
zap_pte_range (which would be incorrect if the pmd become a
pmd_trans_huge()).
The simplest way to fix this is to read the pmd in the local stack
(regardless of what we read, no need of actual CPU barriers, only
compiler barrier needed), and be sure it is not changing under the code
that computes its value. Even if the real pmd is changing under the
value we hold on the stack, we don't care. If we actually end up in
zap_pte_range it means the pmd was not none already and it was not huge,
and it can't become huge from under us (khugepaged locking explained
above).
All we need is to enforce that there is no way anymore that in a code
path like below, pmd_trans_huge can be false, but pmd_none_or_clear_bad
can run into a hugepmd. The overhead of a barrier() is just a compiler
tweak and should not be measurable (I only added it for THP builds). I
don't exclude different compiler versions may have prevented the race
too by caching the value of *pmd on the stack (that hasn't been
verified, but it wouldn't be impossible considering
pmd_none_or_clear_bad, pmd_bad, pmd_trans_huge, pmd_none are all inlines
and there's no external function called in between pmd_trans_huge and
pmd_none_or_clear_bad).
if (pmd_trans_huge(*pmd)) {
if (next-addr != HPAGE_PMD_SIZE) {
VM_BUG_ON(!rwsem_is_locked(&tlb->mm->mmap_sem));
split_huge_page_pmd(vma->vm_mm, pmd);
} else if (zap_huge_pmd(tlb, vma, pmd, addr))
continue;
/* fall through */
}
if (pmd_none_or_clear_bad(pmd))
Because this race condition could be exercised without special
privileges this was reported in CVE-2012-1179.
The race was identified and fully explained by Ulrich who debugged it.
I'm quoting his accurate explanation below, for reference.
====== start quote =======
mapcount 0 page_mapcount 1
kernel BUG at mm/huge_memory.c:1384!
At some point prior to the panic, a "bad pmd ..." message similar to the
following is logged on the console:
mm/memory.c:145: bad pmd ffff8800376e1f98(80000000314000e7).
The "bad pmd ..." message is logged by pmd_clear_bad() before it clears
the page's PMD table entry.
143 void pmd_clear_bad(pmd_t *pmd)
144 {
-> 145 pmd_ERROR(*pmd);
146 pmd_clear(pmd);
147 }
After the PMD table entry has been cleared, there is an inconsistency
between the actual number of PMD table entries that are mapping the page
and the page's map count (_mapcount field in struct page). When the page
is subsequently reclaimed, __split_huge_page() detects this inconsistency.
1381 if (mapcount != page_mapcount(page))
1382 printk(KERN_ERR "mapcount %d page_mapcount %d\n",
1383 mapcount, page_mapcount(page));
-> 1384 BUG_ON(mapcount != page_mapcount(page));
The root cause of the problem is a race of two threads in a multithreaded
process. Thread B incurs a page fault on a virtual address that has never
been accessed (PMD entry is zero) while Thread A is executing an madvise()
system call on a virtual address within the same 2 MB (huge page) range.
virtual address space
.---------------------.
| |
| |
.-|---------------------|
| | |
| | |<-- B(fault)
| | |
2 MB | |/////////////////////|-.
huge < |/////////////////////| > A(range)
page | |/////////////////////|-'
| | |
| | |
'-|---------------------|
| |
| |
'---------------------'
- Thread A is executing an madvise(..., MADV_DONTNEED) system call
on the virtual address range "A(range)" shown in the picture.
sys_madvise
// Acquire the semaphore in shared mode.
down_read(¤t->mm->mmap_sem)
...
madvise_vma
switch (behavior)
case MADV_DONTNEED:
madvise_dontneed
zap_page_range
unmap_vmas
unmap_page_range
zap_pud_range
zap_pmd_range
//
// Assume that this huge page has never been accessed.
// I.e. content of the PMD entry is zero (not mapped).
//
if (pmd_trans_huge(*pmd)) {
// We don't get here due to the above assumption.
}
//
// Assume that Thread B incurred a page fault and
.---------> // sneaks in here as shown below.
| //
| if (pmd_none_or_clear_bad(pmd))
| {
| if (unlikely(pmd_bad(*pmd)))
| pmd_clear_bad
| {
| pmd_ERROR
| // Log "bad pmd ..." message here.
| pmd_clear
| // Clear the page's PMD entry.
| // Thread B incremented the map count
| // in page_add_new_anon_rmap(), but
| // now the page is no longer mapped
| // by a PMD entry (-> inconsistency).
| }
| }
|
v
- Thread B is handling a page fault on virtual address "B(fault)" shown
in the picture.
...
do_page_fault
__do_page_fault
// Acquire the semaphore in shared mode.
down_read_trylock(&mm->mmap_sem)
...
handle_mm_fault
if (pmd_none(*pmd) && transparent_hugepage_enabled(vma))
// We get here due to the above assumption (PMD entry is zero).
do_huge_pmd_anonymous_page
alloc_hugepage_vma
// Allocate a new transparent huge page here.
...
__do_huge_pmd_anonymous_page
...
spin_lock(&mm->page_table_lock)
...
page_add_new_anon_rmap
// Here we increment the page's map count (starts at -1).
atomic_set(&page->_mapcount, 0)
set_pmd_at
// Here we set the page's PMD entry which will be cleared
// when Thread A calls pmd_clear_bad().
...
spin_unlock(&mm->page_table_lock)
The mmap_sem does not prevent the race because both threads are acquiring
it in shared mode (down_read). Thread B holds the page_table_lock while
the page's map count and PMD table entry are updated. However, Thread A
does not synchronize on that lock.
====== end quote =======
[akpm@linux-foundation.org: checkpatch fixes]
Reported-by: Ulrich Obergfell <uobergfe@redhat.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Jones <davej@redhat.com>
Acked-by: Larry Woodman <lwoodman@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org> [2.6.38+]
Cc: Mark Salter <msalter@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-03-21 23:33:42 +00:00
|
|
|
next:
|
2011-05-25 00:12:04 +00:00
|
|
|
cond_resched();
|
|
|
|
} while (pmd++, addr = next, addr != end);
|
2005-11-14 00:06:42 +00:00
|
|
|
|
|
|
|
return addr;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2005-11-14 00:06:42 +00:00
|
|
|
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
|
2017-03-09 14:24:07 +00:00
|
|
|
struct vm_area_struct *vma, p4d_t *p4d,
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long addr, unsigned long end,
|
2011-05-25 00:12:04 +00:00
|
|
|
struct zap_details *details)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pud_t *pud;
|
|
|
|
unsigned long next;
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
pud = pud_offset(p4d, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
do {
|
|
|
|
next = pud_addr_end(addr, end);
|
2017-02-24 22:57:02 +00:00
|
|
|
if (pud_trans_huge(*pud) || pud_devmap(*pud)) {
|
|
|
|
if (next - addr != HPAGE_PUD_SIZE) {
|
2020-06-09 04:33:44 +00:00
|
|
|
mmap_assert_locked(tlb->mm);
|
2017-02-24 22:57:02 +00:00
|
|
|
split_huge_pud(vma, pud, addr);
|
|
|
|
} else if (zap_huge_pud(tlb, vma, pud, addr))
|
|
|
|
goto next;
|
|
|
|
/* fall through */
|
|
|
|
}
|
2011-05-25 00:12:04 +00:00
|
|
|
if (pud_none_or_clear_bad(pud))
|
2005-04-16 22:20:36 +00:00
|
|
|
continue;
|
2011-05-25 00:12:04 +00:00
|
|
|
next = zap_pmd_range(tlb, vma, pud, addr, next, details);
|
2017-02-24 22:57:02 +00:00
|
|
|
next:
|
|
|
|
cond_resched();
|
2011-05-25 00:12:04 +00:00
|
|
|
} while (pud++, addr = next, addr != end);
|
2005-11-14 00:06:42 +00:00
|
|
|
|
|
|
|
return addr;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
static inline unsigned long zap_p4d_range(struct mmu_gather *tlb,
|
|
|
|
struct vm_area_struct *vma, pgd_t *pgd,
|
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
struct zap_details *details)
|
|
|
|
{
|
|
|
|
p4d_t *p4d;
|
|
|
|
unsigned long next;
|
|
|
|
|
|
|
|
p4d = p4d_offset(pgd, addr);
|
|
|
|
do {
|
|
|
|
next = p4d_addr_end(addr, end);
|
|
|
|
if (p4d_none_or_clear_bad(p4d))
|
|
|
|
continue;
|
|
|
|
next = zap_pud_range(tlb, vma, p4d, addr, next, details);
|
|
|
|
} while (p4d++, addr = next, addr != end);
|
|
|
|
|
|
|
|
return addr;
|
|
|
|
}
|
|
|
|
|
2016-03-25 21:20:24 +00:00
|
|
|
void unmap_page_range(struct mmu_gather *tlb,
|
2012-03-05 18:25:09 +00:00
|
|
|
struct vm_area_struct *vma,
|
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
struct zap_details *details)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pgd_t *pgd;
|
|
|
|
unsigned long next;
|
|
|
|
|
|
|
|
BUG_ON(addr >= end);
|
|
|
|
tlb_start_vma(tlb, vma);
|
|
|
|
pgd = pgd_offset(vma->vm_mm, addr);
|
|
|
|
do {
|
|
|
|
next = pgd_addr_end(addr, end);
|
2011-05-25 00:12:04 +00:00
|
|
|
if (pgd_none_or_clear_bad(pgd))
|
2005-04-16 22:20:36 +00:00
|
|
|
continue;
|
2017-03-09 14:24:07 +00:00
|
|
|
next = zap_p4d_range(tlb, vma, pgd, addr, next, details);
|
2011-05-25 00:12:04 +00:00
|
|
|
} while (pgd++, addr = next, addr != end);
|
2005-04-16 22:20:36 +00:00
|
|
|
tlb_end_vma(tlb, vma);
|
|
|
|
}
|
2005-11-14 00:06:42 +00:00
|
|
|
|
2012-03-05 19:14:20 +00:00
|
|
|
|
|
|
|
static void unmap_single_vma(struct mmu_gather *tlb,
|
|
|
|
struct vm_area_struct *vma, unsigned long start_addr,
|
2012-05-06 20:54:06 +00:00
|
|
|
unsigned long end_addr,
|
2023-01-26 19:37:51 +00:00
|
|
|
struct zap_details *details, bool mm_wr_locked)
|
2012-03-05 19:14:20 +00:00
|
|
|
{
|
|
|
|
unsigned long start = max(vma->vm_start, start_addr);
|
|
|
|
unsigned long end;
|
|
|
|
|
|
|
|
if (start >= vma->vm_end)
|
|
|
|
return;
|
|
|
|
end = min(vma->vm_end, end_addr);
|
|
|
|
if (end <= vma->vm_start)
|
|
|
|
return;
|
|
|
|
|
2012-04-11 10:35:27 +00:00
|
|
|
if (vma->vm_file)
|
|
|
|
uprobe_munmap(vma, start, end);
|
|
|
|
|
2012-10-08 23:28:34 +00:00
|
|
|
if (unlikely(vma->vm_flags & VM_PFNMAP))
|
2023-01-26 19:37:51 +00:00
|
|
|
untrack_pfn(vma, 0, 0, mm_wr_locked);
|
2012-03-05 19:14:20 +00:00
|
|
|
|
|
|
|
if (start != end) {
|
|
|
|
if (unlikely(is_vm_hugetlb_page(vma))) {
|
|
|
|
/*
|
|
|
|
* It is undesirable to test vma->vm_file as it
|
|
|
|
* should be non-null for valid hugetlb area.
|
|
|
|
* However, vm_file will be NULL in the error
|
2014-04-07 22:37:01 +00:00
|
|
|
* cleanup path of mmap_region. When
|
2012-03-05 19:14:20 +00:00
|
|
|
* hugetlbfs ->mmap method fails,
|
2014-04-07 22:37:01 +00:00
|
|
|
* mmap_region() nullifies vma->vm_file
|
2012-03-05 19:14:20 +00:00
|
|
|
* before calling this function to clean up.
|
|
|
|
* Since no pte has actually been setup, it is
|
|
|
|
* safe to do nothing in this case.
|
|
|
|
*/
|
2012-07-31 23:42:03 +00:00
|
|
|
if (vma->vm_file) {
|
2022-05-13 03:22:55 +00:00
|
|
|
zap_flags_t zap_flags = details ?
|
|
|
|
details->zap_flags : 0;
|
|
|
|
__unmap_hugepage_range_final(tlb, vma, start, end,
|
|
|
|
NULL, zap_flags);
|
2012-07-31 23:42:03 +00:00
|
|
|
}
|
2012-03-05 19:14:20 +00:00
|
|
|
} else
|
|
|
|
unmap_page_range(tlb, vma, start, end, details);
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* unmap_vmas - unmap a range of memory covered by a list of vma's
|
2011-06-15 22:08:09 +00:00
|
|
|
* @tlb: address of the caller's struct mmu_gather
|
2022-09-06 19:49:06 +00:00
|
|
|
* @mt: the maple tree
|
2005-04-16 22:20:36 +00:00
|
|
|
* @vma: the starting vma
|
|
|
|
* @start_addr: virtual address at which to start unmapping
|
|
|
|
* @end_addr: virtual address at which to end unmapping
|
|
|
|
*
|
2005-10-30 01:16:30 +00:00
|
|
|
* Unmap all pages in the vma list.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
|
|
|
* Only addresses between `start' and `end' will be unmapped.
|
|
|
|
*
|
|
|
|
* The VMA list must be sorted in ascending virtual address order.
|
|
|
|
*
|
|
|
|
* unmap_vmas() assumes that the caller will flush the whole unmapped address
|
|
|
|
* range after unmap_vmas() returns. So the only responsibility here is to
|
|
|
|
* ensure that any thus-far unmapped pages are flushed before unmap_vmas()
|
|
|
|
* drops the lock and schedules.
|
|
|
|
*/
|
2022-09-06 19:49:06 +00:00
|
|
|
void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
|
2005-04-16 22:20:36 +00:00
|
|
|
struct vm_area_struct *vma, unsigned long start_addr,
|
2023-01-26 19:37:51 +00:00
|
|
|
unsigned long end_addr, bool mm_wr_locked)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2018-12-28 08:38:09 +00:00
|
|
|
struct mmu_notifier_range range;
|
2022-05-13 03:22:53 +00:00
|
|
|
struct zap_details details = {
|
2022-11-14 23:55:06 +00:00
|
|
|
.zap_flags = ZAP_FLAG_DROP_MARKER | ZAP_FLAG_UNMAP,
|
2022-05-13 03:22:53 +00:00
|
|
|
/* Careful - we need to zap private pages too! */
|
|
|
|
.even_cows = true,
|
|
|
|
};
|
2022-09-06 19:49:06 +00:00
|
|
|
MA_STATE(mas, mt, vma->vm_end, vma->vm_end);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2023-01-10 02:57:22 +00:00
|
|
|
mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma->vm_mm,
|
mm/mmu_notifier: contextual information for event triggering invalidation
CPU page table update can happens for many reasons, not only as a result
of a syscall (munmap(), mprotect(), mremap(), madvise(), ...) but also as
a result of kernel activities (memory compression, reclaim, migration,
...).
Users of mmu notifier API track changes to the CPU page table and take
specific action for them. While current API only provide range of virtual
address affected by the change, not why the changes is happening.
This patchset do the initial mechanical convertion of all the places that
calls mmu_notifier_range_init to also provide the default MMU_NOTIFY_UNMAP
event as well as the vma if it is know (most invalidation happens against
a given vma). Passing down the vma allows the users of mmu notifier to
inspect the new vma page protection.
The MMU_NOTIFY_UNMAP is always the safe default as users of mmu notifier
should assume that every for the range is going away when that event
happens. A latter patch do convert mm call path to use a more appropriate
events for each call.
This is done as 2 patches so that no call site is forgotten especialy
as it uses this following coccinelle patch:
%<----------------------------------------------------------------------
@@
identifier I1, I2, I3, I4;
@@
static inline void mmu_notifier_range_init(struct mmu_notifier_range *I1,
+enum mmu_notifier_event event,
+unsigned flags,
+struct vm_area_struct *vma,
struct mm_struct *I2, unsigned long I3, unsigned long I4) { ... }
@@
@@
-#define mmu_notifier_range_init(range, mm, start, end)
+#define mmu_notifier_range_init(range, event, flags, vma, mm, start, end)
@@
expression E1, E3, E4;
identifier I1;
@@
<...
mmu_notifier_range_init(E1,
+MMU_NOTIFY_UNMAP, 0, I1,
I1->vm_mm, E3, E4)
...>
@@
expression E1, E2, E3, E4;
identifier FN, VMA;
@@
FN(..., struct vm_area_struct *VMA, ...) {
<...
mmu_notifier_range_init(E1,
+MMU_NOTIFY_UNMAP, 0, VMA,
E2, E3, E4)
...> }
@@
expression E1, E2, E3, E4;
identifier FN, VMA;
@@
FN(...) {
struct vm_area_struct *VMA;
<...
mmu_notifier_range_init(E1,
+MMU_NOTIFY_UNMAP, 0, VMA,
E2, E3, E4)
...> }
@@
expression E1, E2, E3, E4;
identifier FN;
@@
FN(...) {
<...
mmu_notifier_range_init(E1,
+MMU_NOTIFY_UNMAP, 0, NULL,
E2, E3, E4)
...> }
---------------------------------------------------------------------->%
Applied with:
spatch --all-includes --sp-file mmu-notifier.spatch fs/proc/task_mmu.c --in-place
spatch --sp-file mmu-notifier.spatch --dir kernel/events/ --in-place
spatch --sp-file mmu-notifier.spatch --dir mm --in-place
Link: http://lkml.kernel.org/r/20190326164747.24405-6-jglisse@redhat.com
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
Reviewed-by: Ralph Campbell <rcampbell@nvidia.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Ross Zwisler <zwisler@kernel.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krcmar <rkrcmar@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 00:20:49 +00:00
|
|
|
start_addr, end_addr);
|
2018-12-28 08:38:09 +00:00
|
|
|
mmu_notifier_invalidate_range_start(&range);
|
2022-09-06 19:49:06 +00:00
|
|
|
do {
|
2023-01-26 19:37:51 +00:00
|
|
|
unmap_single_vma(tlb, vma, start_addr, end_addr, &details,
|
|
|
|
mm_wr_locked);
|
2022-09-06 19:49:06 +00:00
|
|
|
} while ((vma = mas_find(&mas, end_addr - 1)) != NULL);
|
2018-12-28 08:38:09 +00:00
|
|
|
mmu_notifier_invalidate_range_end(&range);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2012-03-05 19:14:20 +00:00
|
|
|
/**
|
|
|
|
* zap_page_range_single - remove user pages in a given range
|
|
|
|
* @vma: vm_area_struct holding the applicable pages
|
|
|
|
* @address: starting address of pages to zap
|
|
|
|
* @size: number of bytes to zap
|
2015-02-10 22:09:49 +00:00
|
|
|
* @details: details of shared cache invalidation
|
2012-03-05 19:14:20 +00:00
|
|
|
*
|
|
|
|
* The range must fit into one VMA.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2022-11-14 23:55:05 +00:00
|
|
|
void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long size, struct zap_details *details)
|
|
|
|
{
|
2022-11-14 23:55:05 +00:00
|
|
|
const unsigned long end = address + size;
|
2018-12-28 08:38:09 +00:00
|
|
|
struct mmu_notifier_range range;
|
2011-05-25 00:11:45 +00:00
|
|
|
struct mmu_gather tlb;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
lru_add_drain();
|
2023-01-10 02:57:22 +00:00
|
|
|
mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma->vm_mm,
|
2022-11-14 23:55:05 +00:00
|
|
|
address, end);
|
|
|
|
if (is_vm_hugetlb_page(vma))
|
|
|
|
adjust_range_if_pmd_sharing_possible(vma, &range.start,
|
|
|
|
&range.end);
|
2021-01-27 23:53:45 +00:00
|
|
|
tlb_gather_mmu(&tlb, vma->vm_mm);
|
2018-12-28 08:38:09 +00:00
|
|
|
update_hiwater_rss(vma->vm_mm);
|
|
|
|
mmu_notifier_invalidate_range_start(&range);
|
2022-11-14 23:55:05 +00:00
|
|
|
/*
|
|
|
|
* unmap 'address-end' not 'range.start-range.end' as range
|
|
|
|
* could have been expanded for hugetlb pmd sharing.
|
|
|
|
*/
|
2023-01-26 19:37:51 +00:00
|
|
|
unmap_single_vma(&tlb, vma, address, end, details, false);
|
2018-12-28 08:38:09 +00:00
|
|
|
mmu_notifier_invalidate_range_end(&range);
|
2021-01-27 23:53:43 +00:00
|
|
|
tlb_finish_mmu(&tlb);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2008-07-30 05:33:53 +00:00
|
|
|
/**
|
|
|
|
* zap_vma_ptes - remove ptes mapping the vma
|
|
|
|
* @vma: vm_area_struct holding ptes to be zapped
|
|
|
|
* @address: starting address of pages to zap
|
|
|
|
* @size: number of bytes to zap
|
|
|
|
*
|
|
|
|
* This function only unmaps ptes assigned to VM_PFNMAP vmas.
|
|
|
|
*
|
|
|
|
* The entire address range must be fully contained within the vma.
|
|
|
|
*
|
|
|
|
*/
|
2018-05-29 12:14:07 +00:00
|
|
|
void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
|
2008-07-30 05:33:53 +00:00
|
|
|
unsigned long size)
|
|
|
|
{
|
2022-03-22 21:42:30 +00:00
|
|
|
if (!range_in_vma(vma, address, address + size) ||
|
2008-07-30 05:33:53 +00:00
|
|
|
!(vma->vm_flags & VM_PFNMAP))
|
2018-05-29 12:14:07 +00:00
|
|
|
return;
|
|
|
|
|
2012-03-05 19:14:20 +00:00
|
|
|
zap_page_range_single(vma, address, size, NULL);
|
2008-07-30 05:33:53 +00:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(zap_vma_ptes);
|
|
|
|
|
2020-04-10 21:33:01 +00:00
|
|
|
static pmd_t *walk_to_pmd(struct mm_struct *mm, unsigned long addr)
|
2005-11-29 22:03:14 +00:00
|
|
|
{
|
2017-03-09 14:24:07 +00:00
|
|
|
pgd_t *pgd;
|
|
|
|
p4d_t *p4d;
|
|
|
|
pud_t *pud;
|
|
|
|
pmd_t *pmd;
|
|
|
|
|
|
|
|
pgd = pgd_offset(mm, addr);
|
|
|
|
p4d = p4d_alloc(mm, pgd, addr);
|
|
|
|
if (!p4d)
|
|
|
|
return NULL;
|
|
|
|
pud = pud_alloc(mm, p4d, addr);
|
|
|
|
if (!pud)
|
|
|
|
return NULL;
|
|
|
|
pmd = pmd_alloc(mm, pud, addr);
|
|
|
|
if (!pmd)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
VM_BUG_ON(pmd_trans_huge(*pmd));
|
2020-04-10 21:33:01 +00:00
|
|
|
return pmd;
|
|
|
|
}
|
|
|
|
|
|
|
|
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
|
|
|
|
spinlock_t **ptl)
|
|
|
|
{
|
|
|
|
pmd_t *pmd = walk_to_pmd(mm, addr);
|
|
|
|
|
|
|
|
if (!pmd)
|
|
|
|
return NULL;
|
2017-03-09 14:24:07 +00:00
|
|
|
return pte_alloc_map_lock(mm, pmd, addr, ptl);
|
2005-11-29 22:03:14 +00:00
|
|
|
}
|
|
|
|
|
2020-04-10 21:32:51 +00:00
|
|
|
static int validate_page_before_insert(struct page *page)
|
|
|
|
{
|
|
|
|
if (PageAnon(page) || PageSlab(page) || page_has_type(page))
|
|
|
|
return -EINVAL;
|
|
|
|
flush_dcache_page(page);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
static int insert_page_into_pte_locked(struct vm_area_struct *vma, pte_t *pte,
|
2020-04-10 21:32:51 +00:00
|
|
|
unsigned long addr, struct page *page, pgprot_t prot)
|
|
|
|
{
|
|
|
|
if (!pte_none(*pte))
|
|
|
|
return -EBUSY;
|
|
|
|
/* Ok, finally just insert the thing.. */
|
|
|
|
get_page(page);
|
2022-10-24 05:28:41 +00:00
|
|
|
inc_mm_counter(vma->vm_mm, mm_counter_file(page));
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
page_add_file_rmap(page, vma, false);
|
|
|
|
set_pte_at(vma->vm_mm, addr, pte, mk_pte(page, prot));
|
2020-04-10 21:32:51 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-11-29 21:01:56 +00:00
|
|
|
/*
|
|
|
|
* This is the old fallback for page remapping.
|
|
|
|
*
|
|
|
|
* For historical reasons, it only allows reserved pages. Only
|
|
|
|
* old drivers should use this, and they needed to mark their
|
|
|
|
* pages reserved for the old functions anyway.
|
|
|
|
*/
|
2008-04-28 09:13:01 +00:00
|
|
|
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
struct page *page, pgprot_t prot)
|
2005-11-29 21:01:56 +00:00
|
|
|
{
|
|
|
|
int retval;
|
2005-11-29 22:03:14 +00:00
|
|
|
pte_t *pte;
|
2008-02-07 08:13:53 +00:00
|
|
|
spinlock_t *ptl;
|
|
|
|
|
2020-04-10 21:32:51 +00:00
|
|
|
retval = validate_page_before_insert(page);
|
|
|
|
if (retval)
|
2008-10-19 03:28:10 +00:00
|
|
|
goto out;
|
2005-11-29 21:01:56 +00:00
|
|
|
retval = -ENOMEM;
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
pte = get_locked_pte(vma->vm_mm, addr, &ptl);
|
2005-11-29 21:01:56 +00:00
|
|
|
if (!pte)
|
2008-10-19 03:28:10 +00:00
|
|
|
goto out;
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
retval = insert_page_into_pte_locked(vma, pte, addr, page, prot);
|
2005-11-29 21:01:56 +00:00
|
|
|
pte_unmap_unlock(pte, ptl);
|
|
|
|
out:
|
|
|
|
return retval;
|
|
|
|
}
|
|
|
|
|
2020-04-10 21:33:01 +00:00
|
|
|
#ifdef pte_index
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
static int insert_page_in_batch_locked(struct vm_area_struct *vma, pte_t *pte,
|
2020-04-10 21:33:01 +00:00
|
|
|
unsigned long addr, struct page *page, pgprot_t prot)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (!page_count(page))
|
|
|
|
return -EINVAL;
|
|
|
|
err = validate_page_before_insert(page);
|
2020-06-26 03:30:01 +00:00
|
|
|
if (err)
|
|
|
|
return err;
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
return insert_page_into_pte_locked(vma, pte, addr, page, prot);
|
2020-04-10 21:33:01 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* insert_pages() amortizes the cost of spinlock operations
|
|
|
|
* when inserting pages in a loop. Arch *must* define pte_index.
|
|
|
|
*/
|
|
|
|
static int insert_pages(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
struct page **pages, unsigned long *num, pgprot_t prot)
|
|
|
|
{
|
|
|
|
pmd_t *pmd = NULL;
|
2020-06-26 03:30:01 +00:00
|
|
|
pte_t *start_pte, *pte;
|
|
|
|
spinlock_t *pte_lock;
|
2020-04-10 21:33:01 +00:00
|
|
|
struct mm_struct *const mm = vma->vm_mm;
|
|
|
|
unsigned long curr_page_idx = 0;
|
|
|
|
unsigned long remaining_pages_total = *num;
|
|
|
|
unsigned long pages_to_write_in_pmd;
|
|
|
|
int ret;
|
|
|
|
more:
|
|
|
|
ret = -EFAULT;
|
|
|
|
pmd = walk_to_pmd(mm, addr);
|
|
|
|
if (!pmd)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
pages_to_write_in_pmd = min_t(unsigned long,
|
|
|
|
remaining_pages_total, PTRS_PER_PTE - pte_index(addr));
|
|
|
|
|
|
|
|
/* Allocate the PTE if necessary; takes PMD lock once only. */
|
|
|
|
ret = -ENOMEM;
|
|
|
|
if (pte_alloc(mm, pmd))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
while (pages_to_write_in_pmd) {
|
|
|
|
int pte_idx = 0;
|
|
|
|
const int batch_size = min_t(int, pages_to_write_in_pmd, 8);
|
|
|
|
|
2020-06-26 03:30:01 +00:00
|
|
|
start_pte = pte_offset_map_lock(mm, pmd, addr, &pte_lock);
|
|
|
|
for (pte = start_pte; pte_idx < batch_size; ++pte, ++pte_idx) {
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
int err = insert_page_in_batch_locked(vma, pte,
|
2020-04-10 21:33:01 +00:00
|
|
|
addr, pages[curr_page_idx], prot);
|
|
|
|
if (unlikely(err)) {
|
2020-06-26 03:30:01 +00:00
|
|
|
pte_unmap_unlock(start_pte, pte_lock);
|
2020-04-10 21:33:01 +00:00
|
|
|
ret = err;
|
|
|
|
remaining_pages_total -= pte_idx;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
addr += PAGE_SIZE;
|
|
|
|
++curr_page_idx;
|
|
|
|
}
|
2020-06-26 03:30:01 +00:00
|
|
|
pte_unmap_unlock(start_pte, pte_lock);
|
2020-04-10 21:33:01 +00:00
|
|
|
pages_to_write_in_pmd -= batch_size;
|
|
|
|
remaining_pages_total -= batch_size;
|
|
|
|
}
|
|
|
|
if (remaining_pages_total)
|
|
|
|
goto more;
|
|
|
|
ret = 0;
|
|
|
|
out:
|
|
|
|
*num = remaining_pages_total;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
#endif /* ifdef pte_index */
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vm_insert_pages - insert multiple pages into user vma, batching the pmd lock.
|
|
|
|
* @vma: user vma to map to
|
|
|
|
* @addr: target start user address of these pages
|
|
|
|
* @pages: source kernel pages
|
|
|
|
* @num: in: number of pages to map. out: number of pages that were *not*
|
|
|
|
* mapped. (0 means all pages were successfully mapped).
|
|
|
|
*
|
|
|
|
* Preferred over vm_insert_page() when inserting multiple pages.
|
|
|
|
*
|
|
|
|
* In case of error, we may have mapped a subset of the provided
|
|
|
|
* pages. It is the caller's responsibility to account for this case.
|
|
|
|
*
|
|
|
|
* The same restrictions apply as in vm_insert_page().
|
|
|
|
*/
|
|
|
|
int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
struct page **pages, unsigned long *num)
|
|
|
|
{
|
|
|
|
#ifdef pte_index
|
|
|
|
const unsigned long end_addr = addr + (*num * PAGE_SIZE) - 1;
|
|
|
|
|
|
|
|
if (addr < vma->vm_start || end_addr >= vma->vm_end)
|
|
|
|
return -EFAULT;
|
|
|
|
if (!(vma->vm_flags & VM_MIXEDMAP)) {
|
2020-06-09 04:33:25 +00:00
|
|
|
BUG_ON(mmap_read_trylock(vma->vm_mm));
|
2020-04-10 21:33:01 +00:00
|
|
|
BUG_ON(vma->vm_flags & VM_PFNMAP);
|
2023-01-26 19:37:49 +00:00
|
|
|
vm_flags_set(vma, VM_MIXEDMAP);
|
2020-04-10 21:33:01 +00:00
|
|
|
}
|
|
|
|
/* Defer page refcount checking till we're about to map that page. */
|
|
|
|
return insert_pages(vma, addr, pages, num, vma->vm_page_prot);
|
|
|
|
#else
|
|
|
|
unsigned long idx = 0, pgcount = *num;
|
2020-07-24 04:15:18 +00:00
|
|
|
int err = -EINVAL;
|
2020-04-10 21:33:01 +00:00
|
|
|
|
|
|
|
for (; idx < pgcount; ++idx) {
|
|
|
|
err = vm_insert_page(vma, addr + (PAGE_SIZE * idx), pages[idx]);
|
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
*num = pgcount - idx;
|
|
|
|
return err;
|
|
|
|
#endif /* ifdef pte_index */
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(vm_insert_pages);
|
|
|
|
|
2006-09-26 06:31:22 +00:00
|
|
|
/**
|
|
|
|
* vm_insert_page - insert single page into user vma
|
|
|
|
* @vma: user vma to map to
|
|
|
|
* @addr: target user address of this page
|
|
|
|
* @page: source kernel page
|
|
|
|
*
|
2005-11-30 17:35:19 +00:00
|
|
|
* This allows drivers to insert individual pages they've allocated
|
|
|
|
* into a user vma.
|
|
|
|
*
|
|
|
|
* The page has to be a nice clean _individual_ kernel allocation.
|
|
|
|
* If you allocate a compound page, you need to have marked it as
|
|
|
|
* such (__GFP_COMP), or manually just split the page up yourself
|
2006-03-22 08:08:05 +00:00
|
|
|
* (see split_page()).
|
2005-11-30 17:35:19 +00:00
|
|
|
*
|
|
|
|
* NOTE! Traditionally this was done with "remap_pfn_range()" which
|
|
|
|
* took an arbitrary page protection parameter. This doesn't allow
|
|
|
|
* that. Your vma protection will have to be set up correctly, which
|
|
|
|
* means that if you want a shared writable mapping, you'd better
|
|
|
|
* ask for a shared writable mapping!
|
|
|
|
*
|
|
|
|
* The page does not need to be reserved.
|
2012-10-08 23:28:40 +00:00
|
|
|
*
|
|
|
|
* Usually this function is called from f_op->mmap() handler
|
2020-06-09 04:33:54 +00:00
|
|
|
* under mm->mmap_lock write-lock, so it can change vma->vm_flags.
|
2012-10-08 23:28:40 +00:00
|
|
|
* Caller must set VM_MIXEDMAP on vma if it wants to call this
|
|
|
|
* function from other places, for example from page-fault handler.
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: %0 on success, negative error code otherwise.
|
2005-11-30 17:35:19 +00:00
|
|
|
*/
|
2008-04-28 09:13:01 +00:00
|
|
|
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
struct page *page)
|
2005-11-30 17:35:19 +00:00
|
|
|
{
|
|
|
|
if (addr < vma->vm_start || addr >= vma->vm_end)
|
|
|
|
return -EFAULT;
|
|
|
|
if (!page_count(page))
|
|
|
|
return -EINVAL;
|
2012-10-08 23:28:40 +00:00
|
|
|
if (!(vma->vm_flags & VM_MIXEDMAP)) {
|
2020-06-09 04:33:25 +00:00
|
|
|
BUG_ON(mmap_read_trylock(vma->vm_mm));
|
2012-10-08 23:28:40 +00:00
|
|
|
BUG_ON(vma->vm_flags & VM_PFNMAP);
|
2023-01-26 19:37:49 +00:00
|
|
|
vm_flags_set(vma, VM_MIXEDMAP);
|
2012-10-08 23:28:40 +00:00
|
|
|
}
|
2008-04-28 09:13:01 +00:00
|
|
|
return insert_page(vma, addr, page, vma->vm_page_prot);
|
2005-11-30 17:35:19 +00:00
|
|
|
}
|
2005-12-04 04:48:11 +00:00
|
|
|
EXPORT_SYMBOL(vm_insert_page);
|
2005-11-30 17:35:19 +00:00
|
|
|
|
2019-05-14 00:21:56 +00:00
|
|
|
/*
|
|
|
|
* __vm_map_pages - maps range of kernel pages into user vma
|
|
|
|
* @vma: user vma to map to
|
|
|
|
* @pages: pointer to array of source kernel pages
|
|
|
|
* @num: number of pages in page array
|
|
|
|
* @offset: user's requested vm_pgoff
|
|
|
|
*
|
|
|
|
* This allows drivers to map range of kernel pages into a user vma.
|
|
|
|
*
|
|
|
|
* Return: 0 on success and error code otherwise.
|
|
|
|
*/
|
|
|
|
static int __vm_map_pages(struct vm_area_struct *vma, struct page **pages,
|
|
|
|
unsigned long num, unsigned long offset)
|
|
|
|
{
|
|
|
|
unsigned long count = vma_pages(vma);
|
|
|
|
unsigned long uaddr = vma->vm_start;
|
|
|
|
int ret, i;
|
|
|
|
|
|
|
|
/* Fail if the user requested offset is beyond the end of the object */
|
2019-07-12 03:58:47 +00:00
|
|
|
if (offset >= num)
|
2019-05-14 00:21:56 +00:00
|
|
|
return -ENXIO;
|
|
|
|
|
|
|
|
/* Fail if the user requested size exceeds available object size */
|
|
|
|
if (count > num - offset)
|
|
|
|
return -ENXIO;
|
|
|
|
|
|
|
|
for (i = 0; i < count; i++) {
|
|
|
|
ret = vm_insert_page(vma, uaddr, pages[offset + i]);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
uaddr += PAGE_SIZE;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vm_map_pages - maps range of kernel pages starts with non zero offset
|
|
|
|
* @vma: user vma to map to
|
|
|
|
* @pages: pointer to array of source kernel pages
|
|
|
|
* @num: number of pages in page array
|
|
|
|
*
|
|
|
|
* Maps an object consisting of @num pages, catering for the user's
|
|
|
|
* requested vm_pgoff
|
|
|
|
*
|
|
|
|
* If we fail to insert any page into the vma, the function will return
|
|
|
|
* immediately leaving any previously inserted pages present. Callers
|
|
|
|
* from the mmap handler may immediately return the error as their caller
|
|
|
|
* will destroy the vma, removing any successfully inserted pages. Other
|
|
|
|
* callers should make their own arrangements for calling unmap_region().
|
|
|
|
*
|
|
|
|
* Context: Process context. Called by mmap handlers.
|
|
|
|
* Return: 0 on success and error code otherwise.
|
|
|
|
*/
|
|
|
|
int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
|
|
|
|
unsigned long num)
|
|
|
|
{
|
|
|
|
return __vm_map_pages(vma, pages, num, vma->vm_pgoff);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(vm_map_pages);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vm_map_pages_zero - map range of kernel pages starts with zero offset
|
|
|
|
* @vma: user vma to map to
|
|
|
|
* @pages: pointer to array of source kernel pages
|
|
|
|
* @num: number of pages in page array
|
|
|
|
*
|
|
|
|
* Similar to vm_map_pages(), except that it explicitly sets the offset
|
|
|
|
* to 0. This function is intended for the drivers that did not consider
|
|
|
|
* vm_pgoff.
|
|
|
|
*
|
|
|
|
* Context: Process context. Called by mmap handlers.
|
|
|
|
* Return: 0 on success and error code otherwise.
|
|
|
|
*/
|
|
|
|
int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
|
|
|
|
unsigned long num)
|
|
|
|
{
|
|
|
|
return __vm_map_pages(vma, pages, num, 0);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(vm_map_pages_zero);
|
|
|
|
|
2018-10-26 22:04:40 +00:00
|
|
|
static vm_fault_t insert_pfn(struct vm_area_struct *vma, unsigned long addr,
|
2017-09-06 23:18:35 +00:00
|
|
|
pfn_t pfn, pgprot_t prot, bool mkwrite)
|
2008-04-28 09:13:01 +00:00
|
|
|
{
|
|
|
|
struct mm_struct *mm = vma->vm_mm;
|
|
|
|
pte_t *pte, entry;
|
|
|
|
spinlock_t *ptl;
|
|
|
|
|
|
|
|
pte = get_locked_pte(mm, addr, &ptl);
|
|
|
|
if (!pte)
|
2018-10-26 22:04:40 +00:00
|
|
|
return VM_FAULT_OOM;
|
2017-09-06 23:18:35 +00:00
|
|
|
if (!pte_none(*pte)) {
|
|
|
|
if (mkwrite) {
|
|
|
|
/*
|
|
|
|
* For read faults on private mappings the PFN passed
|
|
|
|
* in may not match the PFN we have mapped if the
|
|
|
|
* mapped PFN is a writeable COW page. In the mkwrite
|
|
|
|
* case we are creating a writable PTE for a shared
|
2018-10-30 22:10:47 +00:00
|
|
|
* mapping and we expect the PFNs to match. If they
|
|
|
|
* don't match, we are likely racing with block
|
|
|
|
* allocation and mapping invalidation so just skip the
|
|
|
|
* update.
|
2017-09-06 23:18:35 +00:00
|
|
|
*/
|
2018-10-30 22:10:47 +00:00
|
|
|
if (pte_pfn(*pte) != pfn_t_to_pfn(pfn)) {
|
|
|
|
WARN_ON_ONCE(!is_zero_pfn(pte_pfn(*pte)));
|
2017-09-06 23:18:35 +00:00
|
|
|
goto out_unlock;
|
2018-10-30 22:10:47 +00:00
|
|
|
}
|
2019-03-29 03:43:19 +00:00
|
|
|
entry = pte_mkyoung(*pte);
|
|
|
|
entry = maybe_mkwrite(pte_mkdirty(entry), vma);
|
|
|
|
if (ptep_set_access_flags(vma, addr, pte, entry, 1))
|
|
|
|
update_mmu_cache(vma, addr, pte);
|
|
|
|
}
|
|
|
|
goto out_unlock;
|
2017-09-06 23:18:35 +00:00
|
|
|
}
|
2008-04-28 09:13:01 +00:00
|
|
|
|
|
|
|
/* Ok, finally just insert the thing.. */
|
2016-01-16 00:56:40 +00:00
|
|
|
if (pfn_t_devmap(pfn))
|
|
|
|
entry = pte_mkdevmap(pfn_t_pte(pfn, prot));
|
|
|
|
else
|
|
|
|
entry = pte_mkspecial(pfn_t_pte(pfn, prot));
|
2017-09-06 23:18:35 +00:00
|
|
|
|
|
|
|
if (mkwrite) {
|
|
|
|
entry = pte_mkyoung(entry);
|
|
|
|
entry = maybe_mkwrite(pte_mkdirty(entry), vma);
|
|
|
|
}
|
|
|
|
|
2008-04-28 09:13:01 +00:00
|
|
|
set_pte_at(mm, addr, pte, entry);
|
MM: Pass a PTE pointer to update_mmu_cache() rather than the PTE itself
On VIVT ARM, when we have multiple shared mappings of the same file
in the same MM, we need to ensure that we have coherency across all
copies. We do this via make_coherent() by making the pages
uncacheable.
This used to work fine, until we allowed highmem with highpte - we
now have a page table which is mapped as required, and is not available
for modification via update_mmu_cache().
Ralf Beache suggested getting rid of the PTE value passed to
update_mmu_cache():
On MIPS update_mmu_cache() calls __update_tlb() which walks pagetables
to construct a pointer to the pte again. Passing a pte_t * is much
more elegant. Maybe we might even replace the pte argument with the
pte_t?
Ben Herrenschmidt would also like the pte pointer for PowerPC:
Passing the ptep in there is exactly what I want. I want that
-instead- of the PTE value, because I have issue on some ppc cases,
for I$/D$ coherency, where set_pte_at() may decide to mask out the
_PAGE_EXEC.
So, pass in the mapped page table pointer into update_mmu_cache(), and
remove the PTE value, updating all implementations and call sites to
suit.
Includes a fix from Stephen Rothwell:
sparc: fix fallout from update_mmu_cache API change
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2009-12-18 16:40:18 +00:00
|
|
|
update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
|
2008-04-28 09:13:01 +00:00
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
pte_unmap_unlock(pte, ptl);
|
2018-10-26 22:04:40 +00:00
|
|
|
return VM_FAULT_NOPAGE;
|
2008-04-28 09:13:01 +00:00
|
|
|
}
|
|
|
|
|
2018-10-26 22:04:13 +00:00
|
|
|
/**
|
|
|
|
* vmf_insert_pfn_prot - insert single pfn into user vma with specified pgprot
|
|
|
|
* @vma: user vma to map to
|
|
|
|
* @addr: target user address of this page
|
|
|
|
* @pfn: source kernel pfn
|
|
|
|
* @pgprot: pgprot flags for the inserted page
|
|
|
|
*
|
2020-08-12 01:33:05 +00:00
|
|
|
* This is exactly like vmf_insert_pfn(), except that it allows drivers
|
2018-10-26 22:04:13 +00:00
|
|
|
* to override pgprot on a per-page basis.
|
|
|
|
*
|
|
|
|
* This only makes sense for IO mappings, and it makes no sense for
|
|
|
|
* COW mappings. In general, using multiple vmas is preferable;
|
2018-10-26 22:04:29 +00:00
|
|
|
* vmf_insert_pfn_prot should only be used if using multiple VMAs is
|
2018-10-26 22:04:13 +00:00
|
|
|
* impractical.
|
|
|
|
*
|
2023-03-12 23:40:13 +00:00
|
|
|
* pgprot typically only differs from @vma->vm_page_prot when drivers set
|
|
|
|
* caching- and encryption bits different than those of @vma->vm_page_prot,
|
|
|
|
* because the caching- or encryption mode may not be known at mmap() time.
|
|
|
|
*
|
|
|
|
* This is ok as long as @vma->vm_page_prot is not used by the core vm
|
|
|
|
* to set caching and encryption bits for those vmas (except for COW pages).
|
|
|
|
* This is ensured by core vm only modifying these page table entries using
|
|
|
|
* functions that don't touch caching- or encryption bits, using pte_modify()
|
|
|
|
* if needed. (See for example mprotect()).
|
|
|
|
*
|
|
|
|
* Also when new page-table entries are created, this is only done using the
|
|
|
|
* fault() callback, and never using the value of vma->vm_page_prot,
|
|
|
|
* except for page-table entries that point to anonymous pages as the result
|
|
|
|
* of COW.
|
2019-11-22 08:25:12 +00:00
|
|
|
*
|
2018-10-26 22:04:29 +00:00
|
|
|
* Context: Process context. May allocate using %GFP_KERNEL.
|
2018-10-26 22:04:13 +00:00
|
|
|
* Return: vm_fault_t value.
|
|
|
|
*/
|
|
|
|
vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
unsigned long pfn, pgprot_t pgprot)
|
|
|
|
{
|
2018-10-26 22:04:33 +00:00
|
|
|
/*
|
|
|
|
* Technically, architectures with pte_special can avoid all these
|
|
|
|
* restrictions (same for remap_pfn_range). However we would like
|
|
|
|
* consistency in testing and feature parity among all, so we should
|
|
|
|
* try to keep these invariants in place for everybody.
|
|
|
|
*/
|
|
|
|
BUG_ON(!(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)));
|
|
|
|
BUG_ON((vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)) ==
|
|
|
|
(VM_PFNMAP|VM_MIXEDMAP));
|
|
|
|
BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags));
|
|
|
|
BUG_ON((vma->vm_flags & VM_MIXEDMAP) && pfn_valid(pfn));
|
|
|
|
|
|
|
|
if (addr < vma->vm_start || addr >= vma->vm_end)
|
|
|
|
return VM_FAULT_SIGBUS;
|
|
|
|
|
|
|
|
if (!pfn_modify_allowed(pfn, pgprot))
|
|
|
|
return VM_FAULT_SIGBUS;
|
|
|
|
|
|
|
|
track_pfn_insert(vma, &pgprot, __pfn_to_pfn_t(pfn, PFN_DEV));
|
|
|
|
|
2018-10-26 22:04:40 +00:00
|
|
|
return insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot,
|
2018-10-26 22:04:33 +00:00
|
|
|
false);
|
2018-10-26 22:04:13 +00:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(vmf_insert_pfn_prot);
|
2007-02-12 08:51:36 +00:00
|
|
|
|
2018-10-26 22:04:29 +00:00
|
|
|
/**
|
|
|
|
* vmf_insert_pfn - insert single pfn into user vma
|
|
|
|
* @vma: user vma to map to
|
|
|
|
* @addr: target user address of this page
|
|
|
|
* @pfn: source kernel pfn
|
|
|
|
*
|
|
|
|
* Similar to vm_insert_page, this allows drivers to insert individual pages
|
|
|
|
* they've allocated into a user vma. Same comments apply.
|
|
|
|
*
|
|
|
|
* This function should only be called from a vm_ops->fault handler, and
|
|
|
|
* in that case the handler should return the result of this function.
|
|
|
|
*
|
|
|
|
* vma cannot be a COW mapping.
|
|
|
|
*
|
|
|
|
* As this is called only for pages that do not currently exist, we
|
|
|
|
* do not need to flush old virtual caches or the TLB.
|
|
|
|
*
|
|
|
|
* Context: Process context. May allocate using %GFP_KERNEL.
|
|
|
|
* Return: vm_fault_t value.
|
|
|
|
*/
|
|
|
|
vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
unsigned long pfn)
|
|
|
|
{
|
|
|
|
return vmf_insert_pfn_prot(vma, addr, pfn, vma->vm_page_prot);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(vmf_insert_pfn);
|
|
|
|
|
2017-10-23 14:20:00 +00:00
|
|
|
static bool vm_mixed_ok(struct vm_area_struct *vma, pfn_t pfn)
|
|
|
|
{
|
|
|
|
/* these checks mirror the abort conditions in vm_normal_page */
|
|
|
|
if (vma->vm_flags & VM_MIXEDMAP)
|
|
|
|
return true;
|
|
|
|
if (pfn_t_devmap(pfn))
|
|
|
|
return true;
|
|
|
|
if (pfn_t_special(pfn))
|
|
|
|
return true;
|
|
|
|
if (is_zero_pfn(pfn_t_to_pfn(pfn)))
|
|
|
|
return true;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2018-10-26 22:04:37 +00:00
|
|
|
static vm_fault_t __vm_insert_mixed(struct vm_area_struct *vma,
|
2023-03-12 23:40:13 +00:00
|
|
|
unsigned long addr, pfn_t pfn, bool mkwrite)
|
2008-04-28 09:13:01 +00:00
|
|
|
{
|
2023-03-12 23:40:13 +00:00
|
|
|
pgprot_t pgprot = vma->vm_page_prot;
|
2018-10-26 22:04:37 +00:00
|
|
|
int err;
|
2016-10-08 00:00:18 +00:00
|
|
|
|
2017-10-23 14:20:00 +00:00
|
|
|
BUG_ON(!vm_mixed_ok(vma, pfn));
|
2007-02-12 08:51:36 +00:00
|
|
|
|
2008-04-28 09:13:01 +00:00
|
|
|
if (addr < vma->vm_start || addr >= vma->vm_end)
|
2018-10-26 22:04:37 +00:00
|
|
|
return VM_FAULT_SIGBUS;
|
2016-10-26 17:43:43 +00:00
|
|
|
|
|
|
|
track_pfn_insert(vma, &pgprot, pfn);
|
2007-02-12 08:51:36 +00:00
|
|
|
|
2018-06-13 22:48:27 +00:00
|
|
|
if (!pfn_modify_allowed(pfn_t_to_pfn(pfn), pgprot))
|
2018-10-26 22:04:37 +00:00
|
|
|
return VM_FAULT_SIGBUS;
|
2018-06-13 22:48:27 +00:00
|
|
|
|
2008-04-28 09:13:01 +00:00
|
|
|
/*
|
|
|
|
* If we don't have pte special, then we have to use the pfn_valid()
|
|
|
|
* based VM_MIXEDMAP scheme (see vm_normal_page), and thus we *must*
|
|
|
|
* refcount the page if pfn_valid is true (hence insert_page rather
|
2009-09-22 00:03:34 +00:00
|
|
|
* than insert_pfn). If a zero_pfn were inserted into a VM_MIXEDMAP
|
|
|
|
* without pte special, it would there be refcounted as a normal page.
|
2008-04-28 09:13:01 +00:00
|
|
|
*/
|
2018-06-08 00:06:12 +00:00
|
|
|
if (!IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL) &&
|
|
|
|
!pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
|
2008-04-28 09:13:01 +00:00
|
|
|
struct page *page;
|
|
|
|
|
2016-01-26 17:48:05 +00:00
|
|
|
/*
|
|
|
|
* At this point we are committed to insert_page()
|
|
|
|
* regardless of whether the caller specified flags that
|
|
|
|
* result in pfn_t_has_page() == false.
|
|
|
|
*/
|
|
|
|
page = pfn_to_page(pfn_t_to_pfn(pfn));
|
2018-10-26 22:04:37 +00:00
|
|
|
err = insert_page(vma, addr, page, pgprot);
|
|
|
|
} else {
|
2018-10-26 22:04:40 +00:00
|
|
|
return insert_pfn(vma, addr, pfn, pgprot, mkwrite);
|
2008-04-28 09:13:01 +00:00
|
|
|
}
|
2017-09-06 23:18:35 +00:00
|
|
|
|
2018-10-26 22:04:10 +00:00
|
|
|
if (err == -ENOMEM)
|
|
|
|
return VM_FAULT_OOM;
|
|
|
|
if (err < 0 && err != -EBUSY)
|
|
|
|
return VM_FAULT_SIGBUS;
|
|
|
|
|
|
|
|
return VM_FAULT_NOPAGE;
|
2007-02-12 08:51:36 +00:00
|
|
|
}
|
2018-10-26 22:04:37 +00:00
|
|
|
|
|
|
|
vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
pfn_t pfn)
|
|
|
|
{
|
2023-03-12 23:40:13 +00:00
|
|
|
return __vm_insert_mixed(vma, addr, pfn, false);
|
2018-10-26 22:04:37 +00:00
|
|
|
}
|
2018-10-26 22:04:10 +00:00
|
|
|
EXPORT_SYMBOL(vmf_insert_mixed);
|
2007-02-12 08:51:36 +00:00
|
|
|
|
2018-06-08 00:04:29 +00:00
|
|
|
/*
|
|
|
|
* If the insertion of PTE failed because someone else already added a
|
|
|
|
* different entry in the mean time, we treat that as success as we assume
|
|
|
|
* the same entry was actually inserted.
|
|
|
|
*/
|
|
|
|
vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
|
|
|
|
unsigned long addr, pfn_t pfn)
|
2017-09-06 23:18:35 +00:00
|
|
|
{
|
2023-03-12 23:40:13 +00:00
|
|
|
return __vm_insert_mixed(vma, addr, pfn, true);
|
2017-09-06 23:18:35 +00:00
|
|
|
}
|
2018-06-08 00:04:29 +00:00
|
|
|
EXPORT_SYMBOL(vmf_insert_mixed_mkwrite);
|
2017-09-06 23:18:35 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* maps a range of physical memory into the requested pages. the old
|
|
|
|
* mappings are removed. any references to nonexistent pages results
|
|
|
|
* in null mappings (currently treated as "copy-on-access")
|
|
|
|
*/
|
|
|
|
static int remap_pte_range(struct mm_struct *mm, pmd_t *pmd,
|
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
unsigned long pfn, pgprot_t prot)
|
|
|
|
{
|
2021-02-24 20:04:33 +00:00
|
|
|
pte_t *pte, *mapped_pte;
|
2005-10-30 01:16:23 +00:00
|
|
|
spinlock_t *ptl;
|
2018-06-13 22:48:27 +00:00
|
|
|
int err = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2021-02-24 20:04:33 +00:00
|
|
|
mapped_pte = pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (!pte)
|
|
|
|
return -ENOMEM;
|
2006-10-01 06:29:33 +00:00
|
|
|
arch_enter_lazy_mmu_mode();
|
2005-04-16 22:20:36 +00:00
|
|
|
do {
|
|
|
|
BUG_ON(!pte_none(*pte));
|
2018-06-13 22:48:27 +00:00
|
|
|
if (!pfn_modify_allowed(pfn, prot)) {
|
|
|
|
err = -EACCES;
|
|
|
|
break;
|
|
|
|
}
|
mm: introduce pte_special pte bit
s390 for one, cannot implement VM_MIXEDMAP with pfn_valid, due to their memory
model (which is more dynamic than most). Instead, they had proposed to
implement it with an additional path through vm_normal_page(), using a bit in
the pte to determine whether or not the page should be refcounted:
vm_normal_page()
{
...
if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
if (vma->vm_flags & VM_MIXEDMAP) {
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
#else
if (!pfn_valid(pfn))
return NULL;
#endif
goto out;
}
...
}
This is fine, however if we are allowed to use a bit in the pte to determine
refcountedness, we can use that to _completely_ replace all the vma based
schemes. So instead of adding more cases to the already complex vma-based
scheme, we can have a clearly seperate and simple pte-based scheme (and get
slightly better code generation in the process):
vm_normal_page()
{
#ifdef s390
if (!mixedmap_refcount_pte(pte))
return NULL;
return pte_page(pte);
#else
...
#endif
}
And finally, we may rather make this concept usable by any architecture rather
than making it s390 only, so implement a new type of pte state for this.
Unfortunately the old vma based code must stay, because some architectures may
not be able to spare pte bits. This makes vm_normal_page a little bit more
ugly than we would like, but the 2 cases are clearly seperate.
So introduce a pte_special pte state, and use it in mm/memory.c. It is
currently a noop for all architectures, so this doesn't actually result in any
compiled code changes to mm/memory.o.
BTW:
I haven't put vm_normal_page() into arch code as-per an earlier suggestion.
The reason is that, regardless of where vm_normal_page is actually
implemented, the *abstraction* is still exactly the same. Also, while it
depends on whether the architecture has pte_special or not, that is the
only two possible cases, and it really isn't an arch specific function --
the role of the arch code should be to provide primitive functions and
accessors with which to build the core code; pte_special does that. We do
not want architectures to know or care about vm_normal_page itself, and
we definitely don't want them being able to invent something new there
out of sight of mm/ code. If we made vm_normal_page an arch function, then
we have to make vm_insert_mixed (next patch) an arch function too. So I
don't think moving it to arch code fundamentally improves any abstractions,
while it does practically make the code more difficult to follow, for both
mm and arch developers, and easier to misuse.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Carsten Otte <cotte@de.ibm.com>
Cc: Jared Hulbert <jaredeh@gmail.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:13:00 +00:00
|
|
|
set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
|
2005-04-16 22:20:36 +00:00
|
|
|
pfn++;
|
|
|
|
} while (pte++, addr += PAGE_SIZE, addr != end);
|
2006-10-01 06:29:33 +00:00
|
|
|
arch_leave_lazy_mmu_mode();
|
2021-02-24 20:04:33 +00:00
|
|
|
pte_unmap_unlock(mapped_pte, ptl);
|
2018-06-13 22:48:27 +00:00
|
|
|
return err;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline int remap_pmd_range(struct mm_struct *mm, pud_t *pud,
|
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
unsigned long pfn, pgprot_t prot)
|
|
|
|
{
|
|
|
|
pmd_t *pmd;
|
|
|
|
unsigned long next;
|
2018-06-13 22:48:27 +00:00
|
|
|
int err;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
pfn -= addr >> PAGE_SHIFT;
|
|
|
|
pmd = pmd_alloc(mm, pud, addr);
|
|
|
|
if (!pmd)
|
|
|
|
return -ENOMEM;
|
2011-01-13 23:46:54 +00:00
|
|
|
VM_BUG_ON(pmd_trans_huge(*pmd));
|
2005-04-16 22:20:36 +00:00
|
|
|
do {
|
|
|
|
next = pmd_addr_end(addr, end);
|
2018-06-13 22:48:27 +00:00
|
|
|
err = remap_pte_range(mm, pmd, addr, next,
|
|
|
|
pfn + (addr >> PAGE_SHIFT), prot);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2005-04-16 22:20:36 +00:00
|
|
|
} while (pmd++, addr = next, addr != end);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
static inline int remap_pud_range(struct mm_struct *mm, p4d_t *p4d,
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
unsigned long pfn, pgprot_t prot)
|
|
|
|
{
|
|
|
|
pud_t *pud;
|
|
|
|
unsigned long next;
|
2018-06-13 22:48:27 +00:00
|
|
|
int err;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
pfn -= addr >> PAGE_SHIFT;
|
2017-03-09 14:24:07 +00:00
|
|
|
pud = pud_alloc(mm, p4d, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (!pud)
|
|
|
|
return -ENOMEM;
|
|
|
|
do {
|
|
|
|
next = pud_addr_end(addr, end);
|
2018-06-13 22:48:27 +00:00
|
|
|
err = remap_pmd_range(mm, pud, addr, next,
|
|
|
|
pfn + (addr >> PAGE_SHIFT), prot);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2005-04-16 22:20:36 +00:00
|
|
|
} while (pud++, addr = next, addr != end);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
static inline int remap_p4d_range(struct mm_struct *mm, pgd_t *pgd,
|
|
|
|
unsigned long addr, unsigned long end,
|
|
|
|
unsigned long pfn, pgprot_t prot)
|
|
|
|
{
|
|
|
|
p4d_t *p4d;
|
|
|
|
unsigned long next;
|
2018-06-13 22:48:27 +00:00
|
|
|
int err;
|
2017-03-09 14:24:07 +00:00
|
|
|
|
|
|
|
pfn -= addr >> PAGE_SHIFT;
|
|
|
|
p4d = p4d_alloc(mm, pgd, addr);
|
|
|
|
if (!p4d)
|
|
|
|
return -ENOMEM;
|
|
|
|
do {
|
|
|
|
next = p4d_addr_end(addr, end);
|
2018-06-13 22:48:27 +00:00
|
|
|
err = remap_pud_range(mm, p4d, addr, next,
|
|
|
|
pfn + (addr >> PAGE_SHIFT), prot);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2017-03-09 14:24:07 +00:00
|
|
|
} while (p4d++, addr = next, addr != end);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2021-04-30 05:57:29 +00:00
|
|
|
/*
|
|
|
|
* Variant of remap_pfn_range that does not call track_pfn_remap. The caller
|
|
|
|
* must have pre-validated the caching bits of the pgprot_t.
|
2006-09-26 06:31:22 +00:00
|
|
|
*/
|
2021-04-30 05:57:29 +00:00
|
|
|
int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
unsigned long pfn, unsigned long size, pgprot_t prot)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pgd_t *pgd;
|
|
|
|
unsigned long next;
|
2005-06-25 21:54:33 +00:00
|
|
|
unsigned long end = addr + PAGE_ALIGN(size);
|
2005-04-16 22:20:36 +00:00
|
|
|
struct mm_struct *mm = vma->vm_mm;
|
|
|
|
int err;
|
|
|
|
|
2020-08-07 06:22:24 +00:00
|
|
|
if (WARN_ON_ONCE(!PAGE_ALIGNED(addr)))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Physically remapped pages are special. Tell the
|
|
|
|
* rest of the world about it:
|
|
|
|
* VM_IO tells people not to look at these pages
|
|
|
|
* (accesses can have side effects).
|
2005-11-28 22:34:23 +00:00
|
|
|
* VM_PFNMAP tells the core MM that the base pages are just
|
|
|
|
* raw PFN mappings, and do not have a "struct page" associated
|
|
|
|
* with them.
|
mm: kill vma flag VM_RESERVED and mm->reserved_vm counter
A long time ago, in v2.4, VM_RESERVED kept swapout process off VMA,
currently it lost original meaning but still has some effects:
| effect | alternative flags
-+------------------------+---------------------------------------------
1| account as reserved_vm | VM_IO
2| skip in core dump | VM_IO, VM_DONTDUMP
3| do not merge or expand | VM_IO, VM_DONTEXPAND, VM_HUGETLB, VM_PFNMAP
4| do not mlock | VM_IO, VM_DONTEXPAND, VM_HUGETLB, VM_PFNMAP
This patch removes reserved_vm counter from mm_struct. Seems like nobody
cares about it, it does not exported into userspace directly, it only
reduces total_vm showed in proc.
Thus VM_RESERVED can be replaced with VM_IO or pair VM_DONTEXPAND | VM_DONTDUMP.
remap_pfn_range() and io_remap_pfn_range() set VM_IO|VM_DONTEXPAND|VM_DONTDUMP.
remap_vmalloc_range() set VM_DONTEXPAND | VM_DONTDUMP.
[akpm@linux-foundation.org: drivers/vfio/pci/vfio_pci.c fixup]
Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Carsten Otte <cotte@de.ibm.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Eric Paris <eparis@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Morris <james.l.morris@oracle.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Kentaro Takeda <takedakn@nttdata.co.jp>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Venkatesh Pallipadi <venki@google.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-08 23:29:02 +00:00
|
|
|
* VM_DONTEXPAND
|
|
|
|
* Disable vma merging and expanding with mremap().
|
|
|
|
* VM_DONTDUMP
|
|
|
|
* Omit vma from core dump, even when VM_IO turned off.
|
2005-12-12 03:46:02 +00:00
|
|
|
*
|
|
|
|
* There's a horrible special case to handle copy-on-write
|
|
|
|
* behaviour that some programs depend on. We mark the "original"
|
|
|
|
* un-COW'ed pages by matching them up with "vma->vm_pgoff".
|
2012-10-08 23:28:34 +00:00
|
|
|
* See vm_normal_page() for details.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2012-10-08 23:28:34 +00:00
|
|
|
if (is_cow_mapping(vma->vm_flags)) {
|
|
|
|
if (addr != vma->vm_start || end != vma->vm_end)
|
|
|
|
return -EINVAL;
|
2005-12-12 03:46:02 +00:00
|
|
|
vma->vm_pgoff = pfn;
|
2012-10-08 23:28:34 +00:00
|
|
|
}
|
|
|
|
|
2023-01-26 19:37:49 +00:00
|
|
|
vm_flags_set(vma, VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
BUG_ON(addr >= end);
|
|
|
|
pfn -= addr >> PAGE_SHIFT;
|
|
|
|
pgd = pgd_offset(mm, addr);
|
|
|
|
flush_cache_range(vma, addr, end);
|
|
|
|
do {
|
|
|
|
next = pgd_addr_end(addr, end);
|
2017-03-09 14:24:07 +00:00
|
|
|
err = remap_p4d_range(mm, pgd, addr, next,
|
2005-04-16 22:20:36 +00:00
|
|
|
pfn + (addr >> PAGE_SHIFT), prot);
|
|
|
|
if (err)
|
2021-04-30 05:57:29 +00:00
|
|
|
return err;
|
2005-04-16 22:20:36 +00:00
|
|
|
} while (pgd++, addr = next, addr != end);
|
2008-12-18 19:41:29 +00:00
|
|
|
|
2021-04-30 05:57:29 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* remap_pfn_range - remap kernel memory to userspace
|
|
|
|
* @vma: user vma to map to
|
|
|
|
* @addr: target page aligned user address to start at
|
|
|
|
* @pfn: page frame number of kernel physical memory address
|
|
|
|
* @size: size of mapping area
|
|
|
|
* @prot: page protection flags for this mapping
|
|
|
|
*
|
|
|
|
* Note: this is only safe if the mm semaphore is held when called.
|
|
|
|
*
|
|
|
|
* Return: %0 on success, negative error code otherwise.
|
|
|
|
*/
|
|
|
|
int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
unsigned long pfn, unsigned long size, pgprot_t prot)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = track_pfn_remap(vma, &prot, pfn, addr, PAGE_ALIGN(size));
|
2008-12-18 19:41:29 +00:00
|
|
|
if (err)
|
2021-04-30 05:57:29 +00:00
|
|
|
return -EINVAL;
|
2008-12-18 19:41:29 +00:00
|
|
|
|
2021-04-30 05:57:29 +00:00
|
|
|
err = remap_pfn_range_notrack(vma, addr, pfn, size, prot);
|
|
|
|
if (err)
|
2023-01-26 19:37:51 +00:00
|
|
|
untrack_pfn(vma, pfn, PAGE_ALIGN(size), true);
|
2005-04-16 22:20:36 +00:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(remap_pfn_range);
|
|
|
|
|
2013-04-16 20:45:37 +00:00
|
|
|
/**
|
|
|
|
* vm_iomap_memory - remap memory to userspace
|
|
|
|
* @vma: user vma to map to
|
2020-04-02 04:09:07 +00:00
|
|
|
* @start: start of the physical memory to be mapped
|
2013-04-16 20:45:37 +00:00
|
|
|
* @len: size of area
|
|
|
|
*
|
|
|
|
* This is a simplified io_remap_pfn_range() for common driver use. The
|
|
|
|
* driver just needs to give us the physical memory range to be mapped,
|
|
|
|
* we'll figure out the rest from the vma information.
|
|
|
|
*
|
|
|
|
* NOTE! Some drivers might want to tweak vma->vm_page_prot first to get
|
|
|
|
* whatever write-combining details or similar.
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: %0 on success, negative error code otherwise.
|
2013-04-16 20:45:37 +00:00
|
|
|
*/
|
|
|
|
int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len)
|
|
|
|
{
|
|
|
|
unsigned long vm_len, pfn, pages;
|
|
|
|
|
|
|
|
/* Check that the physical memory area passed in looks valid */
|
|
|
|
if (start + len < start)
|
|
|
|
return -EINVAL;
|
|
|
|
/*
|
|
|
|
* You *really* shouldn't map things that aren't page-aligned,
|
|
|
|
* but we've historically allowed it because IO memory might
|
|
|
|
* just have smaller alignment.
|
|
|
|
*/
|
|
|
|
len += start & ~PAGE_MASK;
|
|
|
|
pfn = start >> PAGE_SHIFT;
|
|
|
|
pages = (len + ~PAGE_MASK) >> PAGE_SHIFT;
|
|
|
|
if (pfn + pages < pfn)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
/* We start the mapping 'vm_pgoff' pages into the area */
|
|
|
|
if (vma->vm_pgoff > pages)
|
|
|
|
return -EINVAL;
|
|
|
|
pfn += vma->vm_pgoff;
|
|
|
|
pages -= vma->vm_pgoff;
|
|
|
|
|
|
|
|
/* Can we fit all of the mapping? */
|
|
|
|
vm_len = vma->vm_end - vma->vm_start;
|
|
|
|
if (vm_len >> PAGE_SHIFT > pages)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
/* Ok, let it rip */
|
|
|
|
return io_remap_pfn_range(vma, vma->vm_start, pfn, vm_len, vma->vm_page_prot);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(vm_iomap_memory);
|
|
|
|
|
2007-05-06 21:48:54 +00:00
|
|
|
static int apply_to_pte_range(struct mm_struct *mm, pmd_t *pmd,
|
|
|
|
unsigned long addr, unsigned long end,
|
2020-09-04 23:35:43 +00:00
|
|
|
pte_fn_t fn, void *data, bool create,
|
|
|
|
pgtbl_mod_mask *mask)
|
2007-05-06 21:48:54 +00:00
|
|
|
{
|
2021-02-24 20:04:42 +00:00
|
|
|
pte_t *pte, *mapped_pte;
|
2019-12-18 04:51:41 +00:00
|
|
|
int err = 0;
|
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
|
|
|
spinlock_t *ptl;
|
2007-05-06 21:48:54 +00:00
|
|
|
|
2019-12-18 04:51:41 +00:00
|
|
|
if (create) {
|
2021-02-24 20:04:42 +00:00
|
|
|
mapped_pte = pte = (mm == &init_mm) ?
|
2020-09-04 23:35:43 +00:00
|
|
|
pte_alloc_kernel_track(pmd, addr, mask) :
|
2019-12-18 04:51:41 +00:00
|
|
|
pte_alloc_map_lock(mm, pmd, addr, &ptl);
|
|
|
|
if (!pte)
|
|
|
|
return -ENOMEM;
|
|
|
|
} else {
|
2021-02-24 20:04:42 +00:00
|
|
|
mapped_pte = pte = (mm == &init_mm) ?
|
2019-12-18 04:51:41 +00:00
|
|
|
pte_offset_kernel(pmd, addr) :
|
|
|
|
pte_offset_map_lock(mm, pmd, addr, &ptl);
|
|
|
|
}
|
2007-05-06 21:48:54 +00:00
|
|
|
|
|
|
|
BUG_ON(pmd_huge(*pmd));
|
|
|
|
|
2009-01-06 22:39:21 +00:00
|
|
|
arch_enter_lazy_mmu_mode();
|
|
|
|
|
2020-10-17 23:15:14 +00:00
|
|
|
if (fn) {
|
|
|
|
do {
|
|
|
|
if (create || !pte_none(*pte)) {
|
|
|
|
err = fn(pte++, addr, data);
|
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
} while (addr += PAGE_SIZE, addr != end);
|
|
|
|
}
|
2020-09-04 23:35:43 +00:00
|
|
|
*mask |= PGTBL_PTE_MODIFIED;
|
2007-05-06 21:48:54 +00:00
|
|
|
|
2009-01-06 22:39:21 +00:00
|
|
|
arch_leave_lazy_mmu_mode();
|
|
|
|
|
2007-05-06 21:48:54 +00:00
|
|
|
if (mm != &init_mm)
|
2021-02-24 20:04:42 +00:00
|
|
|
pte_unmap_unlock(mapped_pte, ptl);
|
2007-05-06 21:48:54 +00:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int apply_to_pmd_range(struct mm_struct *mm, pud_t *pud,
|
|
|
|
unsigned long addr, unsigned long end,
|
2020-09-04 23:35:43 +00:00
|
|
|
pte_fn_t fn, void *data, bool create,
|
|
|
|
pgtbl_mod_mask *mask)
|
2007-05-06 21:48:54 +00:00
|
|
|
{
|
|
|
|
pmd_t *pmd;
|
|
|
|
unsigned long next;
|
2019-12-18 04:51:41 +00:00
|
|
|
int err = 0;
|
2007-05-06 21:48:54 +00:00
|
|
|
|
2008-07-24 04:27:50 +00:00
|
|
|
BUG_ON(pud_huge(*pud));
|
|
|
|
|
2019-12-18 04:51:41 +00:00
|
|
|
if (create) {
|
2020-09-04 23:35:43 +00:00
|
|
|
pmd = pmd_alloc_track(mm, pud, addr, mask);
|
2019-12-18 04:51:41 +00:00
|
|
|
if (!pmd)
|
|
|
|
return -ENOMEM;
|
|
|
|
} else {
|
|
|
|
pmd = pmd_offset(pud, addr);
|
|
|
|
}
|
2007-05-06 21:48:54 +00:00
|
|
|
do {
|
|
|
|
next = pmd_addr_end(addr, end);
|
2021-04-30 05:58:16 +00:00
|
|
|
if (pmd_none(*pmd) && !create)
|
|
|
|
continue;
|
|
|
|
if (WARN_ON_ONCE(pmd_leaf(*pmd)))
|
|
|
|
return -EINVAL;
|
|
|
|
if (!pmd_none(*pmd) && WARN_ON_ONCE(pmd_bad(*pmd))) {
|
|
|
|
if (!create)
|
|
|
|
continue;
|
|
|
|
pmd_clear_bad(pmd);
|
2019-12-18 04:51:41 +00:00
|
|
|
}
|
2021-04-30 05:58:16 +00:00
|
|
|
err = apply_to_pte_range(mm, pmd, addr, next,
|
|
|
|
fn, data, create, mask);
|
|
|
|
if (err)
|
|
|
|
break;
|
2007-05-06 21:48:54 +00:00
|
|
|
} while (pmd++, addr = next, addr != end);
|
2021-04-30 05:58:16 +00:00
|
|
|
|
2007-05-06 21:48:54 +00:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
static int apply_to_pud_range(struct mm_struct *mm, p4d_t *p4d,
|
2007-05-06 21:48:54 +00:00
|
|
|
unsigned long addr, unsigned long end,
|
2020-09-04 23:35:43 +00:00
|
|
|
pte_fn_t fn, void *data, bool create,
|
|
|
|
pgtbl_mod_mask *mask)
|
2007-05-06 21:48:54 +00:00
|
|
|
{
|
|
|
|
pud_t *pud;
|
|
|
|
unsigned long next;
|
2019-12-18 04:51:41 +00:00
|
|
|
int err = 0;
|
2007-05-06 21:48:54 +00:00
|
|
|
|
2019-12-18 04:51:41 +00:00
|
|
|
if (create) {
|
2020-09-04 23:35:43 +00:00
|
|
|
pud = pud_alloc_track(mm, p4d, addr, mask);
|
2019-12-18 04:51:41 +00:00
|
|
|
if (!pud)
|
|
|
|
return -ENOMEM;
|
|
|
|
} else {
|
|
|
|
pud = pud_offset(p4d, addr);
|
|
|
|
}
|
2007-05-06 21:48:54 +00:00
|
|
|
do {
|
|
|
|
next = pud_addr_end(addr, end);
|
2021-04-30 05:58:16 +00:00
|
|
|
if (pud_none(*pud) && !create)
|
|
|
|
continue;
|
|
|
|
if (WARN_ON_ONCE(pud_leaf(*pud)))
|
|
|
|
return -EINVAL;
|
|
|
|
if (!pud_none(*pud) && WARN_ON_ONCE(pud_bad(*pud))) {
|
|
|
|
if (!create)
|
|
|
|
continue;
|
|
|
|
pud_clear_bad(pud);
|
2019-12-18 04:51:41 +00:00
|
|
|
}
|
2021-04-30 05:58:16 +00:00
|
|
|
err = apply_to_pmd_range(mm, pud, addr, next,
|
|
|
|
fn, data, create, mask);
|
|
|
|
if (err)
|
|
|
|
break;
|
2007-05-06 21:48:54 +00:00
|
|
|
} while (pud++, addr = next, addr != end);
|
2021-04-30 05:58:16 +00:00
|
|
|
|
2007-05-06 21:48:54 +00:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
static int apply_to_p4d_range(struct mm_struct *mm, pgd_t *pgd,
|
|
|
|
unsigned long addr, unsigned long end,
|
2020-09-04 23:35:43 +00:00
|
|
|
pte_fn_t fn, void *data, bool create,
|
|
|
|
pgtbl_mod_mask *mask)
|
2017-03-09 14:24:07 +00:00
|
|
|
{
|
|
|
|
p4d_t *p4d;
|
|
|
|
unsigned long next;
|
2019-12-18 04:51:41 +00:00
|
|
|
int err = 0;
|
2017-03-09 14:24:07 +00:00
|
|
|
|
2019-12-18 04:51:41 +00:00
|
|
|
if (create) {
|
2020-09-04 23:35:43 +00:00
|
|
|
p4d = p4d_alloc_track(mm, pgd, addr, mask);
|
2019-12-18 04:51:41 +00:00
|
|
|
if (!p4d)
|
|
|
|
return -ENOMEM;
|
|
|
|
} else {
|
|
|
|
p4d = p4d_offset(pgd, addr);
|
|
|
|
}
|
2017-03-09 14:24:07 +00:00
|
|
|
do {
|
|
|
|
next = p4d_addr_end(addr, end);
|
2021-04-30 05:58:16 +00:00
|
|
|
if (p4d_none(*p4d) && !create)
|
|
|
|
continue;
|
|
|
|
if (WARN_ON_ONCE(p4d_leaf(*p4d)))
|
|
|
|
return -EINVAL;
|
|
|
|
if (!p4d_none(*p4d) && WARN_ON_ONCE(p4d_bad(*p4d))) {
|
|
|
|
if (!create)
|
|
|
|
continue;
|
|
|
|
p4d_clear_bad(p4d);
|
2019-12-18 04:51:41 +00:00
|
|
|
}
|
2021-04-30 05:58:16 +00:00
|
|
|
err = apply_to_pud_range(mm, p4d, addr, next,
|
|
|
|
fn, data, create, mask);
|
|
|
|
if (err)
|
|
|
|
break;
|
2017-03-09 14:24:07 +00:00
|
|
|
} while (p4d++, addr = next, addr != end);
|
2021-04-30 05:58:16 +00:00
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2019-12-18 04:51:41 +00:00
|
|
|
static int __apply_to_page_range(struct mm_struct *mm, unsigned long addr,
|
|
|
|
unsigned long size, pte_fn_t fn,
|
|
|
|
void *data, bool create)
|
2007-05-06 21:48:54 +00:00
|
|
|
{
|
|
|
|
pgd_t *pgd;
|
2020-09-04 23:35:43 +00:00
|
|
|
unsigned long start = addr, next;
|
2010-08-10 00:19:52 +00:00
|
|
|
unsigned long end = addr + size;
|
2020-09-04 23:35:43 +00:00
|
|
|
pgtbl_mod_mask mask = 0;
|
2019-12-18 04:51:41 +00:00
|
|
|
int err = 0;
|
2007-05-06 21:48:54 +00:00
|
|
|
|
2016-03-15 21:56:45 +00:00
|
|
|
if (WARN_ON(addr >= end))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2007-05-06 21:48:54 +00:00
|
|
|
pgd = pgd_offset(mm, addr);
|
|
|
|
do {
|
|
|
|
next = pgd_addr_end(addr, end);
|
2021-04-30 05:58:16 +00:00
|
|
|
if (pgd_none(*pgd) && !create)
|
2019-12-18 04:51:41 +00:00
|
|
|
continue;
|
2021-04-30 05:58:16 +00:00
|
|
|
if (WARN_ON_ONCE(pgd_leaf(*pgd)))
|
|
|
|
return -EINVAL;
|
|
|
|
if (!pgd_none(*pgd) && WARN_ON_ONCE(pgd_bad(*pgd))) {
|
|
|
|
if (!create)
|
|
|
|
continue;
|
|
|
|
pgd_clear_bad(pgd);
|
|
|
|
}
|
|
|
|
err = apply_to_p4d_range(mm, pgd, addr, next,
|
|
|
|
fn, data, create, &mask);
|
2007-05-06 21:48:54 +00:00
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
} while (pgd++, addr = next, addr != end);
|
2010-08-10 00:19:52 +00:00
|
|
|
|
2020-09-04 23:35:43 +00:00
|
|
|
if (mask & ARCH_PAGE_TABLE_SYNC_MASK)
|
|
|
|
arch_sync_kernel_mappings(start, start + size);
|
|
|
|
|
2007-05-06 21:48:54 +00:00
|
|
|
return err;
|
|
|
|
}
|
2019-12-18 04:51:41 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Scan a region of virtual memory, filling in page tables as necessary
|
|
|
|
* and calling a provided function on each leaf page table.
|
|
|
|
*/
|
|
|
|
int apply_to_page_range(struct mm_struct *mm, unsigned long addr,
|
|
|
|
unsigned long size, pte_fn_t fn, void *data)
|
|
|
|
{
|
|
|
|
return __apply_to_page_range(mm, addr, size, fn, data, true);
|
|
|
|
}
|
2007-05-06 21:48:54 +00:00
|
|
|
EXPORT_SYMBOL_GPL(apply_to_page_range);
|
|
|
|
|
2019-12-18 04:51:41 +00:00
|
|
|
/*
|
|
|
|
* Scan a region of virtual memory, calling a provided function on
|
|
|
|
* each leaf page table where it exists.
|
|
|
|
*
|
|
|
|
* Unlike apply_to_page_range, this does _not_ fill in page tables
|
|
|
|
* where they are absent.
|
|
|
|
*/
|
|
|
|
int apply_to_existing_page_range(struct mm_struct *mm, unsigned long addr,
|
|
|
|
unsigned long size, pte_fn_t fn, void *data)
|
|
|
|
{
|
|
|
|
return __apply_to_page_range(mm, addr, size, fn, data, false);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(apply_to_existing_page_range);
|
|
|
|
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
/*
|
2015-02-10 22:09:51 +00:00
|
|
|
* handle_pte_fault chooses page fault handler according to an entry which was
|
|
|
|
* read non-atomically. Before making any commitment, on those architectures
|
|
|
|
* or configurations (e.g. i386 with PAE) which might give a mix of unmatched
|
|
|
|
* parts, do_swap_page must check under lock before unmapping the pte and
|
|
|
|
* proceeding (but do_wp_page is only called after already making such a check;
|
2011-02-10 04:56:28 +00:00
|
|
|
* and do_anonymous_page can safely check later on).
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
*/
|
2021-11-05 20:38:28 +00:00
|
|
|
static inline int pte_unmap_same(struct vm_fault *vmf)
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
{
|
|
|
|
int same = 1;
|
2019-10-15 19:18:12 +00:00
|
|
|
#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPTION)
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
if (sizeof(pte_t) > sizeof(unsigned long)) {
|
2021-11-05 20:38:28 +00:00
|
|
|
spinlock_t *ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
|
[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
|
|
|
spin_lock(ptl);
|
2021-11-05 20:38:28 +00:00
|
|
|
same = pte_same(*vmf->pte, vmf->orig_pte);
|
[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
|
|
|
spin_unlock(ptl);
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
}
|
|
|
|
#endif
|
2021-11-05 20:38:28 +00:00
|
|
|
pte_unmap(vmf->pte);
|
|
|
|
vmf->pte = NULL;
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
return same;
|
|
|
|
}
|
|
|
|
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
/*
|
|
|
|
* Return:
|
|
|
|
* 0: copied succeeded
|
|
|
|
* -EHWPOISON: copy failed due to hwpoison in source page
|
|
|
|
* -EAGAIN: copied failed (some other reason)
|
|
|
|
*/
|
|
|
|
static inline int __wp_page_copy_user(struct page *dst, struct page *src,
|
|
|
|
struct vm_fault *vmf)
|
2005-11-28 22:34:23 +00:00
|
|
|
{
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
int ret;
|
2019-10-11 14:09:39 +00:00
|
|
|
void *kaddr;
|
|
|
|
void __user *uaddr;
|
2020-03-06 06:28:32 +00:00
|
|
|
bool locked = false;
|
2019-10-11 14:09:39 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
|
|
|
struct mm_struct *mm = vma->vm_mm;
|
|
|
|
unsigned long addr = vmf->address;
|
|
|
|
|
|
|
|
if (likely(src)) {
|
2022-10-21 20:01:20 +00:00
|
|
|
if (copy_mc_user_highpage(dst, src, addr, vma)) {
|
|
|
|
memory_failure_queue(page_to_pfn(src), 0);
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
return -EHWPOISON;
|
2022-10-21 20:01:20 +00:00
|
|
|
}
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
return 0;
|
2019-10-11 14:09:39 +00:00
|
|
|
}
|
|
|
|
|
2005-11-28 22:34:23 +00:00
|
|
|
/*
|
|
|
|
* If the source page was a PFN mapping, we don't have
|
|
|
|
* a "struct page" for it. We do a best-effort copy by
|
|
|
|
* just copying from the original user address. If that
|
|
|
|
* fails, we just zero-fill it. Live with it.
|
|
|
|
*/
|
2019-10-11 14:09:39 +00:00
|
|
|
kaddr = kmap_atomic(dst);
|
|
|
|
uaddr = (void __user *)(addr & PAGE_MASK);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* On architectures with software "accessed" bits, we would
|
|
|
|
* take a double page fault, so mark it accessed here.
|
|
|
|
*/
|
mm: x86, arm64: add arch_has_hw_pte_young()
Patch series "Multi-Gen LRU Framework", v14.
What's new
==========
1. OpenWrt, in addition to Android, Arch Linux Zen, Armbian, ChromeOS,
Liquorix, post-factum and XanMod, is now shipping MGLRU on 5.15.
2. Fixed long-tailed direct reclaim latency seen on high-memory (TBs)
machines. The old direct reclaim backoff, which tries to enforce a
minimum fairness among all eligible memcgs, over-swapped by about
(total_mem>>DEF_PRIORITY)-nr_to_reclaim. The new backoff, which
pulls the plug on swapping once the target is met, trades some
fairness for curtailed latency:
https://lore.kernel.org/r/20220918080010.2920238-10-yuzhao@google.com/
3. Fixed minior build warnings and conflicts. More comments and nits.
TLDR
====
The current page reclaim is too expensive in terms of CPU usage and it
often makes poor choices about what to evict. This patchset offers an
alternative solution that is performant, versatile and
straightforward.
Patchset overview
=================
The design and implementation overview is in patch 14:
https://lore.kernel.org/r/20220918080010.2920238-15-yuzhao@google.com/
01. mm: x86, arm64: add arch_has_hw_pte_young()
02. mm: x86: add CONFIG_ARCH_HAS_NONLEAF_PMD_YOUNG
Take advantage of hardware features when trying to clear the accessed
bit in many PTEs.
03. mm/vmscan.c: refactor shrink_node()
04. Revert "include/linux/mm_inline.h: fold __update_lru_size() into
its sole caller"
Minor refactors to improve readability for the following patches.
05. mm: multi-gen LRU: groundwork
Adds the basic data structure and the functions that insert pages to
and remove pages from the multi-gen LRU (MGLRU) lists.
06. mm: multi-gen LRU: minimal implementation
A minimal implementation without optimizations.
07. mm: multi-gen LRU: exploit locality in rmap
Exploits spatial locality to improve efficiency when using the rmap.
08. mm: multi-gen LRU: support page table walks
Further exploits spatial locality by optionally scanning page tables.
09. mm: multi-gen LRU: optimize multiple memcgs
Optimizes the overall performance for multiple memcgs running mixed
types of workloads.
10. mm: multi-gen LRU: kill switch
Adds a kill switch to enable or disable MGLRU at runtime.
11. mm: multi-gen LRU: thrashing prevention
12. mm: multi-gen LRU: debugfs interface
Provide userspace with features like thrashing prevention, working set
estimation and proactive reclaim.
13. mm: multi-gen LRU: admin guide
14. mm: multi-gen LRU: design doc
Add an admin guide and a design doc.
Benchmark results
=================
Independent lab results
-----------------------
Based on the popularity of searches [01] and the memory usage in
Google's public cloud, the most popular open-source memory-hungry
applications, in alphabetical order, are:
Apache Cassandra Memcached
Apache Hadoop MongoDB
Apache Spark PostgreSQL
MariaDB (MySQL) Redis
An independent lab evaluated MGLRU with the most widely used benchmark
suites for the above applications. They posted 960 data points along
with kernel metrics and perf profiles collected over more than 500
hours of total benchmark time. Their final reports show that, with 95%
confidence intervals (CIs), the above applications all performed
significantly better for at least part of their benchmark matrices.
On 5.14:
1. Apache Spark [02] took 95% CIs [9.28, 11.19]% and [12.20, 14.93]%
less wall time to sort three billion random integers, respectively,
under the medium- and the high-concurrency conditions, when
overcommitting memory. There were no statistically significant
changes in wall time for the rest of the benchmark matrix.
2. MariaDB [03] achieved 95% CIs [5.24, 10.71]% and [20.22, 25.97]%
more transactions per minute (TPM), respectively, under the medium-
and the high-concurrency conditions, when overcommitting memory.
There were no statistically significant changes in TPM for the rest
of the benchmark matrix.
3. Memcached [04] achieved 95% CIs [23.54, 32.25]%, [20.76, 41.61]%
and [21.59, 30.02]% more operations per second (OPS), respectively,
for sequential access, random access and Gaussian (distribution)
access, when THP=always; 95% CIs [13.85, 15.97]% and
[23.94, 29.92]% more OPS, respectively, for random access and
Gaussian access, when THP=never. There were no statistically
significant changes in OPS for the rest of the benchmark matrix.
4. MongoDB [05] achieved 95% CIs [2.23, 3.44]%, [6.97, 9.73]% and
[2.16, 3.55]% more operations per second (OPS), respectively, for
exponential (distribution) access, random access and Zipfian
(distribution) access, when underutilizing memory; 95% CIs
[8.83, 10.03]%, [21.12, 23.14]% and [5.53, 6.46]% more OPS,
respectively, for exponential access, random access and Zipfian
access, when overcommitting memory.
On 5.15:
5. Apache Cassandra [06] achieved 95% CIs [1.06, 4.10]%, [1.94, 5.43]%
and [4.11, 7.50]% more operations per second (OPS), respectively,
for exponential (distribution) access, random access and Zipfian
(distribution) access, when swap was off; 95% CIs [0.50, 2.60]%,
[6.51, 8.77]% and [3.29, 6.75]% more OPS, respectively, for
exponential access, random access and Zipfian access, when swap was
on.
6. Apache Hadoop [07] took 95% CIs [5.31, 9.69]% and [2.02, 7.86]%
less average wall time to finish twelve parallel TeraSort jobs,
respectively, under the medium- and the high-concurrency
conditions, when swap was on. There were no statistically
significant changes in average wall time for the rest of the
benchmark matrix.
7. PostgreSQL [08] achieved 95% CI [1.75, 6.42]% more transactions per
minute (TPM) under the high-concurrency condition, when swap was
off; 95% CIs [12.82, 18.69]% and [22.70, 46.86]% more TPM,
respectively, under the medium- and the high-concurrency
conditions, when swap was on. There were no statistically
significant changes in TPM for the rest of the benchmark matrix.
8. Redis [09] achieved 95% CIs [0.58, 5.94]%, [6.55, 14.58]% and
[11.47, 19.36]% more total operations per second (OPS),
respectively, for sequential access, random access and Gaussian
(distribution) access, when THP=always; 95% CIs [1.27, 3.54]%,
[10.11, 14.81]% and [8.75, 13.64]% more total OPS, respectively,
for sequential access, random access and Gaussian access, when
THP=never.
Our lab results
---------------
To supplement the above results, we ran the following benchmark suites
on 5.16-rc7 and found no regressions [10].
fs_fio_bench_hdd_mq pft
fs_lmbench pgsql-hammerdb
fs_parallelio redis
fs_postmark stream
hackbench sysbenchthread
kernbench tpcc_spark
memcached unixbench
multichase vm-scalability
mutilate will-it-scale
nginx
[01] https://trends.google.com
[02] https://lore.kernel.org/r/20211102002002.92051-1-bot@edi.works/
[03] https://lore.kernel.org/r/20211009054315.47073-1-bot@edi.works/
[04] https://lore.kernel.org/r/20211021194103.65648-1-bot@edi.works/
[05] https://lore.kernel.org/r/20211109021346.50266-1-bot@edi.works/
[06] https://lore.kernel.org/r/20211202062806.80365-1-bot@edi.works/
[07] https://lore.kernel.org/r/20211209072416.33606-1-bot@edi.works/
[08] https://lore.kernel.org/r/20211218071041.24077-1-bot@edi.works/
[09] https://lore.kernel.org/r/20211122053248.57311-1-bot@edi.works/
[10] https://lore.kernel.org/r/20220104202247.2903702-1-yuzhao@google.com/
Read-world applications
=======================
Third-party testimonials
------------------------
Konstantin reported [11]:
I have Archlinux with 8G RAM + zswap + swap. While developing, I
have lots of apps opened such as multiple LSP-servers for different
langs, chats, two browsers, etc... Usually, my system gets quickly
to a point of SWAP-storms, where I have to kill LSP-servers,
restart browsers to free memory, etc, otherwise the system lags
heavily and is barely usable.
1.5 day ago I migrated from 5.11.15 kernel to 5.12 + the LRU
patchset, and I started up by opening lots of apps to create memory
pressure, and worked for a day like this. Till now I had not a
single SWAP-storm, and mind you I got 3.4G in SWAP. I was never
getting to the point of 3G in SWAP before without a single
SWAP-storm.
Vaibhav from IBM reported [12]:
In a synthetic MongoDB Benchmark, seeing an average of ~19%
throughput improvement on POWER10(Radix MMU + 64K Page Size) with
MGLRU patches on top of 5.16 kernel for MongoDB + YCSB across
three different request distributions, namely, Exponential, Uniform
and Zipfan.
Shuang from U of Rochester reported [13]:
With the MGLRU, fio achieved 95% CIs [38.95, 40.26]%, [4.12, 6.64]%
and [9.26, 10.36]% higher throughput, respectively, for random
access, Zipfian (distribution) access and Gaussian (distribution)
access, when the average number of jobs per CPU is 1; 95% CIs
[42.32, 49.15]%, [9.44, 9.89]% and [20.99, 22.86]% higher
throughput, respectively, for random access, Zipfian access and
Gaussian access, when the average number of jobs per CPU is 2.
Daniel from Michigan Tech reported [14]:
With Memcached allocating ~100GB of byte-addressable Optante,
performance improvement in terms of throughput (measured as queries
per second) was about 10% for a series of workloads.
Large-scale deployments
-----------------------
We've rolled out MGLRU to tens of millions of ChromeOS users and
about a million Android users. Google's fleetwide profiling [15] shows
an overall 40% decrease in kswapd CPU usage, in addition to
improvements in other UX metrics, e.g., an 85% decrease in the number
of low-memory kills at the 75th percentile and an 18% decrease in
app launch time at the 50th percentile.
The downstream kernels that have been using MGLRU include:
1. Android [16]
2. Arch Linux Zen [17]
3. Armbian [18]
4. ChromeOS [19]
5. Liquorix [20]
6. OpenWrt [21]
7. post-factum [22]
8. XanMod [23]
[11] https://lore.kernel.org/r/140226722f2032c86301fbd326d91baefe3d7d23.camel@yandex.ru/
[12] https://lore.kernel.org/r/87czj3mux0.fsf@vajain21.in.ibm.com/
[13] https://lore.kernel.org/r/20220105024423.26409-1-szhai2@cs.rochester.edu/
[14] https://lore.kernel.org/r/CA+4-3vksGvKd18FgRinxhqHetBS1hQekJE2gwco8Ja-bJWKtFw@mail.gmail.com/
[15] https://dl.acm.org/doi/10.1145/2749469.2750392
[16] https://android.com
[17] https://archlinux.org
[18] https://armbian.com
[19] https://chromium.org
[20] https://liquorix.net
[21] https://openwrt.org
[22] https://codeberg.org/pf-kernel
[23] https://xanmod.org
Summary
=======
The facts are:
1. The independent lab results and the real-world applications
indicate substantial improvements; there are no known regressions.
2. Thrashing prevention, working set estimation and proactive reclaim
work out of the box; there are no equivalent solutions.
3. There is a lot of new code; no smaller changes have been
demonstrated similar effects.
Our options, accordingly, are:
1. Given the amount of evidence, the reported improvements will likely
materialize for a wide range of workloads.
2. Gauging the interest from the past discussions, the new features
will likely be put to use for both personal computers and data
centers.
3. Based on Google's track record, the new code will likely be well
maintained in the long term. It'd be more difficult if not
impossible to achieve similar effects with other approaches.
This patch (of 14):
Some architectures automatically set the accessed bit in PTEs, e.g., x86
and arm64 v8.2. On architectures that do not have this capability,
clearing the accessed bit in a PTE usually triggers a page fault following
the TLB miss of this PTE (to emulate the accessed bit).
Being aware of this capability can help make better decisions, e.g.,
whether to spread the work out over a period of time to reduce bursty page
faults when trying to clear the accessed bit in many PTEs.
Note that theoretically this capability can be unreliable, e.g.,
hotplugged CPUs might be different from builtin ones. Therefore it should
not be used in architecture-independent code that involves correctness,
e.g., to determine whether TLB flushes are required (in combination with
the accessed bit).
Link: https://lkml.kernel.org/r/20220918080010.2920238-1-yuzhao@google.com
Link: https://lkml.kernel.org/r/20220918080010.2920238-2-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Barry Song <baohua@kernel.org>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Acked-by: Will Deacon <will@kernel.org>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: linux-arm-kernel@lists.infradead.org
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-18 07:59:58 +00:00
|
|
|
if (!arch_has_hw_pte_young() && !pte_young(vmf->orig_pte)) {
|
2019-10-11 14:09:39 +00:00
|
|
|
pte_t entry;
|
2005-11-29 22:07:55 +00:00
|
|
|
|
2019-10-11 14:09:39 +00:00
|
|
|
vmf->pte = pte_offset_map_lock(mm, vmf->pmd, addr, &vmf->ptl);
|
2020-03-06 06:28:32 +00:00
|
|
|
locked = true;
|
2019-10-11 14:09:39 +00:00
|
|
|
if (!likely(pte_same(*vmf->pte, vmf->orig_pte))) {
|
|
|
|
/*
|
|
|
|
* Other thread has already handled the fault
|
2020-05-27 02:25:18 +00:00
|
|
|
* and update local tlb only
|
2019-10-11 14:09:39 +00:00
|
|
|
*/
|
2020-05-27 02:25:18 +00:00
|
|
|
update_mmu_tlb(vma, addr, vmf->pte);
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
ret = -EAGAIN;
|
2019-10-11 14:09:39 +00:00
|
|
|
goto pte_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
entry = pte_mkyoung(vmf->orig_pte);
|
|
|
|
if (ptep_set_access_flags(vma, addr, vmf->pte, entry, 0))
|
|
|
|
update_mmu_cache(vma, addr, vmf->pte);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This really shouldn't fail, because the page is there
|
|
|
|
* in the page tables. But it might just be unreadable,
|
|
|
|
* in which case we just give up and fill the result with
|
|
|
|
* zeroes.
|
|
|
|
*/
|
|
|
|
if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE)) {
|
2020-03-06 06:28:32 +00:00
|
|
|
if (locked)
|
|
|
|
goto warn;
|
|
|
|
|
|
|
|
/* Re-validate under PTL if the page is still mapped */
|
|
|
|
vmf->pte = pte_offset_map_lock(mm, vmf->pmd, addr, &vmf->ptl);
|
|
|
|
locked = true;
|
|
|
|
if (!likely(pte_same(*vmf->pte, vmf->orig_pte))) {
|
2020-05-27 02:25:18 +00:00
|
|
|
/* The PTE changed under us, update local tlb */
|
|
|
|
update_mmu_tlb(vma, addr, vmf->pte);
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
ret = -EAGAIN;
|
2020-03-06 06:28:32 +00:00
|
|
|
goto pte_unlock;
|
|
|
|
}
|
|
|
|
|
2005-11-29 22:07:55 +00:00
|
|
|
/*
|
2020-06-04 23:49:43 +00:00
|
|
|
* The same page can be mapped back since last copy attempt.
|
2020-03-06 06:28:32 +00:00
|
|
|
* Try to copy again under PTL.
|
2005-11-29 22:07:55 +00:00
|
|
|
*/
|
2020-03-06 06:28:32 +00:00
|
|
|
if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE)) {
|
|
|
|
/*
|
|
|
|
* Give a warn in case there can be some obscure
|
|
|
|
* use-case
|
|
|
|
*/
|
|
|
|
warn:
|
|
|
|
WARN_ON_ONCE(1);
|
|
|
|
clear_page(kaddr);
|
|
|
|
}
|
2019-10-11 14:09:39 +00:00
|
|
|
}
|
|
|
|
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
ret = 0;
|
2019-10-11 14:09:39 +00:00
|
|
|
|
|
|
|
pte_unlock:
|
2020-03-06 06:28:32 +00:00
|
|
|
if (locked)
|
2019-10-11 14:09:39 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
|
|
|
kunmap_atomic(kaddr);
|
|
|
|
flush_dcache_page(dst);
|
|
|
|
|
|
|
|
return ret;
|
2005-11-28 22:34:23 +00:00
|
|
|
}
|
|
|
|
|
2016-01-14 23:20:12 +00:00
|
|
|
static gfp_t __get_fault_gfp_mask(struct vm_area_struct *vma)
|
|
|
|
{
|
|
|
|
struct file *vm_file = vma->vm_file;
|
|
|
|
|
|
|
|
if (vm_file)
|
|
|
|
return mapping_gfp_mask(vm_file->f_mapping) | __GFP_FS | __GFP_IO;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Special mappings (e.g. VDSO) do not have any file so fake
|
|
|
|
* a default GFP_KERNEL for them.
|
|
|
|
*/
|
|
|
|
return GFP_KERNEL;
|
|
|
|
}
|
|
|
|
|
2014-04-03 21:48:15 +00:00
|
|
|
/*
|
|
|
|
* Notify the address space that the page is about to become writable so that
|
|
|
|
* it can prohibit this or wait for the page to get into an appropriate state.
|
|
|
|
*
|
|
|
|
* We do this without the lock held, so that it can sleep if it needs to.
|
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_page_mkwrite(struct vm_fault *vmf)
|
2014-04-03 21:48:15 +00:00
|
|
|
{
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret;
|
2016-12-14 23:07:30 +00:00
|
|
|
struct page *page = vmf->page;
|
|
|
|
unsigned int old_flags = vmf->flags;
|
2014-04-03 21:48:15 +00:00
|
|
|
|
2016-12-14 23:07:30 +00:00
|
|
|
vmf->flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
|
2014-04-03 21:48:15 +00:00
|
|
|
|
2019-08-20 14:55:16 +00:00
|
|
|
if (vmf->vma->vm_file &&
|
|
|
|
IS_SWAPFILE(vmf->vma->vm_file->f_mapping->host))
|
|
|
|
return VM_FAULT_SIGBUS;
|
|
|
|
|
2017-02-24 22:56:41 +00:00
|
|
|
ret = vmf->vma->vm_ops->page_mkwrite(vmf);
|
2016-12-14 23:07:30 +00:00
|
|
|
/* Restore original flags so that caller is not surprised */
|
|
|
|
vmf->flags = old_flags;
|
2014-04-03 21:48:15 +00:00
|
|
|
if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))
|
|
|
|
return ret;
|
|
|
|
if (unlikely(!(ret & VM_FAULT_LOCKED))) {
|
|
|
|
lock_page(page);
|
|
|
|
if (!page->mapping) {
|
|
|
|
unlock_page(page);
|
|
|
|
return 0; /* retry */
|
|
|
|
}
|
|
|
|
ret |= VM_FAULT_LOCKED;
|
|
|
|
} else
|
|
|
|
VM_BUG_ON_PAGE(!PageLocked(page), page);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2016-12-14 23:07:27 +00:00
|
|
|
/*
|
|
|
|
* Handle dirtying of a page in shared file mapping on a write fault.
|
|
|
|
*
|
|
|
|
* The function expects the page to be locked and unlocks it.
|
|
|
|
*/
|
2019-12-01 01:50:22 +00:00
|
|
|
static vm_fault_t fault_dirty_shared_page(struct vm_fault *vmf)
|
2016-12-14 23:07:27 +00:00
|
|
|
{
|
2019-12-01 01:50:22 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2016-12-14 23:07:27 +00:00
|
|
|
struct address_space *mapping;
|
2019-12-01 01:50:22 +00:00
|
|
|
struct page *page = vmf->page;
|
2016-12-14 23:07:27 +00:00
|
|
|
bool dirtied;
|
|
|
|
bool page_mkwrite = vma->vm_ops && vma->vm_ops->page_mkwrite;
|
|
|
|
|
|
|
|
dirtied = set_page_dirty(page);
|
|
|
|
VM_BUG_ON_PAGE(PageAnon(page), page);
|
|
|
|
/*
|
|
|
|
* Take a local copy of the address_space - page.mapping may be zeroed
|
|
|
|
* by truncate after unlock_page(). The address_space itself remains
|
|
|
|
* pinned by vma->vm_file's reference. We rely on unlock_page()'s
|
|
|
|
* release semantics to prevent the compiler from undoing this copying.
|
|
|
|
*/
|
|
|
|
mapping = page_rmapping(page);
|
|
|
|
unlock_page(page);
|
|
|
|
|
2019-12-01 01:50:22 +00:00
|
|
|
if (!page_mkwrite)
|
|
|
|
file_update_time(vma->vm_file);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Throttle page dirtying rate down to writeback speed.
|
|
|
|
*
|
|
|
|
* mapping may be NULL here because some device drivers do not
|
|
|
|
* set page.mapping but still dirty their pages
|
|
|
|
*
|
2020-06-09 04:33:54 +00:00
|
|
|
* Drop the mmap_lock before waiting on IO, if we can. The file
|
2019-12-01 01:50:22 +00:00
|
|
|
* is pinning the mapping, as per above.
|
|
|
|
*/
|
2016-12-14 23:07:27 +00:00
|
|
|
if ((dirtied || page_mkwrite) && mapping) {
|
2019-12-01 01:50:22 +00:00
|
|
|
struct file *fpin;
|
|
|
|
|
|
|
|
fpin = maybe_unlock_mmap_for_io(vmf, NULL);
|
2016-12-14 23:07:27 +00:00
|
|
|
balance_dirty_pages_ratelimited(mapping);
|
2019-12-01 01:50:22 +00:00
|
|
|
if (fpin) {
|
|
|
|
fput(fpin);
|
mm: avoid unnecessary page fault retires on shared memory types
I observed that for each of the shared file-backed page faults, we're very
likely to retry one more time for the 1st write fault upon no page. It's
because we'll need to release the mmap lock for dirty rate limit purpose
with balance_dirty_pages_ratelimited() (in fault_dirty_shared_page()).
Then after that throttling we return VM_FAULT_RETRY.
We did that probably because VM_FAULT_RETRY is the only way we can return
to the fault handler at that time telling it we've released the mmap lock.
However that's not ideal because it's very likely the fault does not need
to be retried at all since the pgtable was well installed before the
throttling, so the next continuous fault (including taking mmap read lock,
walk the pgtable, etc.) could be in most cases unnecessary.
It's not only slowing down page faults for shared file-backed, but also add
more mmap lock contention which is in most cases not needed at all.
To observe this, one could try to write to some shmem page and look at
"pgfault" value in /proc/vmstat, then we should expect 2 counts for each
shmem write simply because we retried, and vm event "pgfault" will capture
that.
To make it more efficient, add a new VM_FAULT_COMPLETED return code just to
show that we've completed the whole fault and released the lock. It's also
a hint that we should very possibly not need another fault immediately on
this page because we've just completed it.
This patch provides a ~12% perf boost on my aarch64 test VM with a simple
program sequentially dirtying 400MB shmem file being mmap()ed and these are
the time it needs:
Before: 650.980 ms (+-1.94%)
After: 569.396 ms (+-1.38%)
I believe it could help more than that.
We need some special care on GUP and the s390 pgfault handler (for gmap
code before returning from pgfault), the rest changes in the page fault
handlers should be relatively straightforward.
Another thing to mention is that mm_account_fault() does take this new
fault as a generic fault to be accounted, unlike VM_FAULT_RETRY.
I explicitly didn't touch hmm_vma_fault() and break_ksm() because they do
not handle VM_FAULT_RETRY even with existing code, so I'm literally keeping
them as-is.
Link: https://lkml.kernel.org/r/20220530183450.42886-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vineet Gupta <vgupta@kernel.org>
Acked-by: Guo Ren <guoren@kernel.org>
Acked-by: Max Filippov <jcmvbkbc@gmail.com>
Acked-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Acked-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc)
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> [arm part]
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Stafford Horne <shorne@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: Brian Cain <bcain@quicinc.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Richard Weinberger <richard@nod.at>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Janosch Frank <frankja@linux.ibm.com>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Sven Schnelle <svens@linux.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Will Deacon <will@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Chris Zankel <chris@zankel.net>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dinh Nguyen <dinguyen@kernel.org>
Cc: Rich Felker <dalias@libc.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Helge Deller <deller@gmx.de>
Cc: Yoshinori Sato <ysato@users.osdn.me>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-30 18:34:50 +00:00
|
|
|
return VM_FAULT_COMPLETED;
|
2019-12-01 01:50:22 +00:00
|
|
|
}
|
2016-12-14 23:07:27 +00:00
|
|
|
}
|
|
|
|
|
2019-12-01 01:50:22 +00:00
|
|
|
return 0;
|
2016-12-14 23:07:27 +00:00
|
|
|
}
|
|
|
|
|
mm: refactor do_wp_page, extract the reuse case
Currently do_wp_page contains 265 code lines. It also contains 9 goto
statements, of which 5 are targeting labels which are not cleanup
related. This makes the function extremely difficult to understand.
The following patches are an attempt at breaking the function to its
basic components, and making it easier to understand.
The patches are straight forward function extractions from do_wp_page.
As we extract functions, we remove unneeded parameters and simplify the
code as much as possible. However, the functionality is supposed to
remain completely unchanged. The patches also attempt to document the
functionality of each extracted function. In patch 2, we split the
unlock logic to the contain logic relevant to specific needs of each use
case, instead of having huge number of conditional decisions in a single
unlock flow.
This patch (of 4):
When do_wp_page is ending, in several cases it needs to reuse the existing
page. This is achieved by making the page table writable, and possibly
updating the page-cache state.
Currently, this logic was "called" by using a goto jump. This makes
following the control flow of the function harder. It is also against the
coding style guidelines for using goto.
As the code can easily be refactored into a specialized function, refactor
it out and simplify the code flow in do_wp_page.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Haggai Eran <haggaie@mellanox.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Feiner <pfeiner@google.com>
Cc: Michel Lespinasse <walken@google.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:46:25 +00:00
|
|
|
/*
|
|
|
|
* Handle write page faults for pages that can be reused in the current vma
|
|
|
|
*
|
|
|
|
* This can happen either due to the mapping being with the VM_SHARED flag,
|
|
|
|
* or due to us being the last reference standing to the page. In either
|
|
|
|
* case, all we need to do here is to mark the page as writable and update
|
|
|
|
* any related book-keeping.
|
|
|
|
*/
|
2016-12-14 23:07:36 +00:00
|
|
|
static inline void wp_page_reuse(struct vm_fault *vmf)
|
2016-12-14 23:06:58 +00:00
|
|
|
__releases(vmf->ptl)
|
mm: refactor do_wp_page, extract the reuse case
Currently do_wp_page contains 265 code lines. It also contains 9 goto
statements, of which 5 are targeting labels which are not cleanup
related. This makes the function extremely difficult to understand.
The following patches are an attempt at breaking the function to its
basic components, and making it easier to understand.
The patches are straight forward function extractions from do_wp_page.
As we extract functions, we remove unneeded parameters and simplify the
code as much as possible. However, the functionality is supposed to
remain completely unchanged. The patches also attempt to document the
functionality of each extracted function. In patch 2, we split the
unlock logic to the contain logic relevant to specific needs of each use
case, instead of having huge number of conditional decisions in a single
unlock flow.
This patch (of 4):
When do_wp_page is ending, in several cases it needs to reuse the existing
page. This is achieved by making the page table writable, and possibly
updating the page-cache state.
Currently, this logic was "called" by using a goto jump. This makes
following the control flow of the function harder. It is also against the
coding style guidelines for using goto.
As the code can easily be refactored into a specialized function, refactor
it out and simplify the code flow in do_wp_page.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Haggai Eran <haggaie@mellanox.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Feiner <pfeiner@google.com>
Cc: Michel Lespinasse <walken@google.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:46:25 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2016-12-14 23:07:33 +00:00
|
|
|
struct page *page = vmf->page;
|
mm: refactor do_wp_page, extract the reuse case
Currently do_wp_page contains 265 code lines. It also contains 9 goto
statements, of which 5 are targeting labels which are not cleanup
related. This makes the function extremely difficult to understand.
The following patches are an attempt at breaking the function to its
basic components, and making it easier to understand.
The patches are straight forward function extractions from do_wp_page.
As we extract functions, we remove unneeded parameters and simplify the
code as much as possible. However, the functionality is supposed to
remain completely unchanged. The patches also attempt to document the
functionality of each extracted function. In patch 2, we split the
unlock logic to the contain logic relevant to specific needs of each use
case, instead of having huge number of conditional decisions in a single
unlock flow.
This patch (of 4):
When do_wp_page is ending, in several cases it needs to reuse the existing
page. This is achieved by making the page table writable, and possibly
updating the page-cache state.
Currently, this logic was "called" by using a goto jump. This makes
following the control flow of the function harder. It is also against the
coding style guidelines for using goto.
As the code can easily be refactored into a specialized function, refactor
it out and simplify the code flow in do_wp_page.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Haggai Eran <haggaie@mellanox.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Feiner <pfeiner@google.com>
Cc: Michel Lespinasse <walken@google.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:46:25 +00:00
|
|
|
pte_t entry;
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
VM_BUG_ON(!(vmf->flags & FAULT_FLAG_WRITE));
|
2022-07-26 06:24:36 +00:00
|
|
|
VM_BUG_ON(page && PageAnon(page) && !PageAnonExclusive(page));
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
|
mm: refactor do_wp_page, extract the reuse case
Currently do_wp_page contains 265 code lines. It also contains 9 goto
statements, of which 5 are targeting labels which are not cleanup
related. This makes the function extremely difficult to understand.
The following patches are an attempt at breaking the function to its
basic components, and making it easier to understand.
The patches are straight forward function extractions from do_wp_page.
As we extract functions, we remove unneeded parameters and simplify the
code as much as possible. However, the functionality is supposed to
remain completely unchanged. The patches also attempt to document the
functionality of each extracted function. In patch 2, we split the
unlock logic to the contain logic relevant to specific needs of each use
case, instead of having huge number of conditional decisions in a single
unlock flow.
This patch (of 4):
When do_wp_page is ending, in several cases it needs to reuse the existing
page. This is achieved by making the page table writable, and possibly
updating the page-cache state.
Currently, this logic was "called" by using a goto jump. This makes
following the control flow of the function harder. It is also against the
coding style guidelines for using goto.
As the code can easily be refactored into a specialized function, refactor
it out and simplify the code flow in do_wp_page.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Haggai Eran <haggaie@mellanox.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Feiner <pfeiner@google.com>
Cc: Michel Lespinasse <walken@google.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:46:25 +00:00
|
|
|
/*
|
|
|
|
* Clear the pages cpupid information as the existing
|
|
|
|
* information potentially belongs to a now completely
|
|
|
|
* unrelated process.
|
|
|
|
*/
|
|
|
|
if (page)
|
|
|
|
page_cpupid_xchg_last(page, (1 << LAST_CPUPID_SHIFT) - 1);
|
|
|
|
|
2016-12-14 23:07:16 +00:00
|
|
|
flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
|
|
|
|
entry = pte_mkyoung(vmf->orig_pte);
|
mm: refactor do_wp_page, extract the reuse case
Currently do_wp_page contains 265 code lines. It also contains 9 goto
statements, of which 5 are targeting labels which are not cleanup
related. This makes the function extremely difficult to understand.
The following patches are an attempt at breaking the function to its
basic components, and making it easier to understand.
The patches are straight forward function extractions from do_wp_page.
As we extract functions, we remove unneeded parameters and simplify the
code as much as possible. However, the functionality is supposed to
remain completely unchanged. The patches also attempt to document the
functionality of each extracted function. In patch 2, we split the
unlock logic to the contain logic relevant to specific needs of each use
case, instead of having huge number of conditional decisions in a single
unlock flow.
This patch (of 4):
When do_wp_page is ending, in several cases it needs to reuse the existing
page. This is achieved by making the page table writable, and possibly
updating the page-cache state.
Currently, this logic was "called" by using a goto jump. This makes
following the control flow of the function harder. It is also against the
coding style guidelines for using goto.
As the code can easily be refactored into a specialized function, refactor
it out and simplify the code flow in do_wp_page.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Haggai Eran <haggaie@mellanox.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Feiner <pfeiner@google.com>
Cc: Michel Lespinasse <walken@google.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:46:25 +00:00
|
|
|
entry = maybe_mkwrite(pte_mkdirty(entry), vma);
|
2016-12-14 23:06:58 +00:00
|
|
|
if (ptep_set_access_flags(vma, vmf->address, vmf->pte, entry, 1))
|
|
|
|
update_mmu_cache(vma, vmf->address, vmf->pte);
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2020-08-21 23:49:58 +00:00
|
|
|
count_vm_event(PGREUSE);
|
mm: refactor do_wp_page, extract the reuse case
Currently do_wp_page contains 265 code lines. It also contains 9 goto
statements, of which 5 are targeting labels which are not cleanup
related. This makes the function extremely difficult to understand.
The following patches are an attempt at breaking the function to its
basic components, and making it easier to understand.
The patches are straight forward function extractions from do_wp_page.
As we extract functions, we remove unneeded parameters and simplify the
code as much as possible. However, the functionality is supposed to
remain completely unchanged. The patches also attempt to document the
functionality of each extracted function. In patch 2, we split the
unlock logic to the contain logic relevant to specific needs of each use
case, instead of having huge number of conditional decisions in a single
unlock flow.
This patch (of 4):
When do_wp_page is ending, in several cases it needs to reuse the existing
page. This is achieved by making the page table writable, and possibly
updating the page-cache state.
Currently, this logic was "called" by using a goto jump. This makes
following the control flow of the function harder. It is also against the
coding style guidelines for using goto.
As the code can easily be refactored into a specialized function, refactor
it out and simplify the code flow in do_wp_page.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Haggai Eran <haggaie@mellanox.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Feiner <pfeiner@google.com>
Cc: Michel Lespinasse <walken@google.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-14 22:46:25 +00:00
|
|
|
}
|
|
|
|
|
2015-04-14 22:46:32 +00:00
|
|
|
/*
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
* Handle the case of a page which we actually need to copy to a new page,
|
|
|
|
* either due to COW or unsharing.
|
2015-04-14 22:46:32 +00:00
|
|
|
*
|
2020-06-09 04:33:54 +00:00
|
|
|
* Called with mmap_lock locked and the old page referenced, but
|
2015-04-14 22:46:32 +00:00
|
|
|
* without the ptl held.
|
|
|
|
*
|
|
|
|
* High level logic flow:
|
|
|
|
*
|
|
|
|
* - Allocate a page, copy the content of the old page to the new one.
|
|
|
|
* - Handle book keeping and accounting - cgroups, mmu-notifiers, etc.
|
|
|
|
* - Take the PTL. If the pte changed, bail out and release the allocated page
|
|
|
|
* - If the pte is still the way we remember it, update the page table and all
|
|
|
|
* relevant references. This includes dropping the reference the page-table
|
|
|
|
* held to the old page, as well as updating the rmap.
|
|
|
|
* - In any case, unlock the PTL and drop the reference we took to the old page.
|
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t wp_page_copy(struct vm_fault *vmf)
|
2015-04-14 22:46:32 +00:00
|
|
|
{
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
const bool unshare = vmf->flags & FAULT_FLAG_UNSHARE;
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2016-07-26 22:25:20 +00:00
|
|
|
struct mm_struct *mm = vma->vm_mm;
|
2023-01-16 19:18:11 +00:00
|
|
|
struct folio *old_folio = NULL;
|
|
|
|
struct folio *new_folio = NULL;
|
2015-04-14 22:46:32 +00:00
|
|
|
pte_t entry;
|
|
|
|
int page_copied = 0;
|
2018-12-28 08:38:09 +00:00
|
|
|
struct mmu_notifier_range range;
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
int ret;
|
2015-04-14 22:46:32 +00:00
|
|
|
|
2022-06-01 22:55:25 +00:00
|
|
|
delayacct_wpcopy_start();
|
|
|
|
|
2023-01-16 19:18:11 +00:00
|
|
|
if (vmf->page)
|
|
|
|
old_folio = page_folio(vmf->page);
|
2015-04-14 22:46:32 +00:00
|
|
|
if (unlikely(anon_vma_prepare(vma)))
|
|
|
|
goto oom;
|
|
|
|
|
2016-12-14 23:07:16 +00:00
|
|
|
if (is_zero_pfn(pte_pfn(vmf->orig_pte))) {
|
2023-01-16 19:18:09 +00:00
|
|
|
new_folio = vma_alloc_zeroed_movable_folio(vma, vmf->address);
|
|
|
|
if (!new_folio)
|
2015-04-14 22:46:32 +00:00
|
|
|
goto oom;
|
|
|
|
} else {
|
2023-01-16 19:18:11 +00:00
|
|
|
new_folio = vma_alloc_folio(GFP_HIGHUSER_MOVABLE, 0, vma,
|
|
|
|
vmf->address, false);
|
|
|
|
if (!new_folio)
|
2015-04-14 22:46:32 +00:00
|
|
|
goto oom;
|
2019-10-11 14:09:39 +00:00
|
|
|
|
2023-01-16 19:18:11 +00:00
|
|
|
ret = __wp_page_copy_user(&new_folio->page, vmf->page, vmf);
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
if (ret) {
|
2019-10-11 14:09:39 +00:00
|
|
|
/*
|
|
|
|
* COW failed, if the fault was solved by other,
|
|
|
|
* it's fine. If not, userspace would re-fault on
|
|
|
|
* the same address and we will handle the fault
|
|
|
|
* from the second attempt.
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
* The -EHWPOISON case will not be retried.
|
2019-10-11 14:09:39 +00:00
|
|
|
*/
|
2023-01-16 19:18:11 +00:00
|
|
|
folio_put(new_folio);
|
|
|
|
if (old_folio)
|
|
|
|
folio_put(old_folio);
|
2022-06-01 22:55:25 +00:00
|
|
|
|
|
|
|
delayacct_wpcopy_end();
|
mm, hwpoison: try to recover from copy-on write faults
Patch series "Copy-on-write poison recovery", v3.
Part 1 deals with the process that triggered the copy on write fault with
a store to a shared read-only page. That process is send a SIGBUS with
the usual machine check decoration to specify the virtual address of the
lost page, together with the scope.
Part 2 sets up to asynchronously take the page with the uncorrected error
offline to prevent additional machine check faults. H/t to Miaohe Lin
<linmiaohe@huawei.com> and Shuai Xue <xueshuai@linux.alibaba.com> for
pointing me to the existing function to queue a call to memory_failure().
On x86 there is some duplicate reporting (because the error is also
signalled by the memory controller as well as by the core that triggered
the machine check). Console logs look like this:
This patch (of 2):
If the kernel is copying a page as the result of a copy-on-write
fault and runs into an uncorrectable error, Linux will crash because
it does not have recovery code for this case where poison is consumed
by the kernel.
It is easy to set up a test case. Just inject an error into a private
page, fork(2), and have the child process write to the page.
I wrapped that neatly into a test at:
git://git.kernel.org/pub/scm/linux/kernel/git/aegl/ras-tools.git
just enable ACPI error injection and run:
# ./einj_mem-uc -f copy-on-write
Add a new copy_user_highpage_mc() function that uses copy_mc_to_kernel()
on architectures where that is available (currently x86 and powerpc).
When an error is detected during the page copy, return VM_FAULT_HWPOISON
to caller of wp_page_copy(). This propagates up the call stack. Both x86
and powerpc have code in their fault handler to deal with this code by
sending a SIGBUS to the application.
Note that this patch avoids a system crash and signals the process that
triggered the copy-on-write action. It does not take any action for the
memory error that is still in the shared page. To handle that a call to
memory_failure() is needed. But this cannot be done from wp_page_copy()
because it holds mmap_lock(). Perhaps the architecture fault handlers
can deal with this loose end in a subsequent patch?
On Intel/x86 this loose end will often be handled automatically because
the memory controller provides an additional notification of the h/w
poison in memory, the handler for this will call memory_failure(). This
isn't a 100% solution. If there are multiple errors, not all may be
logged in this way.
[tony.luck@intel.com: add call to kmsan_unpoison_memory(), per Miaohe Lin]
Link: https://lkml.kernel.org/r/20221031201029.102123-2-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-1-tony.luck@intel.com
Link: https://lkml.kernel.org/r/20221021200120.175753-2-tony.luck@intel.com
Signed-off-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-10-21 20:01:19 +00:00
|
|
|
return ret == -EHWPOISON ? VM_FAULT_HWPOISON : 0;
|
2019-10-11 14:09:39 +00:00
|
|
|
}
|
2023-01-16 19:18:11 +00:00
|
|
|
kmsan_copy_page_meta(&new_folio->page, vmf->page);
|
2015-04-14 22:46:32 +00:00
|
|
|
}
|
|
|
|
|
2023-01-16 19:18:11 +00:00
|
|
|
if (mem_cgroup_charge(new_folio, mm, GFP_KERNEL))
|
2015-04-14 22:46:32 +00:00
|
|
|
goto oom_free_new;
|
2023-03-02 11:58:32 +00:00
|
|
|
folio_throttle_swaprate(new_folio, GFP_KERNEL);
|
2015-04-14 22:46:32 +00:00
|
|
|
|
2023-01-16 19:18:11 +00:00
|
|
|
__folio_mark_uptodate(new_folio);
|
2015-06-24 23:57:27 +00:00
|
|
|
|
2023-01-10 02:57:22 +00:00
|
|
|
mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, mm,
|
mm/mmu_notifier: contextual information for event triggering invalidation
CPU page table update can happens for many reasons, not only as a result
of a syscall (munmap(), mprotect(), mremap(), madvise(), ...) but also as
a result of kernel activities (memory compression, reclaim, migration,
...).
Users of mmu notifier API track changes to the CPU page table and take
specific action for them. While current API only provide range of virtual
address affected by the change, not why the changes is happening.
This patchset do the initial mechanical convertion of all the places that
calls mmu_notifier_range_init to also provide the default MMU_NOTIFY_UNMAP
event as well as the vma if it is know (most invalidation happens against
a given vma). Passing down the vma allows the users of mmu notifier to
inspect the new vma page protection.
The MMU_NOTIFY_UNMAP is always the safe default as users of mmu notifier
should assume that every for the range is going away when that event
happens. A latter patch do convert mm call path to use a more appropriate
events for each call.
This is done as 2 patches so that no call site is forgotten especialy
as it uses this following coccinelle patch:
%<----------------------------------------------------------------------
@@
identifier I1, I2, I3, I4;
@@
static inline void mmu_notifier_range_init(struct mmu_notifier_range *I1,
+enum mmu_notifier_event event,
+unsigned flags,
+struct vm_area_struct *vma,
struct mm_struct *I2, unsigned long I3, unsigned long I4) { ... }
@@
@@
-#define mmu_notifier_range_init(range, mm, start, end)
+#define mmu_notifier_range_init(range, event, flags, vma, mm, start, end)
@@
expression E1, E3, E4;
identifier I1;
@@
<...
mmu_notifier_range_init(E1,
+MMU_NOTIFY_UNMAP, 0, I1,
I1->vm_mm, E3, E4)
...>
@@
expression E1, E2, E3, E4;
identifier FN, VMA;
@@
FN(..., struct vm_area_struct *VMA, ...) {
<...
mmu_notifier_range_init(E1,
+MMU_NOTIFY_UNMAP, 0, VMA,
E2, E3, E4)
...> }
@@
expression E1, E2, E3, E4;
identifier FN, VMA;
@@
FN(...) {
struct vm_area_struct *VMA;
<...
mmu_notifier_range_init(E1,
+MMU_NOTIFY_UNMAP, 0, VMA,
E2, E3, E4)
...> }
@@
expression E1, E2, E3, E4;
identifier FN;
@@
FN(...) {
<...
mmu_notifier_range_init(E1,
+MMU_NOTIFY_UNMAP, 0, NULL,
E2, E3, E4)
...> }
---------------------------------------------------------------------->%
Applied with:
spatch --all-includes --sp-file mmu-notifier.spatch fs/proc/task_mmu.c --in-place
spatch --sp-file mmu-notifier.spatch --dir kernel/events/ --in-place
spatch --sp-file mmu-notifier.spatch --dir mm --in-place
Link: http://lkml.kernel.org/r/20190326164747.24405-6-jglisse@redhat.com
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
Reviewed-by: Ralph Campbell <rcampbell@nvidia.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Ross Zwisler <zwisler@kernel.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krcmar <rkrcmar@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-05-14 00:20:49 +00:00
|
|
|
vmf->address & PAGE_MASK,
|
2018-12-28 08:38:09 +00:00
|
|
|
(vmf->address & PAGE_MASK) + PAGE_SIZE);
|
|
|
|
mmu_notifier_invalidate_range_start(&range);
|
2015-04-14 22:46:32 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Re-check the pte - we dropped the lock
|
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->pte = pte_offset_map_lock(mm, vmf->pmd, vmf->address, &vmf->ptl);
|
2016-12-14 23:07:16 +00:00
|
|
|
if (likely(pte_same(*vmf->pte, vmf->orig_pte))) {
|
2023-01-16 19:18:11 +00:00
|
|
|
if (old_folio) {
|
|
|
|
if (!folio_test_anon(old_folio)) {
|
|
|
|
dec_mm_counter(mm, mm_counter_file(&old_folio->page));
|
2022-10-24 05:28:41 +00:00
|
|
|
inc_mm_counter(mm, MM_ANONPAGES);
|
2015-04-14 22:46:32 +00:00
|
|
|
}
|
|
|
|
} else {
|
2022-10-24 05:28:41 +00:00
|
|
|
inc_mm_counter(mm, MM_ANONPAGES);
|
2015-04-14 22:46:32 +00:00
|
|
|
}
|
2016-12-14 23:07:16 +00:00
|
|
|
flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
|
2023-01-16 19:18:11 +00:00
|
|
|
entry = mk_pte(&new_folio->page, vma->vm_page_prot);
|
2021-06-05 03:01:08 +00:00
|
|
|
entry = pte_sw_mkyoung(entry);
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
if (unlikely(unshare)) {
|
|
|
|
if (pte_soft_dirty(vmf->orig_pte))
|
|
|
|
entry = pte_mksoft_dirty(entry);
|
|
|
|
if (pte_uffd_wp(vmf->orig_pte))
|
|
|
|
entry = pte_mkuffd_wp(entry);
|
|
|
|
} else {
|
|
|
|
entry = maybe_mkwrite(pte_mkdirty(entry), vma);
|
|
|
|
}
|
2020-12-29 23:14:43 +00:00
|
|
|
|
2015-04-14 22:46:32 +00:00
|
|
|
/*
|
|
|
|
* Clear the pte entry and flush it first, before updating the
|
2020-12-29 23:14:43 +00:00
|
|
|
* pte with the new entry, to keep TLBs on different CPUs in
|
|
|
|
* sync. This code used to set the new PTE then flush TLBs, but
|
|
|
|
* that left a window where the new PTE could be loaded into
|
|
|
|
* some TLBs while the old PTE remains in others.
|
2015-04-14 22:46:32 +00:00
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
|
2023-01-16 19:18:11 +00:00
|
|
|
folio_add_new_anon_rmap(new_folio, vma, vmf->address);
|
|
|
|
folio_add_lru_vma(new_folio, vma);
|
2015-04-14 22:46:32 +00:00
|
|
|
/*
|
|
|
|
* We call the notify macro here because, when using secondary
|
|
|
|
* mmu page tables (such as kvm shadow page tables), we want the
|
|
|
|
* new page to be mapped directly into the secondary page table.
|
|
|
|
*/
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
BUG_ON(unshare && pte_write(entry));
|
2016-12-14 23:06:58 +00:00
|
|
|
set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
|
|
|
|
update_mmu_cache(vma, vmf->address, vmf->pte);
|
2023-01-16 19:18:11 +00:00
|
|
|
if (old_folio) {
|
2015-04-14 22:46:32 +00:00
|
|
|
/*
|
|
|
|
* Only after switching the pte to the new page may
|
|
|
|
* we remove the mapcount here. Otherwise another
|
|
|
|
* process may come and find the rmap count decremented
|
|
|
|
* before the pte is switched to the new page, and
|
|
|
|
* "reuse" the old page writing into it while our pte
|
|
|
|
* here still points into it and can be read by other
|
|
|
|
* threads.
|
|
|
|
*
|
|
|
|
* The critical issue is to order this
|
|
|
|
* page_remove_rmap with the ptp_clear_flush above.
|
|
|
|
* Those stores are ordered by (if nothing else,)
|
|
|
|
* the barrier present in the atomic_add_negative
|
|
|
|
* in page_remove_rmap.
|
|
|
|
*
|
|
|
|
* Then the TLB flush in ptep_clear_flush ensures that
|
|
|
|
* no process can access the old page before the
|
|
|
|
* decremented mapcount is visible. And the old page
|
|
|
|
* cannot be reused until after the decremented
|
|
|
|
* mapcount is visible. So transitively, TLBs to
|
|
|
|
* old page will be flushed before it can be reused.
|
|
|
|
*/
|
2023-01-16 19:18:11 +00:00
|
|
|
page_remove_rmap(vmf->page, vma, false);
|
2015-04-14 22:46:32 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Free the old page.. */
|
2023-01-16 19:18:11 +00:00
|
|
|
new_folio = old_folio;
|
2015-04-14 22:46:32 +00:00
|
|
|
page_copied = 1;
|
|
|
|
} else {
|
2020-05-27 02:25:18 +00:00
|
|
|
update_mmu_tlb(vma, vmf->address, vmf->pte);
|
2015-04-14 22:46:32 +00:00
|
|
|
}
|
|
|
|
|
2023-01-16 19:18:11 +00:00
|
|
|
if (new_folio)
|
|
|
|
folio_put(new_folio);
|
2015-04-14 22:46:32 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2017-11-16 01:34:11 +00:00
|
|
|
/*
|
|
|
|
* No need to double call mmu_notifier->invalidate_range() callback as
|
|
|
|
* the above ptep_clear_flush_notify() did already call it.
|
|
|
|
*/
|
2018-12-28 08:38:09 +00:00
|
|
|
mmu_notifier_invalidate_range_only_end(&range);
|
2023-01-16 19:18:11 +00:00
|
|
|
if (old_folio) {
|
mm: free idle swap cache page after COW
With commit 09854ba94c6a ("mm: do_wp_page() simplification"), after COW,
the idle swap cache page (neither the page nor the corresponding swap
entry is mapped by any process) will be left in the LRU list, even if it's
in the active list or the head of the inactive list. So, the page
reclaimer may take quite some overhead to reclaim these actually unused
pages.
To help the page reclaiming, in this patch, after COW, the idle swap cache
page will be tried to be freed. To avoid to introduce much overhead to
the hot COW code path,
a) there's almost zero overhead for non-swap case via checking
PageSwapCache() firstly.
b) the page lock is acquired via trylock only.
To test the patch, we used pmbench memory accessing benchmark with
working-set larger than available memory on a 2-socket Intel server with a
NVMe SSD as swap device. Test results shows that the pmbench score
increases up to 23.8% with the decreased size of swap cache and swapin
throughput.
Link: https://lkml.kernel.org/r/20210601053143.1380078-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Johannes Weiner <hannes@cmpxchg.org> [use free_swap_cache()]
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@surriel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Tim Chen <tim.c.chen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-29 02:37:12 +00:00
|
|
|
if (page_copied)
|
2023-01-16 19:18:11 +00:00
|
|
|
free_swap_cache(&old_folio->page);
|
|
|
|
folio_put(old_folio);
|
2015-04-14 22:46:32 +00:00
|
|
|
}
|
2022-06-01 22:55:25 +00:00
|
|
|
|
|
|
|
delayacct_wpcopy_end();
|
2022-10-21 10:11:35 +00:00
|
|
|
return 0;
|
2015-04-14 22:46:32 +00:00
|
|
|
oom_free_new:
|
2023-01-16 19:18:11 +00:00
|
|
|
folio_put(new_folio);
|
2015-04-14 22:46:32 +00:00
|
|
|
oom:
|
2023-01-16 19:18:11 +00:00
|
|
|
if (old_folio)
|
|
|
|
folio_put(old_folio);
|
2022-06-01 22:55:25 +00:00
|
|
|
|
|
|
|
delayacct_wpcopy_end();
|
2015-04-14 22:46:32 +00:00
|
|
|
return VM_FAULT_OOM;
|
|
|
|
}
|
|
|
|
|
2016-12-14 23:07:39 +00:00
|
|
|
/**
|
|
|
|
* finish_mkwrite_fault - finish page fault for a shared mapping, making PTE
|
|
|
|
* writeable once the page is prepared
|
|
|
|
*
|
|
|
|
* @vmf: structure describing the fault
|
|
|
|
*
|
|
|
|
* This function handles all that is needed to finish a write page fault in a
|
|
|
|
* shared mapping due to PTE being read-only once the mapped page is prepared.
|
2019-03-05 23:48:42 +00:00
|
|
|
* It handles locking of PTE and modifying it.
|
2016-12-14 23:07:39 +00:00
|
|
|
*
|
|
|
|
* The function expects the page to be locked or other protection against
|
|
|
|
* concurrent faults / writeback (such as DAX radix tree locks).
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
2021-06-29 02:38:47 +00:00
|
|
|
* Return: %0 on success, %VM_FAULT_NOPAGE when PTE got changed before
|
2019-03-05 23:48:42 +00:00
|
|
|
* we acquired PTE lock.
|
2016-12-14 23:07:39 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf)
|
2016-12-14 23:07:39 +00:00
|
|
|
{
|
|
|
|
WARN_ON_ONCE(!(vmf->vma->vm_flags & VM_SHARED));
|
|
|
|
vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd, vmf->address,
|
|
|
|
&vmf->ptl);
|
|
|
|
/*
|
|
|
|
* We might have raced with another page fault while we released the
|
|
|
|
* pte_offset_map_lock.
|
|
|
|
*/
|
|
|
|
if (!pte_same(*vmf->pte, vmf->orig_pte)) {
|
2020-05-27 02:25:18 +00:00
|
|
|
update_mmu_tlb(vmf->vma, vmf->address, vmf->pte);
|
2016-12-14 23:07:39 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2016-12-14 23:07:42 +00:00
|
|
|
return VM_FAULT_NOPAGE;
|
2016-12-14 23:07:39 +00:00
|
|
|
}
|
|
|
|
wp_page_reuse(vmf);
|
2016-12-14 23:07:42 +00:00
|
|
|
return 0;
|
2016-12-14 23:07:39 +00:00
|
|
|
}
|
|
|
|
|
2015-04-15 23:15:11 +00:00
|
|
|
/*
|
|
|
|
* Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
|
|
|
|
* mapping
|
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t wp_pfn_shared(struct vm_fault *vmf)
|
2015-04-15 23:15:11 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2016-07-26 22:25:20 +00:00
|
|
|
|
2015-04-15 23:15:11 +00:00
|
|
|
if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret;
|
2015-04-15 23:15:11 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2016-12-14 23:07:13 +00:00
|
|
|
vmf->flags |= FAULT_FLAG_MKWRITE;
|
2017-02-24 22:56:41 +00:00
|
|
|
ret = vma->vm_ops->pfn_mkwrite(vmf);
|
2016-12-14 23:07:50 +00:00
|
|
|
if (ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))
|
2015-04-15 23:15:11 +00:00
|
|
|
return ret;
|
2016-12-14 23:07:39 +00:00
|
|
|
return finish_mkwrite_fault(vmf);
|
2015-04-15 23:15:11 +00:00
|
|
|
}
|
2016-12-14 23:07:36 +00:00
|
|
|
wp_page_reuse(vmf);
|
2022-10-21 10:11:35 +00:00
|
|
|
return 0;
|
2015-04-15 23:15:11 +00:00
|
|
|
}
|
|
|
|
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t wp_page_shared(struct vm_fault *vmf)
|
2016-12-14 23:06:58 +00:00
|
|
|
__releases(vmf->ptl)
|
2015-04-14 22:46:35 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2022-10-21 10:11:35 +00:00
|
|
|
vm_fault_t ret = 0;
|
2015-04-14 22:46:35 +00:00
|
|
|
|
2016-12-14 23:07:33 +00:00
|
|
|
get_page(vmf->page);
|
2015-04-14 22:46:35 +00:00
|
|
|
|
|
|
|
if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t tmp;
|
2015-04-14 22:46:35 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2016-12-14 23:07:30 +00:00
|
|
|
tmp = do_page_mkwrite(vmf);
|
2015-04-14 22:46:35 +00:00
|
|
|
if (unlikely(!tmp || (tmp &
|
|
|
|
(VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
|
2016-12-14 23:07:33 +00:00
|
|
|
put_page(vmf->page);
|
2015-04-14 22:46:35 +00:00
|
|
|
return tmp;
|
|
|
|
}
|
2016-12-14 23:07:39 +00:00
|
|
|
tmp = finish_mkwrite_fault(vmf);
|
2016-12-14 23:07:42 +00:00
|
|
|
if (unlikely(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
|
2016-12-14 23:07:33 +00:00
|
|
|
unlock_page(vmf->page);
|
|
|
|
put_page(vmf->page);
|
2016-12-14 23:07:39 +00:00
|
|
|
return tmp;
|
2015-04-14 22:46:35 +00:00
|
|
|
}
|
2016-12-14 23:07:39 +00:00
|
|
|
} else {
|
|
|
|
wp_page_reuse(vmf);
|
2016-12-14 23:07:36 +00:00
|
|
|
lock_page(vmf->page);
|
2015-04-14 22:46:35 +00:00
|
|
|
}
|
2019-12-01 01:50:22 +00:00
|
|
|
ret |= fault_dirty_shared_page(vmf);
|
2016-12-14 23:07:36 +00:00
|
|
|
put_page(vmf->page);
|
2015-04-14 22:46:35 +00:00
|
|
|
|
2019-12-01 01:50:22 +00:00
|
|
|
return ret;
|
2015-04-14 22:46:35 +00:00
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
* This routine handles present pages, when
|
|
|
|
* * users try to write to a shared page (FAULT_FLAG_WRITE)
|
|
|
|
* * GUP wants to take a R/O pin on a possibly shared anonymous page
|
|
|
|
* (FAULT_FLAG_UNSHARE)
|
|
|
|
*
|
|
|
|
* It is done by copying the page to a new address and decrementing the
|
|
|
|
* shared-page counter for the old page.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
|
|
|
* Note that this routine assumes that the protection checks have been
|
|
|
|
* done by the caller (the low-level page fault routine in most cases).
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
* Thus, with FAULT_FLAG_WRITE, we can safely just mark it writable once we've
|
|
|
|
* done any necessary COW.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
* In case of FAULT_FLAG_WRITE, we also mark the page dirty at this point even
|
|
|
|
* though the page will change only once the write actually happens. This
|
|
|
|
* avoids a few races, and potentially makes it more efficient.
|
2005-04-16 22:20:36 +00:00
|
|
|
*
|
2020-06-09 04:33:54 +00:00
|
|
|
* We enter with non-exclusive mmap_lock (to exclude vma changes,
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
* but allow concurrent faults), with pte both mapped and locked.
|
2020-06-09 04:33:54 +00:00
|
|
|
* We return with mmap_lock still held, but pte unmapped and unlocked.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_wp_page(struct vm_fault *vmf)
|
2016-12-14 23:06:58 +00:00
|
|
|
__releases(vmf->ptl)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
const bool unshare = vmf->flags & FAULT_FLAG_UNSHARE;
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2022-11-16 10:26:45 +00:00
|
|
|
struct folio *folio = NULL;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
if (likely(!unshare)) {
|
|
|
|
if (userfaultfd_pte_wp(vma, *vmf->pte)) {
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
|
|
|
return handle_userfault(vmf, VM_UFFD_WP);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Userfaultfd write-protect can defer flushes. Ensure the TLB
|
|
|
|
* is flushed in this case before copying.
|
|
|
|
*/
|
|
|
|
if (unlikely(userfaultfd_wp(vmf->vma) &&
|
|
|
|
mm_tlb_flush_pending(vmf->vma->vm_mm)))
|
|
|
|
flush_tlb_page(vmf->vma, vmf->address);
|
|
|
|
}
|
mm/userfaultfd: fix memory corruption due to writeprotect
Userfaultfd self-test fails occasionally, indicating a memory corruption.
Analyzing this problem indicates that there is a real bug since mmap_lock
is only taken for read in mwriteprotect_range() and defers flushes, and
since there is insufficient consideration of concurrent deferred TLB
flushes in wp_page_copy(). Although the PTE is flushed from the TLBs in
wp_page_copy(), this flush takes place after the copy has already been
performed, and therefore changes of the page are possible between the time
of the copy and the time in which the PTE is flushed.
To make matters worse, memory-unprotection using userfaultfd also poses a
problem. Although memory unprotection is logically a promotion of PTE
permissions, and therefore should not require a TLB flush, the current
userrfaultfd code might actually cause a demotion of the architectural PTE
permission: when userfaultfd_writeprotect() unprotects memory region, it
unintentionally *clears* the RW-bit if it was already set. Note that this
unprotecting a PTE that is not write-protected is a valid use-case: the
userfaultfd monitor might ask to unprotect a region that holds both
write-protected and write-unprotected PTEs.
The scenario that happens in selftests/vm/userfaultfd is as follows:
cpu0 cpu1 cpu2
---- ---- ----
[ Writable PTE
cached in TLB ]
userfaultfd_writeprotect()
[ write-*unprotect* ]
mwriteprotect_range()
mmap_read_lock()
change_protection()
change_protection_range()
...
change_pte_range()
[ *clear* “write”-bit ]
[ defer TLB flushes ]
[ page-fault ]
...
wp_page_copy()
cow_user_page()
[ copy page ]
[ write to old
page ]
...
set_pte_at_notify()
A similar scenario can happen:
cpu0 cpu1 cpu2 cpu3
---- ---- ---- ----
[ Writable PTE
cached in TLB ]
userfaultfd_writeprotect()
[ write-protect ]
[ deferred TLB flush ]
userfaultfd_writeprotect()
[ write-unprotect ]
[ deferred TLB flush]
[ page-fault ]
wp_page_copy()
cow_user_page()
[ copy page ]
... [ write to page ]
set_pte_at_notify()
This race exists since commit 292924b26024 ("userfaultfd: wp: apply
_PAGE_UFFD_WP bit"). Yet, as Yu Zhao pointed, these races became apparent
since commit 09854ba94c6a ("mm: do_wp_page() simplification") which made
wp_page_copy() more likely to take place, specifically if page_count(page)
> 1.
To resolve the aforementioned races, check whether there are pending
flushes on uffd-write-protected VMAs, and if there are, perform a flush
before doing the COW.
Further optimizations will follow to avoid during uffd-write-unprotect
unnecassary PTE write-protection and TLB flushes.
Link: https://lkml.kernel.org/r/20210304095423.3825684-1-namit@vmware.com
Fixes: 09854ba94c6a ("mm: do_wp_page() simplification")
Signed-off-by: Nadav Amit <namit@vmware.com>
Suggested-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Tested-by: Peter Xu <peterx@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Will Deacon <will@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <stable@vger.kernel.org> [5.9+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-03-13 05:08:17 +00:00
|
|
|
|
2016-12-14 23:07:33 +00:00
|
|
|
vmf->page = vm_normal_page(vma, vmf->address, vmf->orig_pte);
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
|
2022-11-16 10:26:45 +00:00
|
|
|
/*
|
|
|
|
* Shared mapping: we are guaranteed to have VM_WRITE and
|
|
|
|
* FAULT_FLAG_WRITE set at this point.
|
|
|
|
*/
|
|
|
|
if (vma->vm_flags & (VM_SHARED | VM_MAYSHARE)) {
|
2008-07-04 16:59:24 +00:00
|
|
|
/*
|
mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared
For VMAs that don't want write notifications, PTEs created for read faults
have their write bit set. If the read fault happens after VM_SOFTDIRTY is
cleared, then the PTE's softdirty bit will remain clear after subsequent
writes.
Here's a simple code snippet to demonstrate the bug:
char* m = mmap(NULL, getpagesize(), PROT_READ | PROT_WRITE,
MAP_ANONYMOUS | MAP_SHARED, -1, 0);
system("echo 4 > /proc/$PPID/clear_refs"); /* clear VM_SOFTDIRTY */
assert(*m == '\0'); /* new PTE allows write access */
assert(!soft_dirty(x));
*m = 'x'; /* should dirty the page */
assert(soft_dirty(x)); /* fails */
With this patch, write notifications are enabled when VM_SOFTDIRTY is
cleared. Furthermore, to avoid unnecessary faults, write notifications
are disabled when VM_SOFTDIRTY is set.
As a side effect of enabling and disabling write notifications with
care, this patch fixes a bug in mprotect where vm_page_prot bits set by
drivers were zapped on mprotect. An analogous bug was fixed in mmap by
commit c9d0bf241451 ("mm: uncached vma support with writenotify").
Signed-off-by: Peter Feiner <pfeiner@google.com>
Reported-by: Peter Feiner <pfeiner@google.com>
Suggested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Jamie Liu <jamieliu@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-13 22:55:46 +00:00
|
|
|
* VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
|
|
|
|
* VM_PFNMAP VMA.
|
2008-07-04 16:59:24 +00:00
|
|
|
*
|
|
|
|
* We should not cow pages in a shared writeable mapping.
|
2015-04-15 23:15:11 +00:00
|
|
|
* Just mark the pages writable and/or call ops->pfn_mkwrite.
|
2008-07-04 16:59:24 +00:00
|
|
|
*/
|
2022-11-16 10:26:45 +00:00
|
|
|
if (!vmf->page)
|
2016-12-14 23:07:16 +00:00
|
|
|
return wp_pfn_shared(vmf);
|
2022-11-16 10:26:45 +00:00
|
|
|
return wp_page_shared(vmf);
|
2008-07-04 16:59:24 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2022-11-16 10:26:45 +00:00
|
|
|
if (vmf->page)
|
|
|
|
folio = page_folio(vmf->page);
|
|
|
|
|
2006-09-26 06:30:57 +00:00
|
|
|
/*
|
2022-11-16 10:26:45 +00:00
|
|
|
* Private mapping: create an exclusive anonymous page copy if reuse
|
|
|
|
* is impossible. We might miss VM_WRITE for FOLL_FORCE handling.
|
2006-09-26 06:30:57 +00:00
|
|
|
*/
|
2022-11-16 10:26:45 +00:00
|
|
|
if (folio && folio_test_anon(folio)) {
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
/*
|
|
|
|
* If the page is exclusive to this process we must reuse the
|
|
|
|
* page without further checks.
|
|
|
|
*/
|
2022-09-02 19:46:37 +00:00
|
|
|
if (PageAnonExclusive(vmf->page))
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
goto reuse;
|
|
|
|
|
mm: optimize do_wp_page() for exclusive pages in the swapcache
Patch series "mm: COW fixes part 1: fix the COW security issue for THP and swap", v3.
This series attempts to optimize and streamline the COW logic for ordinary
anon pages and THP anon pages, fixing two remaining instances of
CVE-2020-29374 in do_swap_page() and do_huge_pmd_wp_page(): information
can leak from a parent process to a child process via anonymous pages
shared during fork().
This issue, including other related COW issues, has been summarized in [2]:
"1. Observing Memory Modifications of Private Pages From A Child Process
Long story short: process-private memory might not be as private as you
think once you fork(): successive modifications of private memory
regions in the parent process can still be observed by the child
process, for example, by smart use of vmsplice()+munmap().
The core problem is that pinning pages readable in a child process, such
as done via the vmsplice system call, can result in a child process
observing memory modifications done in the parent process the child is
not supposed to observe. [1] contains an excellent summary and [2]
contains further details. This issue was assigned CVE-2020-29374 [9].
For this to trigger, it's required to use a fork() without subsequent
exec(), for example, as used under Android zygote. Without further
details about an application that forks less-privileged child processes,
one cannot really say what's actually affected and what's not -- see the
details section the end of this mail for a short sshd/openssh analysis.
While commit 17839856fd58 ("gup: document and work around "COW can break
either way" issue") fixed this issue and resulted in other problems
(e.g., ptrace on pmem), commit 09854ba94c6a ("mm: do_wp_page()
simplification") re-introduced part of the problem unfortunately.
The original reproducer can be modified quite easily to use THP [3] and
make the issue appear again on upstream kernels. I modified it to use
hugetlb [4] and it triggers as well. The problem is certainly less
severe with hugetlb than with THP; it merely highlights that we still
have plenty of open holes we should be closing/fixing.
Regarding vmsplice(), the only known workaround is to disallow the
vmsplice() system call ... or disable THP and hugetlb. But who knows
what else is affected (RDMA? O_DIRECT?) to achieve the same goal -- in
the end, it's a more generic issue"
This security issue was first reported by Jann Horn on 27 May 2020 and it
currently affects anonymous pages during swapin, anonymous THP and hugetlb.
This series tackles anonymous pages during swapin and anonymous THP:
- do_swap_page() for handling COW on PTEs during swapin directly
- do_huge_pmd_wp_page() for handling COW on PMD-mapped THP during write
faults
With this series, we'll apply the same COW logic we have in do_wp_page()
to all swappable anon pages: don't reuse (map writable) the page in
case there are additional references (page_count() != 1). All users of
reuse_swap_page() are remove, and consequently reuse_swap_page() is
removed.
In general, we're struggling with the following COW-related issues:
(1) "missed COW": we miss to copy on write and reuse the page (map it
writable) although we must copy because there are pending references
from another process to this page. The result is a security issue.
(2) "wrong COW": we copy on write although we wouldn't have to and
shouldn't: if there are valid GUP references, they will become out
of sync with the pages mapped into the page table. We fail to detect
that such a page can be reused safely, especially if never more than
a single process mapped the page. The result is an intra process
memory corruption.
(3) "unnecessary COW": we copy on write although we wouldn't have to:
performance degradation and temporary increases swap+memory
consumption can be the result.
While this series fixes (1) for swappable anon pages, it tries to reduce
reported cases of (3) first as good and easy as possible to limit the
impact when streamlining. The individual patches try to describe in
which cases we will run into (3).
This series certainly makes (2) worse for THP, because a THP will now
get PTE-mapped on write faults if there are additional references, even
if there was only ever a single process involved: once PTE-mapped, we'll
copy each and every subpage and won't reuse any subpage as long as the
underlying compound page wasn't split.
I'm working on an approach to fix (2) and improve (3): PageAnonExclusive
to mark anon pages that are exclusive to a single process, allow GUP
pins only on such exclusive pages, and allow turning exclusive pages
shared (clearing PageAnonExclusive) only if there are no GUP pins. Anon
pages with PageAnonExclusive set never have to be copied during write
faults, but eventually during fork() if they cannot be turned shared.
The improved reuse logic in this series will essentially also be the
logic to reset PageAnonExclusive. This work will certainly take a
while, but I'm planning on sharing details before having code fully
ready.
#1-#5 can be applied independently of the rest. #6-#9 are mostly only
cleanups related to reuse_swap_page().
Notes:
* For now, I'll leave hugetlb code untouched: "unnecessary COW" might
easily break existing setups because hugetlb pages are a scarce resource
and we could just end up having to crash the application when we run out
of hugetlb pages. We have to be very careful and the security aspect with
hugetlb is most certainly less relevant than for unprivileged anon pages.
* Instead of lru_add_drain() we might actually just drain the lru_add list
or even just remove the single page of interest from the lru_add list.
This would require a new helper function, and could be added if the
conditional lru_add_drain() turn out to be a problem.
* I extended the test case already included in [1] to also test for the
newly found do_swap_page() case. I'll send that out separately once/if
this part was merged.
[1] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[2] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 9):
Liang Zhang reported [1] that the current COW logic in do_wp_page() is
sub-optimal when it comes to swap+read fault+write fault of anonymous
pages that have a single user, visible via a performance degradation in
the redis benchmark. Something similar was previously reported [2] by
Nadav with a simple reproducer.
After we put an anon page into the swapcache and unmapped it from a single
process, that process might read that page again and refault it read-only.
If that process then writes to that page, the process is actually the
exclusive user of the page, however, the COW logic in do_co_page() won't
be able to reuse it due to the additional reference from the swapcache.
Let's optimize for pages that have been added to the swapcache but only
have an exclusive user. Try removing the swapcache reference if there is
hope that we're the exclusive user.
We will fail removing the swapcache reference in two scenarios:
(1) There are additional swap entries referencing the page: copying
instead of reusing is the right thing to do.
(2) The page is under writeback: theoretically we might be able to reuse
in some cases, however, we cannot remove the additional reference
and will have to copy.
Note that we'll only try removing the page from the swapcache when it's
highly likely that we'll be the exclusive owner after removing the page
from the swapache. As we're about to map that page writable and redirty
it, that should not affect reclaim but is rather the right thing to do.
Further, we might have additional references from the LRU pagevecs, which
will force us to copy instead of being able to reuse. We'll try handling
such references for some scenarios next. Concurrent writeback cannot be
handled easily and we'll always have to copy.
While at it, remove the superfluous page_mapcount() check: it's
implicitly covered by the page_count() for ordinary anon pages.
[1] https://lkml.kernel.org/r/20220113140318.11117-1-zhangliang5@huawei.com
[2] https://lkml.kernel.org/r/0480D692-D9B2-429A-9A88-9BBA1331AC3A@gmail.com
Link: https://lkml.kernel.org/r/20220131162940.210846-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reported-by: Liang Zhang <zhangliang5@huawei.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:30 +00:00
|
|
|
/*
|
2022-09-02 19:46:37 +00:00
|
|
|
* We have to verify under folio lock: these early checks are
|
|
|
|
* just an optimization to avoid locking the folio and freeing
|
mm: optimize do_wp_page() for exclusive pages in the swapcache
Patch series "mm: COW fixes part 1: fix the COW security issue for THP and swap", v3.
This series attempts to optimize and streamline the COW logic for ordinary
anon pages and THP anon pages, fixing two remaining instances of
CVE-2020-29374 in do_swap_page() and do_huge_pmd_wp_page(): information
can leak from a parent process to a child process via anonymous pages
shared during fork().
This issue, including other related COW issues, has been summarized in [2]:
"1. Observing Memory Modifications of Private Pages From A Child Process
Long story short: process-private memory might not be as private as you
think once you fork(): successive modifications of private memory
regions in the parent process can still be observed by the child
process, for example, by smart use of vmsplice()+munmap().
The core problem is that pinning pages readable in a child process, such
as done via the vmsplice system call, can result in a child process
observing memory modifications done in the parent process the child is
not supposed to observe. [1] contains an excellent summary and [2]
contains further details. This issue was assigned CVE-2020-29374 [9].
For this to trigger, it's required to use a fork() without subsequent
exec(), for example, as used under Android zygote. Without further
details about an application that forks less-privileged child processes,
one cannot really say what's actually affected and what's not -- see the
details section the end of this mail for a short sshd/openssh analysis.
While commit 17839856fd58 ("gup: document and work around "COW can break
either way" issue") fixed this issue and resulted in other problems
(e.g., ptrace on pmem), commit 09854ba94c6a ("mm: do_wp_page()
simplification") re-introduced part of the problem unfortunately.
The original reproducer can be modified quite easily to use THP [3] and
make the issue appear again on upstream kernels. I modified it to use
hugetlb [4] and it triggers as well. The problem is certainly less
severe with hugetlb than with THP; it merely highlights that we still
have plenty of open holes we should be closing/fixing.
Regarding vmsplice(), the only known workaround is to disallow the
vmsplice() system call ... or disable THP and hugetlb. But who knows
what else is affected (RDMA? O_DIRECT?) to achieve the same goal -- in
the end, it's a more generic issue"
This security issue was first reported by Jann Horn on 27 May 2020 and it
currently affects anonymous pages during swapin, anonymous THP and hugetlb.
This series tackles anonymous pages during swapin and anonymous THP:
- do_swap_page() for handling COW on PTEs during swapin directly
- do_huge_pmd_wp_page() for handling COW on PMD-mapped THP during write
faults
With this series, we'll apply the same COW logic we have in do_wp_page()
to all swappable anon pages: don't reuse (map writable) the page in
case there are additional references (page_count() != 1). All users of
reuse_swap_page() are remove, and consequently reuse_swap_page() is
removed.
In general, we're struggling with the following COW-related issues:
(1) "missed COW": we miss to copy on write and reuse the page (map it
writable) although we must copy because there are pending references
from another process to this page. The result is a security issue.
(2) "wrong COW": we copy on write although we wouldn't have to and
shouldn't: if there are valid GUP references, they will become out
of sync with the pages mapped into the page table. We fail to detect
that such a page can be reused safely, especially if never more than
a single process mapped the page. The result is an intra process
memory corruption.
(3) "unnecessary COW": we copy on write although we wouldn't have to:
performance degradation and temporary increases swap+memory
consumption can be the result.
While this series fixes (1) for swappable anon pages, it tries to reduce
reported cases of (3) first as good and easy as possible to limit the
impact when streamlining. The individual patches try to describe in
which cases we will run into (3).
This series certainly makes (2) worse for THP, because a THP will now
get PTE-mapped on write faults if there are additional references, even
if there was only ever a single process involved: once PTE-mapped, we'll
copy each and every subpage and won't reuse any subpage as long as the
underlying compound page wasn't split.
I'm working on an approach to fix (2) and improve (3): PageAnonExclusive
to mark anon pages that are exclusive to a single process, allow GUP
pins only on such exclusive pages, and allow turning exclusive pages
shared (clearing PageAnonExclusive) only if there are no GUP pins. Anon
pages with PageAnonExclusive set never have to be copied during write
faults, but eventually during fork() if they cannot be turned shared.
The improved reuse logic in this series will essentially also be the
logic to reset PageAnonExclusive. This work will certainly take a
while, but I'm planning on sharing details before having code fully
ready.
#1-#5 can be applied independently of the rest. #6-#9 are mostly only
cleanups related to reuse_swap_page().
Notes:
* For now, I'll leave hugetlb code untouched: "unnecessary COW" might
easily break existing setups because hugetlb pages are a scarce resource
and we could just end up having to crash the application when we run out
of hugetlb pages. We have to be very careful and the security aspect with
hugetlb is most certainly less relevant than for unprivileged anon pages.
* Instead of lru_add_drain() we might actually just drain the lru_add list
or even just remove the single page of interest from the lru_add list.
This would require a new helper function, and could be added if the
conditional lru_add_drain() turn out to be a problem.
* I extended the test case already included in [1] to also test for the
newly found do_swap_page() case. I'll send that out separately once/if
this part was merged.
[1] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[2] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 9):
Liang Zhang reported [1] that the current COW logic in do_wp_page() is
sub-optimal when it comes to swap+read fault+write fault of anonymous
pages that have a single user, visible via a performance degradation in
the redis benchmark. Something similar was previously reported [2] by
Nadav with a simple reproducer.
After we put an anon page into the swapcache and unmapped it from a single
process, that process might read that page again and refault it read-only.
If that process then writes to that page, the process is actually the
exclusive user of the page, however, the COW logic in do_co_page() won't
be able to reuse it due to the additional reference from the swapcache.
Let's optimize for pages that have been added to the swapcache but only
have an exclusive user. Try removing the swapcache reference if there is
hope that we're the exclusive user.
We will fail removing the swapcache reference in two scenarios:
(1) There are additional swap entries referencing the page: copying
instead of reusing is the right thing to do.
(2) The page is under writeback: theoretically we might be able to reuse
in some cases, however, we cannot remove the additional reference
and will have to copy.
Note that we'll only try removing the page from the swapcache when it's
highly likely that we'll be the exclusive owner after removing the page
from the swapache. As we're about to map that page writable and redirty
it, that should not affect reclaim but is rather the right thing to do.
Further, we might have additional references from the LRU pagevecs, which
will force us to copy instead of being able to reuse. We'll try handling
such references for some scenarios next. Concurrent writeback cannot be
handled easily and we'll always have to copy.
While at it, remove the superfluous page_mapcount() check: it's
implicitly covered by the page_count() for ordinary anon pages.
[1] https://lkml.kernel.org/r/20220113140318.11117-1-zhangliang5@huawei.com
[2] https://lkml.kernel.org/r/0480D692-D9B2-429A-9A88-9BBA1331AC3A@gmail.com
Link: https://lkml.kernel.org/r/20220131162940.210846-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reported-by: Liang Zhang <zhangliang5@huawei.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:30 +00:00
|
|
|
* the swapcache if there is little hope that we can reuse.
|
|
|
|
*
|
2022-09-02 19:46:37 +00:00
|
|
|
* KSM doesn't necessarily raise the folio refcount.
|
mm: optimize do_wp_page() for exclusive pages in the swapcache
Patch series "mm: COW fixes part 1: fix the COW security issue for THP and swap", v3.
This series attempts to optimize and streamline the COW logic for ordinary
anon pages and THP anon pages, fixing two remaining instances of
CVE-2020-29374 in do_swap_page() and do_huge_pmd_wp_page(): information
can leak from a parent process to a child process via anonymous pages
shared during fork().
This issue, including other related COW issues, has been summarized in [2]:
"1. Observing Memory Modifications of Private Pages From A Child Process
Long story short: process-private memory might not be as private as you
think once you fork(): successive modifications of private memory
regions in the parent process can still be observed by the child
process, for example, by smart use of vmsplice()+munmap().
The core problem is that pinning pages readable in a child process, such
as done via the vmsplice system call, can result in a child process
observing memory modifications done in the parent process the child is
not supposed to observe. [1] contains an excellent summary and [2]
contains further details. This issue was assigned CVE-2020-29374 [9].
For this to trigger, it's required to use a fork() without subsequent
exec(), for example, as used under Android zygote. Without further
details about an application that forks less-privileged child processes,
one cannot really say what's actually affected and what's not -- see the
details section the end of this mail for a short sshd/openssh analysis.
While commit 17839856fd58 ("gup: document and work around "COW can break
either way" issue") fixed this issue and resulted in other problems
(e.g., ptrace on pmem), commit 09854ba94c6a ("mm: do_wp_page()
simplification") re-introduced part of the problem unfortunately.
The original reproducer can be modified quite easily to use THP [3] and
make the issue appear again on upstream kernels. I modified it to use
hugetlb [4] and it triggers as well. The problem is certainly less
severe with hugetlb than with THP; it merely highlights that we still
have plenty of open holes we should be closing/fixing.
Regarding vmsplice(), the only known workaround is to disallow the
vmsplice() system call ... or disable THP and hugetlb. But who knows
what else is affected (RDMA? O_DIRECT?) to achieve the same goal -- in
the end, it's a more generic issue"
This security issue was first reported by Jann Horn on 27 May 2020 and it
currently affects anonymous pages during swapin, anonymous THP and hugetlb.
This series tackles anonymous pages during swapin and anonymous THP:
- do_swap_page() for handling COW on PTEs during swapin directly
- do_huge_pmd_wp_page() for handling COW on PMD-mapped THP during write
faults
With this series, we'll apply the same COW logic we have in do_wp_page()
to all swappable anon pages: don't reuse (map writable) the page in
case there are additional references (page_count() != 1). All users of
reuse_swap_page() are remove, and consequently reuse_swap_page() is
removed.
In general, we're struggling with the following COW-related issues:
(1) "missed COW": we miss to copy on write and reuse the page (map it
writable) although we must copy because there are pending references
from another process to this page. The result is a security issue.
(2) "wrong COW": we copy on write although we wouldn't have to and
shouldn't: if there are valid GUP references, they will become out
of sync with the pages mapped into the page table. We fail to detect
that such a page can be reused safely, especially if never more than
a single process mapped the page. The result is an intra process
memory corruption.
(3) "unnecessary COW": we copy on write although we wouldn't have to:
performance degradation and temporary increases swap+memory
consumption can be the result.
While this series fixes (1) for swappable anon pages, it tries to reduce
reported cases of (3) first as good and easy as possible to limit the
impact when streamlining. The individual patches try to describe in
which cases we will run into (3).
This series certainly makes (2) worse for THP, because a THP will now
get PTE-mapped on write faults if there are additional references, even
if there was only ever a single process involved: once PTE-mapped, we'll
copy each and every subpage and won't reuse any subpage as long as the
underlying compound page wasn't split.
I'm working on an approach to fix (2) and improve (3): PageAnonExclusive
to mark anon pages that are exclusive to a single process, allow GUP
pins only on such exclusive pages, and allow turning exclusive pages
shared (clearing PageAnonExclusive) only if there are no GUP pins. Anon
pages with PageAnonExclusive set never have to be copied during write
faults, but eventually during fork() if they cannot be turned shared.
The improved reuse logic in this series will essentially also be the
logic to reset PageAnonExclusive. This work will certainly take a
while, but I'm planning on sharing details before having code fully
ready.
#1-#5 can be applied independently of the rest. #6-#9 are mostly only
cleanups related to reuse_swap_page().
Notes:
* For now, I'll leave hugetlb code untouched: "unnecessary COW" might
easily break existing setups because hugetlb pages are a scarce resource
and we could just end up having to crash the application when we run out
of hugetlb pages. We have to be very careful and the security aspect with
hugetlb is most certainly less relevant than for unprivileged anon pages.
* Instead of lru_add_drain() we might actually just drain the lru_add list
or even just remove the single page of interest from the lru_add list.
This would require a new helper function, and could be added if the
conditional lru_add_drain() turn out to be a problem.
* I extended the test case already included in [1] to also test for the
newly found do_swap_page() case. I'll send that out separately once/if
this part was merged.
[1] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[2] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 9):
Liang Zhang reported [1] that the current COW logic in do_wp_page() is
sub-optimal when it comes to swap+read fault+write fault of anonymous
pages that have a single user, visible via a performance degradation in
the redis benchmark. Something similar was previously reported [2] by
Nadav with a simple reproducer.
After we put an anon page into the swapcache and unmapped it from a single
process, that process might read that page again and refault it read-only.
If that process then writes to that page, the process is actually the
exclusive user of the page, however, the COW logic in do_co_page() won't
be able to reuse it due to the additional reference from the swapcache.
Let's optimize for pages that have been added to the swapcache but only
have an exclusive user. Try removing the swapcache reference if there is
hope that we're the exclusive user.
We will fail removing the swapcache reference in two scenarios:
(1) There are additional swap entries referencing the page: copying
instead of reusing is the right thing to do.
(2) The page is under writeback: theoretically we might be able to reuse
in some cases, however, we cannot remove the additional reference
and will have to copy.
Note that we'll only try removing the page from the swapcache when it's
highly likely that we'll be the exclusive owner after removing the page
from the swapache. As we're about to map that page writable and redirty
it, that should not affect reclaim but is rather the right thing to do.
Further, we might have additional references from the LRU pagevecs, which
will force us to copy instead of being able to reuse. We'll try handling
such references for some scenarios next. Concurrent writeback cannot be
handled easily and we'll always have to copy.
While at it, remove the superfluous page_mapcount() check: it's
implicitly covered by the page_count() for ordinary anon pages.
[1] https://lkml.kernel.org/r/20220113140318.11117-1-zhangliang5@huawei.com
[2] https://lkml.kernel.org/r/0480D692-D9B2-429A-9A88-9BBA1331AC3A@gmail.com
Link: https://lkml.kernel.org/r/20220131162940.210846-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reported-by: Liang Zhang <zhangliang5@huawei.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:30 +00:00
|
|
|
*/
|
2022-09-02 19:46:37 +00:00
|
|
|
if (folio_test_ksm(folio) || folio_ref_count(folio) > 3)
|
mm: optimize do_wp_page() for fresh pages in local LRU pagevecs
For example, if a page just got swapped in via a read fault, the LRU
pagevecs might still hold a reference to the page. If we trigger a write
fault on such a page, the additional reference from the LRU pagevecs will
prohibit reusing the page.
Let's conditionally drain the local LRU pagevecs when we stumble over a
!PageLRU() page. We cannot easily drain remote LRU pagevecs and it might
not be desirable performance-wise. Consequently, this will only avoid
copying in some cases.
Add a simple "page_count(page) > 3" check first but keep the
"page_count(page) > 1 + PageSwapCache(page)" check in place, as we want to
minimize cases where we remove a page from the swapcache but won't be able
to reuse it, for example, because another process has it mapped R/O, to
not affect reclaim.
We cannot easily handle the following cases and we will always have to
copy:
(1) The page is referenced in the LRU pagevecs of other CPUs. We really
would have to drain the LRU pagevecs of all CPUs -- most probably
copying is much cheaper.
(2) The page is already PageLRU() but is getting moved between LRU
lists, for example, for activation (e.g., mark_page_accessed()),
deactivation (MADV_COLD), or lazyfree (MADV_FREE). We'd have to
drain mostly unconditionally, which might be bad performance-wise.
Most probably this won't happen too often in practice.
Note that there are other reasons why an anon page might temporarily not
be PageLRU(): for example, compaction and migration have to isolate LRU
pages from the LRU lists first (isolate_lru_page()), moving them to
temporary local lists and clearing PageLRU() and holding an additional
reference on the page. In that case, we'll always copy.
This change seems to be fairly effective with the reproducer [1] shared by
Nadav, as long as writeback is done synchronously, for example, using
zram. However, with asynchronous writeback, we'll usually fail to free
the swapcache because the page is still under writeback: something we
cannot easily optimize for, and maybe it's not really relevant in
practice.
[1] https://lkml.kernel.org/r/0480D692-D9B2-429A-9A88-9BBA1331AC3A@gmail.com
Link: https://lkml.kernel.org/r/20220131162940.210846-3-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:34 +00:00
|
|
|
goto copy;
|
2022-09-02 19:46:37 +00:00
|
|
|
if (!folio_test_lru(folio))
|
mm: optimize do_wp_page() for fresh pages in local LRU pagevecs
For example, if a page just got swapped in via a read fault, the LRU
pagevecs might still hold a reference to the page. If we trigger a write
fault on such a page, the additional reference from the LRU pagevecs will
prohibit reusing the page.
Let's conditionally drain the local LRU pagevecs when we stumble over a
!PageLRU() page. We cannot easily drain remote LRU pagevecs and it might
not be desirable performance-wise. Consequently, this will only avoid
copying in some cases.
Add a simple "page_count(page) > 3" check first but keep the
"page_count(page) > 1 + PageSwapCache(page)" check in place, as we want to
minimize cases where we remove a page from the swapcache but won't be able
to reuse it, for example, because another process has it mapped R/O, to
not affect reclaim.
We cannot easily handle the following cases and we will always have to
copy:
(1) The page is referenced in the LRU pagevecs of other CPUs. We really
would have to drain the LRU pagevecs of all CPUs -- most probably
copying is much cheaper.
(2) The page is already PageLRU() but is getting moved between LRU
lists, for example, for activation (e.g., mark_page_accessed()),
deactivation (MADV_COLD), or lazyfree (MADV_FREE). We'd have to
drain mostly unconditionally, which might be bad performance-wise.
Most probably this won't happen too often in practice.
Note that there are other reasons why an anon page might temporarily not
be PageLRU(): for example, compaction and migration have to isolate LRU
pages from the LRU lists first (isolate_lru_page()), moving them to
temporary local lists and clearing PageLRU() and holding an additional
reference on the page. In that case, we'll always copy.
This change seems to be fairly effective with the reproducer [1] shared by
Nadav, as long as writeback is done synchronously, for example, using
zram. However, with asynchronous writeback, we'll usually fail to free
the swapcache because the page is still under writeback: something we
cannot easily optimize for, and maybe it's not really relevant in
practice.
[1] https://lkml.kernel.org/r/0480D692-D9B2-429A-9A88-9BBA1331AC3A@gmail.com
Link: https://lkml.kernel.org/r/20220131162940.210846-3-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:34 +00:00
|
|
|
/*
|
|
|
|
* Note: We cannot easily detect+handle references from
|
2022-09-02 19:46:37 +00:00
|
|
|
* remote LRU pagevecs or references to LRU folios.
|
mm: optimize do_wp_page() for fresh pages in local LRU pagevecs
For example, if a page just got swapped in via a read fault, the LRU
pagevecs might still hold a reference to the page. If we trigger a write
fault on such a page, the additional reference from the LRU pagevecs will
prohibit reusing the page.
Let's conditionally drain the local LRU pagevecs when we stumble over a
!PageLRU() page. We cannot easily drain remote LRU pagevecs and it might
not be desirable performance-wise. Consequently, this will only avoid
copying in some cases.
Add a simple "page_count(page) > 3" check first but keep the
"page_count(page) > 1 + PageSwapCache(page)" check in place, as we want to
minimize cases where we remove a page from the swapcache but won't be able
to reuse it, for example, because another process has it mapped R/O, to
not affect reclaim.
We cannot easily handle the following cases and we will always have to
copy:
(1) The page is referenced in the LRU pagevecs of other CPUs. We really
would have to drain the LRU pagevecs of all CPUs -- most probably
copying is much cheaper.
(2) The page is already PageLRU() but is getting moved between LRU
lists, for example, for activation (e.g., mark_page_accessed()),
deactivation (MADV_COLD), or lazyfree (MADV_FREE). We'd have to
drain mostly unconditionally, which might be bad performance-wise.
Most probably this won't happen too often in practice.
Note that there are other reasons why an anon page might temporarily not
be PageLRU(): for example, compaction and migration have to isolate LRU
pages from the LRU lists first (isolate_lru_page()), moving them to
temporary local lists and clearing PageLRU() and holding an additional
reference on the page. In that case, we'll always copy.
This change seems to be fairly effective with the reproducer [1] shared by
Nadav, as long as writeback is done synchronously, for example, using
zram. However, with asynchronous writeback, we'll usually fail to free
the swapcache because the page is still under writeback: something we
cannot easily optimize for, and maybe it's not really relevant in
practice.
[1] https://lkml.kernel.org/r/0480D692-D9B2-429A-9A88-9BBA1331AC3A@gmail.com
Link: https://lkml.kernel.org/r/20220131162940.210846-3-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:34 +00:00
|
|
|
*/
|
|
|
|
lru_add_drain();
|
2022-09-02 19:46:37 +00:00
|
|
|
if (folio_ref_count(folio) > 1 + folio_test_swapcache(folio))
|
2020-08-21 23:49:55 +00:00
|
|
|
goto copy;
|
2022-09-02 19:46:37 +00:00
|
|
|
if (!folio_trylock(folio))
|
2020-08-21 23:49:55 +00:00
|
|
|
goto copy;
|
2022-09-02 19:46:37 +00:00
|
|
|
if (folio_test_swapcache(folio))
|
|
|
|
folio_free_swap(folio);
|
|
|
|
if (folio_test_ksm(folio) || folio_ref_count(folio) != 1) {
|
|
|
|
folio_unlock(folio);
|
mm: reuse only-pte-mapped KSM page in do_wp_page()
Add an optimization for KSM pages almost in the same way that we have
for ordinary anonymous pages. If there is a write fault in a page,
which is mapped to an only pte, and it is not related to swap cache; the
page may be reused without copying its content.
[ Note that we do not consider PageSwapCache() pages at least for now,
since we don't want to complicate __get_ksm_page(), which has nice
optimization based on this (for the migration case). Currenly it is
spinning on PageSwapCache() pages, waiting for when they have
unfreezed counters (i.e., for the migration finish). But we don't want
to make it also spinning on swap cache pages, which we try to reuse,
since there is not a very high probability to reuse them. So, for now
we do not consider PageSwapCache() pages at all. ]
So in reuse_ksm_page() we check for 1) PageSwapCache() and 2)
page_stable_node(), to skip a page, which KSM is currently trying to
link to stable tree. Then we do page_ref_freeze() to prohibit KSM to
merge one more page into the page, we are reusing. After that, nobody
can refer to the reusing page: KSM skips !PageSwapCache() pages with
zero refcount; and the protection against of all other participants is
the same as for reused ordinary anon pages pte lock, page lock and
mmap_sem.
[akpm@linux-foundation.org: replace BUG_ON()s with WARN_ON()s]
Link: http://lkml.kernel.org/r/154471491016.31352.1168978849911555609.stgit@localhost.localdomain
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Reviewed-by: Yang Shi <yang.shi@linux.alibaba.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Claudio Imbrenda <imbrenda@linux.vnet.ibm.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-03-05 23:43:06 +00:00
|
|
|
goto copy;
|
do_wp_page: remove the 'reuse' flag
mlocking a shared, writable vma currently causes the corresponding pages
to be marked as dirty and queued for writeback. This seems rather
unnecessary given that the pages are not being actually modified during
mlock. It is understood that for non-shared mappings (file or anon) we
want to use a write fault in order to break COW, but there is just no such
need for shared mappings.
The first two patches in this series do not introduce any behavior change.
The intent there is to make it obvious that dirtying file pages is only
done in the (writable, shared) case. I think this clarifies the code, but
I wouldn't mind dropping these two patches if there is no consensus about
them.
The last patch is where we actually avoid dirtying shared mappings during
mlock. Note that as a side effect of this, we won't call page_mkwrite()
for the mappings that define it, and won't be pre-allocating data blocks
at the FS level if the mapped file was sparsely allocated. My
understanding is that mlock does not need to provide such guarantee, as
evidenced by the fact that it never did for the filesystems that don't
define page_mkwrite() - including some common ones like ext3. However, I
would like to gather feedback on this from filesystem people as a
precaution. If this turns out to be a showstopper, maybe block
preallocation can be added back on using a different interface.
Large shared mlocks are getting significantly (>2x) faster in my tests, as
the disk can be fully used for reading the file instead of having to share
between this and writeback.
This patch:
Reorganize the code to remove the 'reuse' flag. No behavior changes.
Signed-off-by: Michel Lespinasse <walken@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Kosaki Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Theodore Tso <tytso@google.com>
Cc: Michael Rubin <mrubin@google.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:46:07 +00:00
|
|
|
}
|
2020-08-21 23:49:55 +00:00
|
|
|
/*
|
2022-09-02 19:46:37 +00:00
|
|
|
* Ok, we've got the only folio reference from our mapping
|
|
|
|
* and the folio is locked, it's dark out, and we're wearing
|
mm: optimize do_wp_page() for exclusive pages in the swapcache
Patch series "mm: COW fixes part 1: fix the COW security issue for THP and swap", v3.
This series attempts to optimize and streamline the COW logic for ordinary
anon pages and THP anon pages, fixing two remaining instances of
CVE-2020-29374 in do_swap_page() and do_huge_pmd_wp_page(): information
can leak from a parent process to a child process via anonymous pages
shared during fork().
This issue, including other related COW issues, has been summarized in [2]:
"1. Observing Memory Modifications of Private Pages From A Child Process
Long story short: process-private memory might not be as private as you
think once you fork(): successive modifications of private memory
regions in the parent process can still be observed by the child
process, for example, by smart use of vmsplice()+munmap().
The core problem is that pinning pages readable in a child process, such
as done via the vmsplice system call, can result in a child process
observing memory modifications done in the parent process the child is
not supposed to observe. [1] contains an excellent summary and [2]
contains further details. This issue was assigned CVE-2020-29374 [9].
For this to trigger, it's required to use a fork() without subsequent
exec(), for example, as used under Android zygote. Without further
details about an application that forks less-privileged child processes,
one cannot really say what's actually affected and what's not -- see the
details section the end of this mail for a short sshd/openssh analysis.
While commit 17839856fd58 ("gup: document and work around "COW can break
either way" issue") fixed this issue and resulted in other problems
(e.g., ptrace on pmem), commit 09854ba94c6a ("mm: do_wp_page()
simplification") re-introduced part of the problem unfortunately.
The original reproducer can be modified quite easily to use THP [3] and
make the issue appear again on upstream kernels. I modified it to use
hugetlb [4] and it triggers as well. The problem is certainly less
severe with hugetlb than with THP; it merely highlights that we still
have plenty of open holes we should be closing/fixing.
Regarding vmsplice(), the only known workaround is to disallow the
vmsplice() system call ... or disable THP and hugetlb. But who knows
what else is affected (RDMA? O_DIRECT?) to achieve the same goal -- in
the end, it's a more generic issue"
This security issue was first reported by Jann Horn on 27 May 2020 and it
currently affects anonymous pages during swapin, anonymous THP and hugetlb.
This series tackles anonymous pages during swapin and anonymous THP:
- do_swap_page() for handling COW on PTEs during swapin directly
- do_huge_pmd_wp_page() for handling COW on PMD-mapped THP during write
faults
With this series, we'll apply the same COW logic we have in do_wp_page()
to all swappable anon pages: don't reuse (map writable) the page in
case there are additional references (page_count() != 1). All users of
reuse_swap_page() are remove, and consequently reuse_swap_page() is
removed.
In general, we're struggling with the following COW-related issues:
(1) "missed COW": we miss to copy on write and reuse the page (map it
writable) although we must copy because there are pending references
from another process to this page. The result is a security issue.
(2) "wrong COW": we copy on write although we wouldn't have to and
shouldn't: if there are valid GUP references, they will become out
of sync with the pages mapped into the page table. We fail to detect
that such a page can be reused safely, especially if never more than
a single process mapped the page. The result is an intra process
memory corruption.
(3) "unnecessary COW": we copy on write although we wouldn't have to:
performance degradation and temporary increases swap+memory
consumption can be the result.
While this series fixes (1) for swappable anon pages, it tries to reduce
reported cases of (3) first as good and easy as possible to limit the
impact when streamlining. The individual patches try to describe in
which cases we will run into (3).
This series certainly makes (2) worse for THP, because a THP will now
get PTE-mapped on write faults if there are additional references, even
if there was only ever a single process involved: once PTE-mapped, we'll
copy each and every subpage and won't reuse any subpage as long as the
underlying compound page wasn't split.
I'm working on an approach to fix (2) and improve (3): PageAnonExclusive
to mark anon pages that are exclusive to a single process, allow GUP
pins only on such exclusive pages, and allow turning exclusive pages
shared (clearing PageAnonExclusive) only if there are no GUP pins. Anon
pages with PageAnonExclusive set never have to be copied during write
faults, but eventually during fork() if they cannot be turned shared.
The improved reuse logic in this series will essentially also be the
logic to reset PageAnonExclusive. This work will certainly take a
while, but I'm planning on sharing details before having code fully
ready.
#1-#5 can be applied independently of the rest. #6-#9 are mostly only
cleanups related to reuse_swap_page().
Notes:
* For now, I'll leave hugetlb code untouched: "unnecessary COW" might
easily break existing setups because hugetlb pages are a scarce resource
and we could just end up having to crash the application when we run out
of hugetlb pages. We have to be very careful and the security aspect with
hugetlb is most certainly less relevant than for unprivileged anon pages.
* Instead of lru_add_drain() we might actually just drain the lru_add list
or even just remove the single page of interest from the lru_add list.
This would require a new helper function, and could be added if the
conditional lru_add_drain() turn out to be a problem.
* I extended the test case already included in [1] to also test for the
newly found do_swap_page() case. I'll send that out separately once/if
this part was merged.
[1] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[2] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 9):
Liang Zhang reported [1] that the current COW logic in do_wp_page() is
sub-optimal when it comes to swap+read fault+write fault of anonymous
pages that have a single user, visible via a performance degradation in
the redis benchmark. Something similar was previously reported [2] by
Nadav with a simple reproducer.
After we put an anon page into the swapcache and unmapped it from a single
process, that process might read that page again and refault it read-only.
If that process then writes to that page, the process is actually the
exclusive user of the page, however, the COW logic in do_co_page() won't
be able to reuse it due to the additional reference from the swapcache.
Let's optimize for pages that have been added to the swapcache but only
have an exclusive user. Try removing the swapcache reference if there is
hope that we're the exclusive user.
We will fail removing the swapcache reference in two scenarios:
(1) There are additional swap entries referencing the page: copying
instead of reusing is the right thing to do.
(2) The page is under writeback: theoretically we might be able to reuse
in some cases, however, we cannot remove the additional reference
and will have to copy.
Note that we'll only try removing the page from the swapcache when it's
highly likely that we'll be the exclusive owner after removing the page
from the swapache. As we're about to map that page writable and redirty
it, that should not affect reclaim but is rather the right thing to do.
Further, we might have additional references from the LRU pagevecs, which
will force us to copy instead of being able to reuse. We'll try handling
such references for some scenarios next. Concurrent writeback cannot be
handled easily and we'll always have to copy.
While at it, remove the superfluous page_mapcount() check: it's
implicitly covered by the page_count() for ordinary anon pages.
[1] https://lkml.kernel.org/r/20220113140318.11117-1-zhangliang5@huawei.com
[2] https://lkml.kernel.org/r/0480D692-D9B2-429A-9A88-9BBA1331AC3A@gmail.com
Link: https://lkml.kernel.org/r/20220131162940.210846-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reported-by: Liang Zhang <zhangliang5@huawei.com>
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:30 +00:00
|
|
|
* sunglasses. Hit it.
|
2020-08-21 23:49:55 +00:00
|
|
|
*/
|
2022-09-02 19:46:37 +00:00
|
|
|
page_move_anon_rmap(vmf->page, vma);
|
|
|
|
folio_unlock(folio);
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
reuse:
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
if (unlikely(unshare)) {
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
|
|
|
return 0;
|
|
|
|
}
|
mm: fix misplaced unlock_page in do_wp_page()
Commit 09854ba94c6a ("mm: do_wp_page() simplification") reorganized all
the code around the page re-use vs copy, but in the process also moved
the final unlock_page() around to after the wp_page_reuse() call.
That normally doesn't matter - but it means that the unlock_page() is
now done after releasing the page table lock. Again, not a big deal,
you'd think.
But it turns out that it's very wrong indeed, because once we've
released the page table lock, we've basically lost our only reference to
the page - the page tables - and it could now be free'd at any time. We
do hold the mmap_sem, so no actual unmap() can happen, but madvise can
come in and a MADV_DONTNEED will zap the page range - and free the page.
So now the page may be free'd just as we're unlocking it, which in turn
will usually trigger a "Bad page state" error in the freeing path. To
make matters more confusing, by the time the debug code prints out the
page state, the unlock has typically completed and everything looks fine
again.
This all doesn't happen in any normal situations, but it does trigger
with the dirtyc0w_child LTP test. And it seems to trigger much more
easily (but not expclusively) on s390 than elsewhere, probably because
s390 doesn't do the "batch pages up for freeing after the TLB flush"
that gives the unlock_page() more time to complete and makes the race
harder to hit.
Fixes: 09854ba94c6a ("mm: do_wp_page() simplification")
Link: https://lore.kernel.org/lkml/a46e9bbef2ed4e17778f5615e818526ef848d791.camel@redhat.com/
Link: https://lore.kernel.org/linux-mm/c41149a8-211e-390b-af1d-d5eee690fecb@linux.alibaba.com/
Reported-by: Qian Cai <cai@redhat.com>
Reported-by: Alex Shi <alex.shi@linux.alibaba.com>
Bisected-and-analyzed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Tested-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-09-24 15:41:32 +00:00
|
|
|
wp_page_reuse(vmf);
|
2022-10-21 10:11:35 +00:00
|
|
|
return 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
mm: reuse only-pte-mapped KSM page in do_wp_page()
Add an optimization for KSM pages almost in the same way that we have
for ordinary anonymous pages. If there is a write fault in a page,
which is mapped to an only pte, and it is not related to swap cache; the
page may be reused without copying its content.
[ Note that we do not consider PageSwapCache() pages at least for now,
since we don't want to complicate __get_ksm_page(), which has nice
optimization based on this (for the migration case). Currenly it is
spinning on PageSwapCache() pages, waiting for when they have
unfreezed counters (i.e., for the migration finish). But we don't want
to make it also spinning on swap cache pages, which we try to reuse,
since there is not a very high probability to reuse them. So, for now
we do not consider PageSwapCache() pages at all. ]
So in reuse_ksm_page() we check for 1) PageSwapCache() and 2)
page_stable_node(), to skip a page, which KSM is currently trying to
link to stable tree. Then we do page_ref_freeze() to prohibit KSM to
merge one more page into the page, we are reusing. After that, nobody
can refer to the reusing page: KSM skips !PageSwapCache() pages with
zero refcount; and the protection against of all other participants is
the same as for reused ordinary anon pages pte lock, page lock and
mmap_sem.
[akpm@linux-foundation.org: replace BUG_ON()s with WARN_ON()s]
Link: http://lkml.kernel.org/r/154471491016.31352.1168978849911555609.stgit@localhost.localdomain
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Reviewed-by: Yang Shi <yang.shi@linux.alibaba.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Claudio Imbrenda <imbrenda@linux.vnet.ibm.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-03-05 23:43:06 +00:00
|
|
|
copy:
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Ok, we need to copy. Oh, well..
|
|
|
|
*/
|
2022-11-16 10:26:45 +00:00
|
|
|
if (folio)
|
|
|
|
folio_get(folio);
|
2015-04-14 22:46:29 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2022-04-29 06:16:16 +00:00
|
|
|
#ifdef CONFIG_KSM
|
2022-11-16 10:26:45 +00:00
|
|
|
if (folio && folio_test_ksm(folio))
|
2022-04-29 06:16:16 +00:00
|
|
|
count_vm_event(COW_KSM);
|
|
|
|
#endif
|
2016-12-14 23:07:33 +00:00
|
|
|
return wp_page_copy(vmf);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2011-05-25 00:12:04 +00:00
|
|
|
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long start_addr, unsigned long end_addr,
|
|
|
|
struct zap_details *details)
|
|
|
|
{
|
2012-03-05 19:14:20 +00:00
|
|
|
zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2017-09-08 23:15:08 +00:00
|
|
|
static inline void unmap_mapping_range_tree(struct rb_root_cached *root,
|
2021-11-05 20:38:31 +00:00
|
|
|
pgoff_t first_index,
|
|
|
|
pgoff_t last_index,
|
2005-04-16 22:20:36 +00:00
|
|
|
struct zap_details *details)
|
|
|
|
{
|
|
|
|
struct vm_area_struct *vma;
|
|
|
|
pgoff_t vba, vea, zba, zea;
|
|
|
|
|
2021-11-05 20:38:31 +00:00
|
|
|
vma_interval_tree_foreach(vma, root, first_index, last_index) {
|
2005-04-16 22:20:36 +00:00
|
|
|
vba = vma->vm_pgoff;
|
2013-07-03 22:01:26 +00:00
|
|
|
vea = vba + vma_pages(vma) - 1;
|
2022-03-22 21:42:33 +00:00
|
|
|
zba = max(first_index, vba);
|
|
|
|
zea = min(last_index, vea);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2011-05-25 00:12:04 +00:00
|
|
|
unmap_mapping_range_vma(vma,
|
2005-04-16 22:20:36 +00:00
|
|
|
((zba - vba) << PAGE_SHIFT) + vma->vm_start,
|
|
|
|
((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
|
2011-05-25 00:12:04 +00:00
|
|
|
details);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
/**
|
2021-11-28 19:53:35 +00:00
|
|
|
* unmap_mapping_folio() - Unmap single folio from processes.
|
|
|
|
* @folio: The locked folio to be unmapped.
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
*
|
2021-11-28 19:53:35 +00:00
|
|
|
* Unmap this folio from any userspace process which still has it mmaped.
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
* Typically, for efficiency, the range of nearby pages has already been
|
|
|
|
* unmapped by unmap_mapping_pages() or unmap_mapping_range(). But once
|
2021-11-28 19:53:35 +00:00
|
|
|
* truncation or invalidation holds the lock on a folio, it may find that
|
|
|
|
* the page has been remapped again: and then uses unmap_mapping_folio()
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
* to unmap it finally.
|
|
|
|
*/
|
2021-11-28 19:53:35 +00:00
|
|
|
void unmap_mapping_folio(struct folio *folio)
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
{
|
2021-11-28 19:53:35 +00:00
|
|
|
struct address_space *mapping = folio->mapping;
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
struct zap_details details = { };
|
2021-11-05 20:38:31 +00:00
|
|
|
pgoff_t first_index;
|
|
|
|
pgoff_t last_index;
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
|
2021-11-28 19:53:35 +00:00
|
|
|
VM_BUG_ON(!folio_test_locked(folio));
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
|
2021-11-28 19:53:35 +00:00
|
|
|
first_index = folio->index;
|
|
|
|
last_index = folio->index + folio_nr_pages(folio) - 1;
|
2021-11-05 20:38:31 +00:00
|
|
|
|
mm: change zap_details.zap_mapping into even_cows
Currently we have a zap_mapping pointer maintained in zap_details, when
it is specified we only want to zap the pages that has the same mapping
with what the caller has specified.
But what we want to do is actually simpler: we want to skip zapping
private (COW-ed) pages in some cases. We can refer to
unmap_mapping_pages() callers where we could have passed in different
even_cows values. The other user is unmap_mapping_folio() where we
always want to skip private pages.
According to Hugh, we used a mapping pointer for historical reason, as
explained here:
https://lore.kernel.org/lkml/391aa58d-ce84-9d4-d68d-d98a9c533255@google.com/
Quoting partly from Hugh:
Which raises the question again of why I did not just use a boolean flag
there originally: aah, I think I've found why. In those days there was a
horrible "optimization", for better performance on some benchmark I guess,
which when you read from /dev/zero into a private mapping, would map the zero
page there (look up read_zero_pagealigned() and zeromap_page_range() if you
dare). So there was another category of page to be skipped along with the
anon COWs, and I didn't want multiple tests in the zap loop, so checking
check_mapping against page->mapping did both. I think nowadays you could do
it by checking for PageAnon page (or genuine swap entry) instead.
This patch replaces the zap_details.zap_mapping pointer into the even_cows
boolean, then we check it against PageAnon.
Link: https://lkml.kernel.org/r/20220216094810.60572-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:21 +00:00
|
|
|
details.even_cows = false;
|
2021-11-28 19:53:35 +00:00
|
|
|
details.single_folio = folio;
|
2022-05-13 03:22:53 +00:00
|
|
|
details.zap_flags = ZAP_FLAG_DROP_MARKER;
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
|
2022-03-25 01:14:02 +00:00
|
|
|
i_mmap_lock_read(mapping);
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root)))
|
2021-11-05 20:38:31 +00:00
|
|
|
unmap_mapping_range_tree(&mapping->i_mmap, first_index,
|
|
|
|
last_index, &details);
|
2022-03-25 01:14:02 +00:00
|
|
|
i_mmap_unlock_read(mapping);
|
mm/thp: unmap_mapping_page() to fix THP truncate_cleanup_page()
There is a race between THP unmapping and truncation, when truncate sees
pmd_none() and skips the entry, after munmap's zap_huge_pmd() cleared
it, but before its page_remove_rmap() gets to decrement
compound_mapcount: generating false "BUG: Bad page cache" reports that
the page is still mapped when deleted. This commit fixes that, but not
in the way I hoped.
The first attempt used try_to_unmap(page, TTU_SYNC|TTU_IGNORE_MLOCK)
instead of unmap_mapping_range() in truncate_cleanup_page(): it has
often been an annoyance that we usually call unmap_mapping_range() with
no pages locked, but there apply it to a single locked page.
try_to_unmap() looks more suitable for a single locked page.
However, try_to_unmap_one() contains a VM_BUG_ON_PAGE(!pvmw.pte,page):
it is used to insert THP migration entries, but not used to unmap THPs.
Copy zap_huge_pmd() and add THP handling now? Perhaps, but their TLB
needs are different, I'm too ignorant of the DAX cases, and couldn't
decide how far to go for anon+swap. Set that aside.
The second attempt took a different tack: make no change in truncate.c,
but modify zap_huge_pmd() to insert an invalidated huge pmd instead of
clearing it initially, then pmd_clear() between page_remove_rmap() and
unlocking at the end. Nice. But powerpc blows that approach out of the
water, with its serialize_against_pte_lookup(), and interesting pgtable
usage. It would need serious help to get working on powerpc (with a
minor optimization issue on s390 too). Set that aside.
Just add an "if (page_mapped(page)) synchronize_rcu();" or other such
delay, after unmapping in truncate_cleanup_page()? Perhaps, but though
that's likely to reduce or eliminate the number of incidents, it would
give less assurance of whether we had identified the problem correctly.
This successful iteration introduces "unmap_mapping_page(page)" instead
of try_to_unmap(), and goes the usual unmap_mapping_range_tree() route,
with an addition to details. Then zap_pmd_range() watches for this
case, and does spin_unlock(pmd_lock) if so - just like
page_vma_mapped_walk() now does in the PVMW_SYNC case. Not pretty, but
safe.
Note that unmap_mapping_page() is doing a VM_BUG_ON(!PageLocked) to
assert its interface; but currently that's only used to make sure that
page->mapping is stable, and zap_pmd_range() doesn't care if the page is
locked or not. Along these lines, in invalidate_inode_pages2_range()
move the initial unmap_mapping_range() out from under page lock, before
then calling unmap_mapping_page() under page lock if still mapped.
Link: https://lkml.kernel.org/r/a2a4a148-cdd8-942c-4ef8-51b77f643dbe@google.com
Fixes: fc127da085c2 ("truncate: handle file thp")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jue Wang <juew@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wang Yugui <wangyugui@e16-tech.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-16 01:24:03 +00:00
|
|
|
}
|
|
|
|
|
2018-02-01 00:17:36 +00:00
|
|
|
/**
|
|
|
|
* unmap_mapping_pages() - Unmap pages from processes.
|
|
|
|
* @mapping: The address space containing pages to be unmapped.
|
|
|
|
* @start: Index of first page to be unmapped.
|
|
|
|
* @nr: Number of pages to be unmapped. 0 to unmap to end of file.
|
|
|
|
* @even_cows: Whether to unmap even private COWed pages.
|
|
|
|
*
|
|
|
|
* Unmap the pages in this address space from any userspace process which
|
|
|
|
* has them mmaped. Generally, you want to remove COWed pages as well when
|
|
|
|
* a file is being truncated, but not when invalidating pages from the page
|
|
|
|
* cache.
|
|
|
|
*/
|
|
|
|
void unmap_mapping_pages(struct address_space *mapping, pgoff_t start,
|
|
|
|
pgoff_t nr, bool even_cows)
|
|
|
|
{
|
|
|
|
struct zap_details details = { };
|
2021-11-05 20:38:31 +00:00
|
|
|
pgoff_t first_index = start;
|
|
|
|
pgoff_t last_index = start + nr - 1;
|
2018-02-01 00:17:36 +00:00
|
|
|
|
mm: change zap_details.zap_mapping into even_cows
Currently we have a zap_mapping pointer maintained in zap_details, when
it is specified we only want to zap the pages that has the same mapping
with what the caller has specified.
But what we want to do is actually simpler: we want to skip zapping
private (COW-ed) pages in some cases. We can refer to
unmap_mapping_pages() callers where we could have passed in different
even_cows values. The other user is unmap_mapping_folio() where we
always want to skip private pages.
According to Hugh, we used a mapping pointer for historical reason, as
explained here:
https://lore.kernel.org/lkml/391aa58d-ce84-9d4-d68d-d98a9c533255@google.com/
Quoting partly from Hugh:
Which raises the question again of why I did not just use a boolean flag
there originally: aah, I think I've found why. In those days there was a
horrible "optimization", for better performance on some benchmark I guess,
which when you read from /dev/zero into a private mapping, would map the zero
page there (look up read_zero_pagealigned() and zeromap_page_range() if you
dare). So there was another category of page to be skipped along with the
anon COWs, and I didn't want multiple tests in the zap loop, so checking
check_mapping against page->mapping did both. I think nowadays you could do
it by checking for PageAnon page (or genuine swap entry) instead.
This patch replaces the zap_details.zap_mapping pointer into the even_cows
boolean, then we check it against PageAnon.
Link: https://lkml.kernel.org/r/20220216094810.60572-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:42:21 +00:00
|
|
|
details.even_cows = even_cows;
|
2021-11-05 20:38:31 +00:00
|
|
|
if (last_index < first_index)
|
|
|
|
last_index = ULONG_MAX;
|
2018-02-01 00:17:36 +00:00
|
|
|
|
2022-03-25 01:14:02 +00:00
|
|
|
i_mmap_lock_read(mapping);
|
2018-02-01 00:17:36 +00:00
|
|
|
if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root)))
|
2021-11-05 20:38:31 +00:00
|
|
|
unmap_mapping_range_tree(&mapping->i_mmap, first_index,
|
|
|
|
last_index, &details);
|
2022-03-25 01:14:02 +00:00
|
|
|
i_mmap_unlock_read(mapping);
|
2018-02-01 00:17:36 +00:00
|
|
|
}
|
afs: Fix mmap coherency vs 3rd-party changes
Fix the coherency management of mmap'd data such that 3rd-party changes
become visible as soon as possible after the callback notification is
delivered by the fileserver. This is done by the following means:
(1) When we break a callback on a vnode specified by the CB.CallBack call
from the server, we queue a work item (vnode->cb_work) to go and
clobber all the PTEs mapping to that inode.
This causes the CPU to trip through the ->map_pages() and
->page_mkwrite() handlers if userspace attempts to access the page(s)
again.
(Ideally, this would be done in the service handler for CB.CallBack,
but the server is waiting for our reply before considering, and we
have a list of vnodes, all of which need breaking - and the process of
getting the mmap_lock and stripping the PTEs on all CPUs could be
quite slow.)
(2) Call afs_validate() from the ->map_pages() handler to check to see if
the file has changed and to get a new callback promise from the
server.
Also handle the fileserver telling us that it's dropping all callbacks,
possibly after it's been restarted by sending us a CB.InitCallBackState*
call by the following means:
(3) Maintain a per-cell list of afs files that are currently mmap'd
(cell->fs_open_mmaps).
(4) Add a work item to each server that is invoked if there are any open
mmaps when CB.InitCallBackState happens. This work item goes through
the aforementioned list and invokes the vnode->cb_work work item for
each one that is currently using this server.
This causes the PTEs to be cleared, causing ->map_pages() or
->page_mkwrite() to be called again, thereby calling afs_validate()
again.
I've chosen to simply strip the PTEs at the point of notification reception
rather than invalidate all the pages as well because (a) it's faster, (b)
we may get a notification for other reasons than the data being altered (in
which case we don't want to clobber the pagecache) and (c) we need to ask
the server to find out - and I don't want to wait for the reply before
holding up userspace.
This was tested using the attached test program:
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <sys/mman.h>
int main(int argc, char *argv[])
{
size_t size = getpagesize();
unsigned char *p;
bool mod = (argc == 3);
int fd;
if (argc != 2 && argc != 3) {
fprintf(stderr, "Format: %s <file> [mod]\n", argv[0]);
exit(2);
}
fd = open(argv[1], mod ? O_RDWR : O_RDONLY);
if (fd < 0) {
perror(argv[1]);
exit(1);
}
p = mmap(NULL, size, mod ? PROT_READ|PROT_WRITE : PROT_READ,
MAP_SHARED, fd, 0);
if (p == MAP_FAILED) {
perror("mmap");
exit(1);
}
for (;;) {
if (mod) {
p[0]++;
msync(p, size, MS_ASYNC);
fsync(fd);
}
printf("%02x", p[0]);
fflush(stdout);
sleep(1);
}
}
It runs in two modes: in one mode, it mmaps a file, then sits in a loop
reading the first byte, printing it and sleeping for a second; in the
second mode it mmaps a file, then sits in a loop incrementing the first
byte and flushing, then printing and sleeping.
Two instances of this program can be run on different machines, one doing
the reading and one doing the writing. The reader should see the changes
made by the writer, but without this patch, they aren't because validity
checking is being done lazily - only on entry to the filesystem.
Testing the InitCallBackState change is more complicated. The server has
to be taken offline, the saved callback state file removed and then the
server restarted whilst the reading-mode program continues to run. The
client machine then has to poke the server to trigger the InitCallBackState
call.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Markus Suvanto <markus.suvanto@gmail.com>
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/163111668833.283156.382633263709075739.stgit@warthog.procyon.org.uk/
2021-09-02 15:43:10 +00:00
|
|
|
EXPORT_SYMBOL_GPL(unmap_mapping_pages);
|
2018-02-01 00:17:36 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/**
|
2015-02-10 22:09:49 +00:00
|
|
|
* unmap_mapping_range - unmap the portion of all mmaps in the specified
|
2018-02-01 00:17:36 +00:00
|
|
|
* address_space corresponding to the specified byte range in the underlying
|
2015-02-10 22:09:49 +00:00
|
|
|
* file.
|
|
|
|
*
|
2005-06-24 05:05:21 +00:00
|
|
|
* @mapping: the address space containing mmaps to be unmapped.
|
2005-04-16 22:20:36 +00:00
|
|
|
* @holebegin: byte in first page to unmap, relative to the start of
|
|
|
|
* the underlying file. This will be rounded down to a PAGE_SIZE
|
2009-08-20 16:35:05 +00:00
|
|
|
* boundary. Note that this is different from truncate_pagecache(), which
|
2005-04-16 22:20:36 +00:00
|
|
|
* must keep the partial page. In contrast, we must get rid of
|
|
|
|
* partial pages.
|
|
|
|
* @holelen: size of prospective hole in bytes. This will be rounded
|
|
|
|
* up to a PAGE_SIZE boundary. A holelen of zero truncates to the
|
|
|
|
* end of the file.
|
|
|
|
* @even_cows: 1 when truncating a file, unmap even private COWed pages;
|
|
|
|
* but 0 when invalidating pagecache, don't throw away private data.
|
|
|
|
*/
|
|
|
|
void unmap_mapping_range(struct address_space *mapping,
|
|
|
|
loff_t const holebegin, loff_t const holelen, int even_cows)
|
|
|
|
{
|
|
|
|
pgoff_t hba = holebegin >> PAGE_SHIFT;
|
|
|
|
pgoff_t hlen = (holelen + PAGE_SIZE - 1) >> PAGE_SHIFT;
|
|
|
|
|
|
|
|
/* Check for overflow. */
|
|
|
|
if (sizeof(holelen) > sizeof(hlen)) {
|
|
|
|
long long holeend =
|
|
|
|
(holebegin + holelen + PAGE_SIZE - 1) >> PAGE_SHIFT;
|
|
|
|
if (holeend & ~(long long)ULONG_MAX)
|
|
|
|
hlen = ULONG_MAX - hba + 1;
|
|
|
|
}
|
|
|
|
|
2018-02-01 00:17:36 +00:00
|
|
|
unmap_mapping_pages(mapping, hba, hlen, even_cows);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(unmap_mapping_range);
|
|
|
|
|
2021-07-01 01:54:25 +00:00
|
|
|
/*
|
|
|
|
* Restore a potential device exclusive pte to a working pte entry
|
|
|
|
*/
|
|
|
|
static vm_fault_t remove_device_exclusive_entry(struct vm_fault *vmf)
|
|
|
|
{
|
2022-09-02 19:46:53 +00:00
|
|
|
struct folio *folio = page_folio(vmf->page);
|
2021-07-01 01:54:25 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
|
|
|
struct mmu_notifier_range range;
|
|
|
|
|
2023-03-30 01:25:19 +00:00
|
|
|
/*
|
|
|
|
* We need a reference to lock the folio because we don't hold
|
|
|
|
* the PTL so a racing thread can remove the device-exclusive
|
|
|
|
* entry and unmap it. If the folio is free the entry must
|
|
|
|
* have been removed already. If it happens to have already
|
|
|
|
* been re-allocated after being freed all we do is lock and
|
|
|
|
* unlock it.
|
|
|
|
*/
|
|
|
|
if (!folio_try_get(folio))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (!folio_lock_or_retry(folio, vma->vm_mm, vmf->flags)) {
|
|
|
|
folio_put(folio);
|
2021-07-01 01:54:25 +00:00
|
|
|
return VM_FAULT_RETRY;
|
2023-03-30 01:25:19 +00:00
|
|
|
}
|
2023-01-10 02:57:22 +00:00
|
|
|
mmu_notifier_range_init_owner(&range, MMU_NOTIFY_EXCLUSIVE, 0,
|
2021-07-01 01:54:25 +00:00
|
|
|
vma->vm_mm, vmf->address & PAGE_MASK,
|
|
|
|
(vmf->address & PAGE_MASK) + PAGE_SIZE, NULL);
|
|
|
|
mmu_notifier_invalidate_range_start(&range);
|
|
|
|
|
|
|
|
vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
|
|
|
|
&vmf->ptl);
|
|
|
|
if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
|
2022-09-02 19:46:53 +00:00
|
|
|
restore_exclusive_pte(vma, vmf->page, vmf->address, vmf->pte);
|
2021-07-01 01:54:25 +00:00
|
|
|
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2022-09-02 19:46:53 +00:00
|
|
|
folio_unlock(folio);
|
2023-03-30 01:25:19 +00:00
|
|
|
folio_put(folio);
|
2021-07-01 01:54:25 +00:00
|
|
|
|
|
|
|
mmu_notifier_invalidate_range_end(&range);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2022-09-02 19:46:42 +00:00
|
|
|
static inline bool should_try_to_free_swap(struct folio *folio,
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
struct vm_area_struct *vma,
|
|
|
|
unsigned int fault_flags)
|
|
|
|
{
|
2022-09-02 19:46:42 +00:00
|
|
|
if (!folio_test_swapcache(folio))
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
return false;
|
2022-09-02 19:46:43 +00:00
|
|
|
if (mem_cgroup_swap_full(folio) || (vma->vm_flags & VM_LOCKED) ||
|
2022-09-02 19:46:42 +00:00
|
|
|
folio_test_mlocked(folio))
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
return true;
|
|
|
|
/*
|
|
|
|
* If we want to map a page that's in the swapcache writable, we
|
|
|
|
* have to detect via the refcount if we're really the exclusive
|
|
|
|
* user. Try freeing the swapcache to get rid of the swapcache
|
|
|
|
* reference only in case it's likely that we'll be the exlusive user.
|
|
|
|
*/
|
2022-09-02 19:46:42 +00:00
|
|
|
return (fault_flags & FAULT_FLAG_WRITE) && !folio_test_ksm(folio) &&
|
|
|
|
folio_ref_count(folio) == 2;
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
}
|
|
|
|
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
static vm_fault_t pte_marker_clear(struct vm_fault *vmf)
|
|
|
|
{
|
|
|
|
vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd,
|
|
|
|
vmf->address, &vmf->ptl);
|
|
|
|
/*
|
|
|
|
* Be careful so that we will only recover a special uffd-wp pte into a
|
|
|
|
* none pte. Otherwise it means the pte could have changed, so retry.
|
2022-12-14 20:04:53 +00:00
|
|
|
*
|
|
|
|
* This should also cover the case where e.g. the pte changed
|
|
|
|
* quickly from a PTE_MARKER_UFFD_WP into PTE_MARKER_SWAPIN_ERROR.
|
|
|
|
* So is_pte_marker() check is not enough to safely drop the pte.
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
*/
|
2022-12-14 20:04:53 +00:00
|
|
|
if (pte_same(vmf->orig_pte, *vmf->pte))
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
pte_clear(vmf->vma->vm_mm, vmf->address, vmf->pte);
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
static vm_fault_t do_pte_missing(struct vm_fault *vmf)
|
|
|
|
{
|
|
|
|
if (vma_is_anonymous(vmf->vma))
|
|
|
|
return do_anonymous_page(vmf);
|
|
|
|
else
|
|
|
|
return do_fault(vmf);
|
|
|
|
}
|
|
|
|
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
/*
|
|
|
|
* This is actually a page-missing access, but with uffd-wp special pte
|
|
|
|
* installed. It means this pte was wr-protected before being unmapped.
|
|
|
|
*/
|
|
|
|
static vm_fault_t pte_marker_handle_uffd_wp(struct vm_fault *vmf)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Just in case there're leftover special ptes even after the region
|
2023-02-15 20:58:00 +00:00
|
|
|
* got unregistered - we can simply clear them.
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
*/
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
if (unlikely(!userfaultfd_wp(vmf->vma)))
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
return pte_marker_clear(vmf);
|
|
|
|
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
return do_pte_missing(vmf);
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
}
|
|
|
|
|
2022-05-13 03:22:52 +00:00
|
|
|
static vm_fault_t handle_pte_marker(struct vm_fault *vmf)
|
|
|
|
{
|
|
|
|
swp_entry_t entry = pte_to_swp_entry(vmf->orig_pte);
|
|
|
|
unsigned long marker = pte_marker_get(entry);
|
|
|
|
|
|
|
|
/*
|
2022-10-30 21:41:50 +00:00
|
|
|
* PTE markers should never be empty. If anything weird happened,
|
|
|
|
* the best thing to do is to kill the process along with its mm.
|
2022-05-13 03:22:52 +00:00
|
|
|
*/
|
2022-10-30 21:41:50 +00:00
|
|
|
if (WARN_ON_ONCE(!marker))
|
2022-05-13 03:22:52 +00:00
|
|
|
return VM_FAULT_SIGBUS;
|
|
|
|
|
2022-10-30 21:41:51 +00:00
|
|
|
/* Higher priority than uffd-wp when data corrupted */
|
|
|
|
if (marker & PTE_MARKER_SWAPIN_ERROR)
|
|
|
|
return VM_FAULT_SIGBUS;
|
|
|
|
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
if (pte_marker_entry_uffd_wp(entry))
|
|
|
|
return pte_marker_handle_uffd_wp(vmf);
|
|
|
|
|
|
|
|
/* This is an unknown pte marker */
|
|
|
|
return VM_FAULT_SIGBUS;
|
2022-05-13 03:22:52 +00:00
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
2020-06-09 04:33:54 +00:00
|
|
|
* We enter with non-exclusive mmap_lock (to exclude vma changes,
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
* but allow concurrent faults), and pte mapped but not yet locked.
|
2014-08-06 23:07:24 +00:00
|
|
|
* We return with pte unmapped and unlocked.
|
|
|
|
*
|
2020-06-09 04:33:54 +00:00
|
|
|
* We return with the mmap_lock locked or unlocked in the same cases
|
2014-08-06 23:07:24 +00:00
|
|
|
* as does filemap_fault().
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t do_swap_page(struct vm_fault *vmf)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2022-09-02 19:46:11 +00:00
|
|
|
struct folio *swapcache, *folio = NULL;
|
|
|
|
struct page *page;
|
2021-06-29 02:36:50 +00:00
|
|
|
struct swap_info_struct *si = NULL;
|
2022-05-10 01:20:43 +00:00
|
|
|
rmap_t rmap_flags = RMAP_NONE;
|
mm/swap: remember PG_anon_exclusive via a swp pte bit
Patch series "mm: COW fixes part 3: reliable GUP R/W FOLL_GET of anonymous pages", v2.
This series fixes memory corruptions when a GUP R/W reference (FOLL_WRITE
| FOLL_GET) was taken on an anonymous page and COW logic fails to detect
exclusivity of the page to then replacing the anonymous page by a copy in
the page table: The GUP reference lost synchronicity with the pages mapped
into the page tables. This series focuses on x86, arm64, s390x and
ppc64/book3s -- other architectures are fairly easy to support by
implementing __HAVE_ARCH_PTE_SWP_EXCLUSIVE.
This primarily fixes the O_DIRECT memory corruptions that can happen on
concurrent swapout, whereby we lose DMA reads to a page (modifying the
user page by writing to it).
O_DIRECT currently uses FOLL_GET for short-term (!FOLL_LONGTERM) DMA
from/to a user page. In the long run, we want to convert it to properly
use FOLL_PIN, and John is working on it, but that might take a while and
might not be easy to backport. In the meantime, let's restore what used
to work before we started modifying our COW logic: make R/W FOLL_GET
references reliable as long as there is no fork() after GUP involved.
This is just the natural follow-up of part 2, that will also further
reduce "wrong COW" on the swapin path, for example, when we cannot remove
a page from the swapcache due to concurrent writeback, or if we have two
threads faulting on the same swapped-out page. Fixing O_DIRECT is just a
nice side-product
This issue, including other related COW issues, has been summarized in [3]
under 2):
"
2. Intra Process Memory Corruptions due to Wrong COW (FOLL_GET)
It was discovered that we can create a memory corruption by reading a
file via O_DIRECT to a part (e.g., first 512 bytes) of a page,
concurrently writing to an unrelated part (e.g., last byte) of the same
page, and concurrently write-protecting the page via clear_refs
SOFTDIRTY tracking [6].
For the reproducer, the issue is that O_DIRECT grabs a reference of the
target page (via FOLL_GET) and clear_refs write-protects the relevant
page table entry. On successive write access to the page from the
process itself, we wrongly COW the page when resolving the write fault,
resulting in a loss of synchronicity and consequently a memory corruption.
While some people might think that using clear_refs in this combination
is a corner cases, it turns out to be a more generic problem unfortunately.
For example, it was just recently discovered that we can similarly
create a memory corruption without clear_refs, simply by concurrently
swapping out the buffer pages [7]. Note that we nowadays even use the
swap infrastructure in Linux without an actual swap disk/partition: the
prime example is zram which is enabled as default under Fedora [10].
The root issue is that a write-fault on a page that has additional
references results in a COW and thereby a loss of synchronicity
and consequently a memory corruption if two parties believe they are
referencing the same page.
"
We don't particularly care about R/O FOLL_GET references: they were never
reliable and O_DIRECT doesn't expect to observe modifications from a page
after DMA was started.
Note that:
* this only fixes the issue on x86, arm64, s390x and ppc64/book3s
("enterprise architectures"). Other architectures have to implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE to achieve the same.
* this does *not * consider any kind of fork() after taking the reference:
fork() after GUP never worked reliably with FOLL_GET.
* Not losing PG_anon_exclusive during swapout was the last remaining
piece. KSM already makes sure that there are no other references on
a page before considering it for sharing. Page migration maintains
PG_anon_exclusive and simply fails when there are additional references
(freezing the refcount fails). Only swapout code dropped the
PG_anon_exclusive flag because it requires more work to remember +
restore it.
With this series in place, most COW issues of [3] are fixed on said
architectures. Other architectures can implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE fairly easily.
[1] https://lkml.kernel.org/r/20220329160440.193848-1-david@redhat.com
[2] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[3] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 8):
Currently, we clear PG_anon_exclusive in try_to_unmap() and forget about
it. We do this, to keep fork() logic on swap entries easy and efficient:
for example, if we wouldn't clear it when unmapping, we'd have to lookup
the page in the swapcache for each and every swap entry during fork() and
clear PG_anon_exclusive if set.
Instead, we want to store that information directly in the swap pte,
protected by the page table lock, similarly to how we handle
SWP_MIGRATION_READ_EXCLUSIVE for migration entries. However, for actual
swap entries, we don't want to mess with the swap type (e.g., still one
bit) because it overcomplicates swap code.
In try_to_unmap(), we already reject to unmap in case the page might be
pinned, because we must not lose PG_anon_exclusive on pinned pages ever.
Checking if there are other unexpected references reliably *before*
completely unmapping a page is unfortunately not really possible: THP
heavily overcomplicate the situation. Once fully unmapped it's easier --
we, for example, make sure that there are no unexpected references *after*
unmapping a page before starting writeback on that page.
So, we currently might end up unmapping a page and clearing
PG_anon_exclusive if that page has additional references, for example, due
to a FOLL_GET.
do_swap_page() has to re-determine if a page is exclusive, which will
easily fail if there are other references on a page, most prominently GUP
references via FOLL_GET. This can currently result in memory corruptions
when taking a FOLL_GET | FOLL_WRITE reference on a page even when fork()
is never involved: try_to_unmap() will succeed, and when refaulting the
page, it cannot be marked exclusive and will get replaced by a copy in the
page tables on the next write access, resulting in writes via the GUP
reference to the page being lost.
In an ideal world, everybody that uses GUP and wants to modify page
content, such as O_DIRECT, would properly use FOLL_PIN. However, that
conversion will take a while. It's easier to fix what used to work in the
past (FOLL_GET | FOLL_WRITE) remembering PG_anon_exclusive. In addition,
by remembering PG_anon_exclusive we can further reduce unnecessary COW in
some cases, so it's the natural thing to do.
So let's transfer the PG_anon_exclusive information to the swap pte and
store it via an architecture-dependant pte bit; use that information when
restoring the swap pte in do_swap_page() and unuse_pte(). During fork(),
we simply have to clear the pte bit and are done.
Of course, there is one corner case to handle: swap backends that don't
support concurrent page modifications while the page is under writeback.
Special case these, and drop the exclusive marker. Add a comment why that
is just fine (also, reuse_swap_page() would have done the same in the
past).
In the future, we'll hopefully have all architectures support
__HAVE_ARCH_PTE_SWP_EXCLUSIVE, such that we can get rid of the empty stubs
and the define completely. Then, we can also convert
SWP_MIGRATION_READ_EXCLUSIVE. For architectures it's fairly easy to
support: either simply use a yet unused pte bit that can be used for swap
entries, steal one from the arch type bits if they exceed 5, or steal one
from the offset bits.
Note: R/O FOLL_GET references were never really reliable, especially when
taking one on a shared page and then writing to the page (e.g., GUP after
fork()). FOLL_GET, including R/W references, were never really reliable
once fork was involved (e.g., GUP before fork(), GUP during fork()). KSM
steps back in case it stumbles over unexpected references and is,
therefore, fine.
[david@redhat.com: fix SWP_STABLE_WRITES test]
Link: https://lkml.kernel.org/r/ac725bcb-313a-4fff-250a-68ba9a8f85fb@redhat.comLink: https://lkml.kernel.org/r/20220329164329.208407-1-david@redhat.com
Link: https://lkml.kernel.org/r/20220329164329.208407-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
bool exclusive = false;
|
[PATCH] mm: page fault handlers tidyup
Impose a little more consistency on the page fault handlers do_wp_page,
do_swap_page, do_anonymous_page, do_no_page, do_file_page: why not pass their
arguments in the same order, called the same names?
break_cow is all very well, but what it did was inlined elsewhere: easier to
compare if it's brought back into do_wp_page.
do_file_page's fallback to do_no_page dates from a time when we were testing
pte_file by using it wherever possible: currently it's peculiar to nonlinear
vmas, so just check that. BUG_ON if not? Better not, it's probably page
table corruption, so just show the pte: hmm, there's a pte_ERROR macro, let's
use that for do_wp_page's invalid pfn too.
Hah! Someone in the ppc64 world noticed pte_ERROR was unused so removed it:
restored (and say "pud" not "pmd" in its pud_ERROR).
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 01:15:59 +00:00
|
|
|
swp_entry_t entry;
|
2005-04-16 22:20:36 +00:00
|
|
|
pte_t pte;
|
2010-10-26 21:21:57 +00:00
|
|
|
int locked;
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret = 0;
|
2020-08-12 01:30:50 +00:00
|
|
|
void *shadow = NULL;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2021-11-05 20:38:28 +00:00
|
|
|
if (!pte_unmap_same(vmf))
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
goto out;
|
[PATCH] mm: page fault handlers tidyup
Impose a little more consistency on the page fault handlers do_wp_page,
do_swap_page, do_anonymous_page, do_no_page, do_file_page: why not pass their
arguments in the same order, called the same names?
break_cow is all very well, but what it did was inlined elsewhere: easier to
compare if it's brought back into do_wp_page.
do_file_page's fallback to do_no_page dates from a time when we were testing
pte_file by using it wherever possible: currently it's peculiar to nonlinear
vmas, so just check that. BUG_ON if not? Better not, it's probably page
table corruption, so just show the pte: hmm, there's a pte_ERROR macro, let's
use that for do_wp_page's invalid pfn too.
Hah! Someone in the ppc64 world noticed pte_ERROR was unused so removed it:
restored (and say "pud" not "pmd" in its pud_ERROR).
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 01:15:59 +00:00
|
|
|
|
2023-02-27 17:36:25 +00:00
|
|
|
if (vmf->flags & FAULT_FLAG_VMA_LOCK) {
|
|
|
|
ret = VM_FAULT_RETRY;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2016-12-14 23:07:16 +00:00
|
|
|
entry = pte_to_swp_entry(vmf->orig_pte);
|
2009-09-16 09:50:06 +00:00
|
|
|
if (unlikely(non_swap_entry(entry))) {
|
|
|
|
if (is_migration_entry(entry)) {
|
2016-12-14 23:06:58 +00:00
|
|
|
migration_entry_wait(vma->vm_mm, vmf->pmd,
|
|
|
|
vmf->address);
|
2021-07-01 01:54:25 +00:00
|
|
|
} else if (is_device_exclusive_entry(entry)) {
|
|
|
|
vmf->page = pfn_swap_entry_to_page(entry);
|
|
|
|
ret = remove_device_exclusive_entry(vmf);
|
2017-09-08 23:11:43 +00:00
|
|
|
} else if (is_device_private_entry(entry)) {
|
2021-07-01 01:54:06 +00:00
|
|
|
vmf->page = pfn_swap_entry_to_page(entry);
|
2022-09-28 12:01:15 +00:00
|
|
|
vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
|
|
|
|
vmf->address, &vmf->ptl);
|
|
|
|
if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
|
|
|
|
spin_unlock(vmf->ptl);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get a page reference while we know the page can't be
|
|
|
|
* freed.
|
|
|
|
*/
|
|
|
|
get_page(vmf->page);
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2022-11-14 11:55:37 +00:00
|
|
|
ret = vmf->page->pgmap->ops->migrate_to_ram(vmf);
|
2022-09-28 12:01:15 +00:00
|
|
|
put_page(vmf->page);
|
2009-09-16 09:50:06 +00:00
|
|
|
} else if (is_hwpoison_entry(entry)) {
|
|
|
|
ret = VM_FAULT_HWPOISON;
|
2022-05-13 03:22:52 +00:00
|
|
|
} else if (is_pte_marker_entry(entry)) {
|
|
|
|
ret = handle_pte_marker(vmf);
|
2009-09-16 09:50:06 +00:00
|
|
|
} else {
|
2016-12-14 23:07:16 +00:00
|
|
|
print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
|
2009-12-15 01:59:04 +00:00
|
|
|
ret = VM_FAULT_SIGBUS;
|
2009-09-16 09:50:06 +00:00
|
|
|
}
|
[PATCH] Swapless page migration: add R/W migration entries
Implement read/write migration ptes
We take the upper two swapfiles for the two types of migration ptes and define
a series of macros in swapops.h.
The VM is modified to handle the migration entries. migration entries can
only be encountered when the page they are pointing to is locked. This limits
the number of places one has to fix. We also check in copy_pte_range and in
mprotect_pte_range() for migration ptes.
We check for migration ptes in do_swap_cache and call a function that will
then wait on the page lock. This allows us to effectively stop all accesses
to apge.
Migration entries are created by try_to_unmap if called for migration and
removed by local functions in migrate.c
From: Hugh Dickins <hugh@veritas.com>
Several times while testing swapless page migration (I've no NUMA, just
hacking it up to migrate recklessly while running load), I've hit the
BUG_ON(!PageLocked(p)) in migration_entry_to_page.
This comes from an orphaned migration entry, unrelated to the current
correctly locked migration, but hit by remove_anon_migration_ptes as it
checks an address in each vma of the anon_vma list.
Such an orphan may be left behind if an earlier migration raced with fork:
copy_one_pte can duplicate a migration entry from parent to child, after
remove_anon_migration_ptes has checked the child vma, but before it has
removed it from the parent vma. (If the process were later to fault on this
orphaned entry, it would hit the same BUG from migration_entry_wait.)
This could be fixed by locking anon_vma in copy_one_pte, but we'd rather
not. There's no such problem with file pages, because vma_prio_tree_add
adds child vma after parent vma, and the page table locking at each end is
enough to serialize. Follow that example with anon_vma: add new vmas to the
tail instead of the head.
(There's no corresponding problem when inserting migration entries,
because a missed pte will leave the page count and mapcount high, which is
allowed for. And there's no corresponding problem when migrating via swap,
because a leftover swap entry will be correctly faulted. But the swapless
method has no refcounting of its entries.)
From: Ingo Molnar <mingo@elte.hu>
pte_unmap_unlock() takes the pte pointer as an argument.
From: Hugh Dickins <hugh@veritas.com>
Several times while testing swapless page migration, gcc has tried to exec
a pointer instead of a string: smells like COW mappings are not being
properly write-protected on fork.
The protection in copy_one_pte looks very convincing, until at last you
realize that the second arg to make_migration_entry is a boolean "write",
and SWP_MIGRATION_READ is 30.
Anyway, it's better done like in change_pte_range, using
is_write_migration_entry and make_migration_entry_read.
From: Hugh Dickins <hugh@veritas.com>
Remove unnecessary obfuscation from sys_swapon's range check on swap type,
which blew up causing memory corruption once swapless migration made
MAX_SWAPFILES no longer 2 ^ MAX_SWAPFILES_SHIFT.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Christoph Lameter <clameter@engr.sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
From: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-23 09:03:35 +00:00
|
|
|
goto out;
|
|
|
|
}
|
2017-11-16 01:33:07 +00:00
|
|
|
|
2021-06-29 02:36:50 +00:00
|
|
|
/* Prevent swapoff from happening to us. */
|
|
|
|
si = get_swap_device(entry);
|
|
|
|
if (unlikely(!si))
|
|
|
|
goto out;
|
2017-11-16 01:33:07 +00:00
|
|
|
|
2022-09-02 19:46:33 +00:00
|
|
|
folio = swap_cache_get_folio(entry, vma, vmf->address);
|
|
|
|
if (folio)
|
|
|
|
page = folio_file_page(folio, swp_offset(entry));
|
2022-09-02 19:46:11 +00:00
|
|
|
swapcache = folio;
|
2018-01-19 00:33:50 +00:00
|
|
|
|
2022-09-02 19:46:11 +00:00
|
|
|
if (!folio) {
|
2020-08-15 00:31:31 +00:00
|
|
|
if (data_race(si->flags & SWP_SYNCHRONOUS_IO) &&
|
|
|
|
__swap_count(entry) == 1) {
|
2017-11-16 01:33:07 +00:00
|
|
|
/* skip swapcache */
|
2022-09-02 19:46:10 +00:00
|
|
|
folio = vma_alloc_folio(GFP_HIGHUSER_MOVABLE, 0,
|
|
|
|
vma, vmf->address, false);
|
|
|
|
page = &folio->page;
|
|
|
|
if (folio) {
|
|
|
|
__folio_set_locked(folio);
|
|
|
|
__folio_set_swapbacked(folio);
|
2020-06-03 23:02:17 +00:00
|
|
|
|
2022-09-02 19:46:12 +00:00
|
|
|
if (mem_cgroup_swapin_charge_folio(folio,
|
2022-09-02 19:46:10 +00:00
|
|
|
vma->vm_mm, GFP_KERNEL,
|
|
|
|
entry)) {
|
2020-06-26 03:29:21 +00:00
|
|
|
ret = VM_FAULT_OOM;
|
2020-06-03 23:02:17 +00:00
|
|
|
goto out_page;
|
2020-06-26 03:29:21 +00:00
|
|
|
}
|
2021-04-30 05:56:36 +00:00
|
|
|
mem_cgroup_swapin_uncharge_swap(entry);
|
2020-06-03 23:02:17 +00:00
|
|
|
|
2020-08-12 01:30:50 +00:00
|
|
|
shadow = get_shadow_from_swap_cache(entry);
|
|
|
|
if (shadow)
|
2022-09-02 19:46:10 +00:00
|
|
|
workingset_refault(folio, shadow);
|
2020-06-26 03:30:37 +00:00
|
|
|
|
2022-09-02 19:46:10 +00:00
|
|
|
folio_add_lru(folio);
|
2021-04-30 05:56:36 +00:00
|
|
|
|
|
|
|
/* To provide entry to swap_readpage() */
|
2022-09-02 19:46:10 +00:00
|
|
|
folio_set_swap_entry(folio, entry);
|
2022-05-10 01:20:49 +00:00
|
|
|
swap_readpage(page, true, NULL);
|
2022-09-02 19:46:10 +00:00
|
|
|
folio->private = NULL;
|
2017-11-16 01:33:07 +00:00
|
|
|
}
|
2017-11-16 01:33:11 +00:00
|
|
|
} else {
|
2018-04-05 23:23:42 +00:00
|
|
|
page = swapin_readahead(entry, GFP_HIGHUSER_MOVABLE,
|
|
|
|
vmf);
|
2022-09-02 19:46:10 +00:00
|
|
|
if (page)
|
|
|
|
folio = page_folio(page);
|
2022-09-02 19:46:11 +00:00
|
|
|
swapcache = folio;
|
2017-11-16 01:33:07 +00:00
|
|
|
}
|
|
|
|
|
2022-09-02 19:46:11 +00:00
|
|
|
if (!folio) {
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
* Back out if somebody else faulted in this pte
|
|
|
|
* while we released the pte lock.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
|
|
|
|
vmf->address, &vmf->ptl);
|
2016-12-14 23:07:16 +00:00
|
|
|
if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
|
2005-04-16 22:20:36 +00:00
|
|
|
ret = VM_FAULT_OOM;
|
[PATCH] mm: page fault handlers tidyup
Impose a little more consistency on the page fault handlers do_wp_page,
do_swap_page, do_anonymous_page, do_no_page, do_file_page: why not pass their
arguments in the same order, called the same names?
break_cow is all very well, but what it did was inlined elsewhere: easier to
compare if it's brought back into do_wp_page.
do_file_page's fallback to do_no_page dates from a time when we were testing
pte_file by using it wherever possible: currently it's peculiar to nonlinear
vmas, so just check that. BUG_ON if not? Better not, it's probably page
table corruption, so just show the pte: hmm, there's a pte_ERROR macro, let's
use that for do_wp_page's invalid pfn too.
Hah! Someone in the ppc64 world noticed pte_ERROR was unused so removed it:
restored (and say "pud" not "pmd" in its pud_ERROR).
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 01:15:59 +00:00
|
|
|
goto unlock;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Had to read the page from swap area: Major fault */
|
|
|
|
ret = VM_FAULT_MAJOR;
|
2006-06-30 08:55:45 +00:00
|
|
|
count_vm_event(PGMAJFAULT);
|
2017-07-06 22:40:25 +00:00
|
|
|
count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
|
2009-09-16 09:50:06 +00:00
|
|
|
} else if (PageHWPoison(page)) {
|
2009-12-16 11:19:58 +00:00
|
|
|
/*
|
|
|
|
* hwpoisoned dirty swapcache pages are kept for killing
|
|
|
|
* owner processes (which may be unknown at hwpoison time)
|
|
|
|
*/
|
2009-09-16 09:50:06 +00:00
|
|
|
ret = VM_FAULT_HWPOISON;
|
2009-10-13 23:51:41 +00:00
|
|
|
goto out_release;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2022-09-02 19:46:53 +00:00
|
|
|
locked = folio_lock_or_retry(folio, vma->vm_mm, vmf->flags);
|
2012-05-29 22:06:18 +00:00
|
|
|
|
2010-10-26 21:21:57 +00:00
|
|
|
if (!locked) {
|
|
|
|
ret |= VM_FAULT_RETRY;
|
|
|
|
goto out_release;
|
|
|
|
}
|
2008-10-19 03:28:08 +00:00
|
|
|
|
2022-03-25 01:13:37 +00:00
|
|
|
if (swapcache) {
|
|
|
|
/*
|
2022-09-02 19:46:44 +00:00
|
|
|
* Make sure folio_free_swap() or swapoff did not release the
|
2022-03-25 01:13:37 +00:00
|
|
|
* swapcache from under us. The page pin, and pte_same test
|
|
|
|
* below, are not enough to exclude that. Even if it is still
|
|
|
|
* swapcache, we need to check that the page's swap has not
|
|
|
|
* changed.
|
|
|
|
*/
|
2022-09-02 19:46:10 +00:00
|
|
|
if (unlikely(!folio_test_swapcache(folio) ||
|
2022-03-25 01:13:37 +00:00
|
|
|
page_private(page) != entry.val))
|
|
|
|
goto out_page;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* KSM sometimes has to copy on read faults, for example, if
|
|
|
|
* page->index of !PageKSM() pages would be nonlinear inside the
|
|
|
|
* anon VMA -- PageKSM() is lost on actual swapout.
|
|
|
|
*/
|
|
|
|
page = ksm_might_need_to_copy(page, vma, vmf->address);
|
|
|
|
if (unlikely(!page)) {
|
|
|
|
ret = VM_FAULT_OOM;
|
|
|
|
goto out_page;
|
mm: hwpoison: support recovery from ksm_might_need_to_copy()
When the kernel copies a page from ksm_might_need_to_copy(), but runs into
an uncorrectable error, it will crash since poisoned page is consumed by
kernel, this is similar to the issue recently fixed by Copy-on-write
poison recovery.
When an error is detected during the page copy, return VM_FAULT_HWPOISON
in do_swap_page(), and install a hwpoison entry in unuse_pte() when
swapoff, which help us to avoid system crash. Note, memory failure on a
KSM page will be skipped, but still call memory_failure_queue() to be
consistent with general memory failure process, and we could support KSM
page recovery in the feature.
[wangkefeng.wang@huawei.com: enhance unuse_pte(), fix issue found by lkp]
Link: https://lkml.kernel.org/r/20221213120523.141588-1-wangkefeng.wang@huawei.com
[wangkefeng.wang@huawei.com: update changelog, alter ksm_might_need_to_copy(), restore unlikely() in unuse_pte()]
Link: https://lkml.kernel.org/r/20230201074433.96641-1-wangkefeng.wang@huawei.com
Link: https://lkml.kernel.org/r/20221209072801.193221-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Tony Luck <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-12-09 07:28:01 +00:00
|
|
|
} else if (unlikely(PTR_ERR(page) == -EHWPOISON)) {
|
|
|
|
ret = VM_FAULT_HWPOISON;
|
|
|
|
goto out_page;
|
2022-03-25 01:13:37 +00:00
|
|
|
}
|
2022-09-02 19:46:10 +00:00
|
|
|
folio = page_folio(page);
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If we want to map a page that's in the swapcache writable, we
|
|
|
|
* have to detect via the refcount if we're really the exclusive
|
|
|
|
* owner. Try removing the extra reference from the local LRU
|
|
|
|
* pagevecs if required.
|
|
|
|
*/
|
2022-09-02 19:46:11 +00:00
|
|
|
if ((vmf->flags & FAULT_FLAG_WRITE) && folio == swapcache &&
|
2022-09-02 19:46:10 +00:00
|
|
|
!folio_test_ksm(folio) && !folio_test_lru(folio))
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
lru_add_drain();
|
ksm: let shared pages be swappable
Initial implementation for swapping out KSM's shared pages: add
page_referenced_ksm() and try_to_unmap_ksm(), which rmap.c calls when
faced with a PageKsm page.
Most of what's needed can be got from the rmap_items listed from the
stable_node of the ksm page, without discovering the actual vma: so in
this patch just fake up a struct vma for page_referenced_one() or
try_to_unmap_one(), then refine that in the next patch.
Add VM_NONLINEAR to ksm_madvise()'s list of exclusions: it has always been
implicit there (being only set with VM_SHARED, already excluded), but
let's make it explicit, to help justify the lack of nonlinear unmap.
Rely on the page lock to protect against concurrent modifications to that
page's node of the stable tree.
The awkward part is not swapout but swapin: do_swap_page() and
page_add_anon_rmap() now have to allow for new possibilities - perhaps a
ksm page still in swapcache, perhaps a swapcache page associated with one
location in one anon_vma now needed for another location or anon_vma.
(And the vma might even be no longer VM_MERGEABLE when that happens.)
ksm_might_need_to_copy() checks for that case, and supplies a duplicate
page when necessary, simply leaving it to a subsequent pass of ksmd to
rediscover the identity and merge them back into one ksm page.
Disappointingly primitive: but the alternative would have to accumulate
unswappable info about the swapped out ksm pages, limiting swappability.
Remove page_add_ksm_rmap(): page_add_anon_rmap() now has to allow for the
particular case it was handling, so just use it instead.
Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Cc: Izik Eidus <ieidus@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Chris Wright <chrisw@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-12-15 01:59:24 +00:00
|
|
|
}
|
|
|
|
|
2023-03-02 11:58:30 +00:00
|
|
|
folio_throttle_swaprate(folio, GFP_KERNEL);
|
2008-02-07 08:13:53 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
* Back out if somebody else already faulted in this pte.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
|
|
|
|
&vmf->ptl);
|
2016-12-14 23:07:16 +00:00
|
|
|
if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte)))
|
2005-05-17 04:53:50 +00:00
|
|
|
goto out_nomap;
|
|
|
|
|
2022-09-02 19:46:10 +00:00
|
|
|
if (unlikely(!folio_test_uptodate(folio))) {
|
2005-05-17 04:53:50 +00:00
|
|
|
ret = VM_FAULT_SIGBUS;
|
|
|
|
goto out_nomap;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
mm/page-flags: reuse PG_mappedtodisk as PG_anon_exclusive for PageAnon() pages
The basic question we would like to have a reliable and efficient answer
to is: is this anonymous page exclusive to a single process or might it be
shared? We need that information for ordinary/single pages, hugetlb
pages, and possibly each subpage of a THP.
Introduce a way to mark an anonymous page as exclusive, with the ultimate
goal of teaching our COW logic to not do "wrong COWs", whereby GUP pins
lose consistency with the pages mapped into the page table, resulting in
reported memory corruptions.
Most pageflags already have semantics for anonymous pages, however,
PG_mappedtodisk should never apply to pages in the swapcache, so let's
reuse that flag.
As PG_has_hwpoisoned also uses that flag on the second tail page of a
compound page, convert it to PG_error instead, which is marked as
PF_NO_TAIL, so never used for tail pages.
Use custom page flag modification functions such that we can do additional
sanity checks. The semantics we'll put into some kernel doc in the future
are:
"
PG_anon_exclusive is *usually* only expressive in combination with a
page table entry. Depending on the page table entry type it might
store the following information:
Is what's mapped via this page table entry exclusive to the
single process and can be mapped writable without further
checks? If not, it might be shared and we might have to COW.
For now, we only expect PTE-mapped THPs to make use of
PG_anon_exclusive in subpages. For other anonymous compound
folios (i.e., hugetlb), only the head page is logically mapped and
holds this information.
For example, an exclusive, PMD-mapped THP only has PG_anon_exclusive
set on the head page. When replacing the PMD by a page table full
of PTEs, PG_anon_exclusive, if set on the head page, will be set on
all tail pages accordingly. Note that converting from a PTE-mapping
to a PMD mapping using the same compound page is currently not
possible and consequently doesn't require care.
If GUP wants to take a reliable pin (FOLL_PIN) on an anonymous page,
it should only pin if the relevant PG_anon_exclusive is set. In that
case, the pin will be fully reliable and stay consistent with the pages
mapped into the page table, as the bit cannot get cleared (e.g., by
fork(), KSM) while the page is pinned. For anonymous pages that
are mapped R/W, PG_anon_exclusive can be assumed to always be set
because such pages cannot possibly be shared.
The page table lock protecting the page table entry is the primary
synchronization mechanism for PG_anon_exclusive; GUP-fast that does
not take the PT lock needs special care when trying to clear the
flag.
Page table entry types and PG_anon_exclusive:
* Present: PG_anon_exclusive applies.
* Swap: the information is lost. PG_anon_exclusive was cleared.
* Migration: the entry holds this information instead.
PG_anon_exclusive was cleared.
* Device private: PG_anon_exclusive applies.
* Device exclusive: PG_anon_exclusive applies.
* HW Poison: PG_anon_exclusive is stale and not changed.
If the page may be pinned (FOLL_PIN), clearing PG_anon_exclusive is
not allowed and the flag will stick around until the page is freed
and folio->mapping is cleared.
"
We won't be clearing PG_anon_exclusive on destructive unmapping (i.e.,
zapping) of page table entries, page freeing code will handle that when
also invalidate page->mapping to not indicate PageAnon() anymore. Letting
information about exclusivity stick around will be an important property
when adding sanity checks to unpinning code.
Note that we properly clear the flag in free_pages_prepare() via
PAGE_FLAGS_CHECK_AT_PREP for each individual subpage of a compound page,
so there is no need to manually clear the flag.
Link: https://lkml.kernel.org/r/20220428083441.37290-12-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
/*
|
|
|
|
* PG_anon_exclusive reuses PG_mappedtodisk for anon pages. A swap pte
|
|
|
|
* must never point at an anonymous page in the swapcache that is
|
|
|
|
* PG_anon_exclusive. Sanity check that this holds and especially, that
|
|
|
|
* no filesystem set PG_mappedtodisk on a page in the swapcache. Sanity
|
|
|
|
* check after taking the PT lock and making sure that nobody
|
|
|
|
* concurrently faulted in this page and set PG_anon_exclusive.
|
|
|
|
*/
|
2022-09-02 19:46:10 +00:00
|
|
|
BUG_ON(!folio_test_anon(folio) && folio_test_mappedtodisk(folio));
|
|
|
|
BUG_ON(folio_test_anon(folio) && PageAnonExclusive(page));
|
mm/page-flags: reuse PG_mappedtodisk as PG_anon_exclusive for PageAnon() pages
The basic question we would like to have a reliable and efficient answer
to is: is this anonymous page exclusive to a single process or might it be
shared? We need that information for ordinary/single pages, hugetlb
pages, and possibly each subpage of a THP.
Introduce a way to mark an anonymous page as exclusive, with the ultimate
goal of teaching our COW logic to not do "wrong COWs", whereby GUP pins
lose consistency with the pages mapped into the page table, resulting in
reported memory corruptions.
Most pageflags already have semantics for anonymous pages, however,
PG_mappedtodisk should never apply to pages in the swapcache, so let's
reuse that flag.
As PG_has_hwpoisoned also uses that flag on the second tail page of a
compound page, convert it to PG_error instead, which is marked as
PF_NO_TAIL, so never used for tail pages.
Use custom page flag modification functions such that we can do additional
sanity checks. The semantics we'll put into some kernel doc in the future
are:
"
PG_anon_exclusive is *usually* only expressive in combination with a
page table entry. Depending on the page table entry type it might
store the following information:
Is what's mapped via this page table entry exclusive to the
single process and can be mapped writable without further
checks? If not, it might be shared and we might have to COW.
For now, we only expect PTE-mapped THPs to make use of
PG_anon_exclusive in subpages. For other anonymous compound
folios (i.e., hugetlb), only the head page is logically mapped and
holds this information.
For example, an exclusive, PMD-mapped THP only has PG_anon_exclusive
set on the head page. When replacing the PMD by a page table full
of PTEs, PG_anon_exclusive, if set on the head page, will be set on
all tail pages accordingly. Note that converting from a PTE-mapping
to a PMD mapping using the same compound page is currently not
possible and consequently doesn't require care.
If GUP wants to take a reliable pin (FOLL_PIN) on an anonymous page,
it should only pin if the relevant PG_anon_exclusive is set. In that
case, the pin will be fully reliable and stay consistent with the pages
mapped into the page table, as the bit cannot get cleared (e.g., by
fork(), KSM) while the page is pinned. For anonymous pages that
are mapped R/W, PG_anon_exclusive can be assumed to always be set
because such pages cannot possibly be shared.
The page table lock protecting the page table entry is the primary
synchronization mechanism for PG_anon_exclusive; GUP-fast that does
not take the PT lock needs special care when trying to clear the
flag.
Page table entry types and PG_anon_exclusive:
* Present: PG_anon_exclusive applies.
* Swap: the information is lost. PG_anon_exclusive was cleared.
* Migration: the entry holds this information instead.
PG_anon_exclusive was cleared.
* Device private: PG_anon_exclusive applies.
* Device exclusive: PG_anon_exclusive applies.
* HW Poison: PG_anon_exclusive is stale and not changed.
If the page may be pinned (FOLL_PIN), clearing PG_anon_exclusive is
not allowed and the flag will stick around until the page is freed
and folio->mapping is cleared.
"
We won't be clearing PG_anon_exclusive on destructive unmapping (i.e.,
zapping) of page table entries, page freeing code will handle that when
also invalidate page->mapping to not indicate PageAnon() anymore. Letting
information about exclusivity stick around will be an important property
when adding sanity checks to unpinning code.
Note that we properly clear the flag in free_pages_prepare() via
PAGE_FLAGS_CHECK_AT_PREP for each individual subpage of a compound page,
so there is no need to manually clear the flag.
Link: https://lkml.kernel.org/r/20220428083441.37290-12-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
|
mm/swap: remember PG_anon_exclusive via a swp pte bit
Patch series "mm: COW fixes part 3: reliable GUP R/W FOLL_GET of anonymous pages", v2.
This series fixes memory corruptions when a GUP R/W reference (FOLL_WRITE
| FOLL_GET) was taken on an anonymous page and COW logic fails to detect
exclusivity of the page to then replacing the anonymous page by a copy in
the page table: The GUP reference lost synchronicity with the pages mapped
into the page tables. This series focuses on x86, arm64, s390x and
ppc64/book3s -- other architectures are fairly easy to support by
implementing __HAVE_ARCH_PTE_SWP_EXCLUSIVE.
This primarily fixes the O_DIRECT memory corruptions that can happen on
concurrent swapout, whereby we lose DMA reads to a page (modifying the
user page by writing to it).
O_DIRECT currently uses FOLL_GET for short-term (!FOLL_LONGTERM) DMA
from/to a user page. In the long run, we want to convert it to properly
use FOLL_PIN, and John is working on it, but that might take a while and
might not be easy to backport. In the meantime, let's restore what used
to work before we started modifying our COW logic: make R/W FOLL_GET
references reliable as long as there is no fork() after GUP involved.
This is just the natural follow-up of part 2, that will also further
reduce "wrong COW" on the swapin path, for example, when we cannot remove
a page from the swapcache due to concurrent writeback, or if we have two
threads faulting on the same swapped-out page. Fixing O_DIRECT is just a
nice side-product
This issue, including other related COW issues, has been summarized in [3]
under 2):
"
2. Intra Process Memory Corruptions due to Wrong COW (FOLL_GET)
It was discovered that we can create a memory corruption by reading a
file via O_DIRECT to a part (e.g., first 512 bytes) of a page,
concurrently writing to an unrelated part (e.g., last byte) of the same
page, and concurrently write-protecting the page via clear_refs
SOFTDIRTY tracking [6].
For the reproducer, the issue is that O_DIRECT grabs a reference of the
target page (via FOLL_GET) and clear_refs write-protects the relevant
page table entry. On successive write access to the page from the
process itself, we wrongly COW the page when resolving the write fault,
resulting in a loss of synchronicity and consequently a memory corruption.
While some people might think that using clear_refs in this combination
is a corner cases, it turns out to be a more generic problem unfortunately.
For example, it was just recently discovered that we can similarly
create a memory corruption without clear_refs, simply by concurrently
swapping out the buffer pages [7]. Note that we nowadays even use the
swap infrastructure in Linux without an actual swap disk/partition: the
prime example is zram which is enabled as default under Fedora [10].
The root issue is that a write-fault on a page that has additional
references results in a COW and thereby a loss of synchronicity
and consequently a memory corruption if two parties believe they are
referencing the same page.
"
We don't particularly care about R/O FOLL_GET references: they were never
reliable and O_DIRECT doesn't expect to observe modifications from a page
after DMA was started.
Note that:
* this only fixes the issue on x86, arm64, s390x and ppc64/book3s
("enterprise architectures"). Other architectures have to implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE to achieve the same.
* this does *not * consider any kind of fork() after taking the reference:
fork() after GUP never worked reliably with FOLL_GET.
* Not losing PG_anon_exclusive during swapout was the last remaining
piece. KSM already makes sure that there are no other references on
a page before considering it for sharing. Page migration maintains
PG_anon_exclusive and simply fails when there are additional references
(freezing the refcount fails). Only swapout code dropped the
PG_anon_exclusive flag because it requires more work to remember +
restore it.
With this series in place, most COW issues of [3] are fixed on said
architectures. Other architectures can implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE fairly easily.
[1] https://lkml.kernel.org/r/20220329160440.193848-1-david@redhat.com
[2] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[3] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 8):
Currently, we clear PG_anon_exclusive in try_to_unmap() and forget about
it. We do this, to keep fork() logic on swap entries easy and efficient:
for example, if we wouldn't clear it when unmapping, we'd have to lookup
the page in the swapcache for each and every swap entry during fork() and
clear PG_anon_exclusive if set.
Instead, we want to store that information directly in the swap pte,
protected by the page table lock, similarly to how we handle
SWP_MIGRATION_READ_EXCLUSIVE for migration entries. However, for actual
swap entries, we don't want to mess with the swap type (e.g., still one
bit) because it overcomplicates swap code.
In try_to_unmap(), we already reject to unmap in case the page might be
pinned, because we must not lose PG_anon_exclusive on pinned pages ever.
Checking if there are other unexpected references reliably *before*
completely unmapping a page is unfortunately not really possible: THP
heavily overcomplicate the situation. Once fully unmapped it's easier --
we, for example, make sure that there are no unexpected references *after*
unmapping a page before starting writeback on that page.
So, we currently might end up unmapping a page and clearing
PG_anon_exclusive if that page has additional references, for example, due
to a FOLL_GET.
do_swap_page() has to re-determine if a page is exclusive, which will
easily fail if there are other references on a page, most prominently GUP
references via FOLL_GET. This can currently result in memory corruptions
when taking a FOLL_GET | FOLL_WRITE reference on a page even when fork()
is never involved: try_to_unmap() will succeed, and when refaulting the
page, it cannot be marked exclusive and will get replaced by a copy in the
page tables on the next write access, resulting in writes via the GUP
reference to the page being lost.
In an ideal world, everybody that uses GUP and wants to modify page
content, such as O_DIRECT, would properly use FOLL_PIN. However, that
conversion will take a while. It's easier to fix what used to work in the
past (FOLL_GET | FOLL_WRITE) remembering PG_anon_exclusive. In addition,
by remembering PG_anon_exclusive we can further reduce unnecessary COW in
some cases, so it's the natural thing to do.
So let's transfer the PG_anon_exclusive information to the swap pte and
store it via an architecture-dependant pte bit; use that information when
restoring the swap pte in do_swap_page() and unuse_pte(). During fork(),
we simply have to clear the pte bit and are done.
Of course, there is one corner case to handle: swap backends that don't
support concurrent page modifications while the page is under writeback.
Special case these, and drop the exclusive marker. Add a comment why that
is just fine (also, reuse_swap_page() would have done the same in the
past).
In the future, we'll hopefully have all architectures support
__HAVE_ARCH_PTE_SWP_EXCLUSIVE, such that we can get rid of the empty stubs
and the define completely. Then, we can also convert
SWP_MIGRATION_READ_EXCLUSIVE. For architectures it's fairly easy to
support: either simply use a yet unused pte bit that can be used for swap
entries, steal one from the arch type bits if they exceed 5, or steal one
from the offset bits.
Note: R/O FOLL_GET references were never really reliable, especially when
taking one on a shared page and then writing to the page (e.g., GUP after
fork()). FOLL_GET, including R/W references, were never really reliable
once fork was involved (e.g., GUP before fork(), GUP during fork()). KSM
steps back in case it stumbles over unexpected references and is,
therefore, fine.
[david@redhat.com: fix SWP_STABLE_WRITES test]
Link: https://lkml.kernel.org/r/ac725bcb-313a-4fff-250a-68ba9a8f85fb@redhat.comLink: https://lkml.kernel.org/r/20220329164329.208407-1-david@redhat.com
Link: https://lkml.kernel.org/r/20220329164329.208407-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
/*
|
|
|
|
* Check under PT lock (to protect against concurrent fork() sharing
|
|
|
|
* the swap entry concurrently) for certainly exclusive pages.
|
|
|
|
*/
|
2022-09-02 19:46:10 +00:00
|
|
|
if (!folio_test_ksm(folio)) {
|
mm/swap: remember PG_anon_exclusive via a swp pte bit
Patch series "mm: COW fixes part 3: reliable GUP R/W FOLL_GET of anonymous pages", v2.
This series fixes memory corruptions when a GUP R/W reference (FOLL_WRITE
| FOLL_GET) was taken on an anonymous page and COW logic fails to detect
exclusivity of the page to then replacing the anonymous page by a copy in
the page table: The GUP reference lost synchronicity with the pages mapped
into the page tables. This series focuses on x86, arm64, s390x and
ppc64/book3s -- other architectures are fairly easy to support by
implementing __HAVE_ARCH_PTE_SWP_EXCLUSIVE.
This primarily fixes the O_DIRECT memory corruptions that can happen on
concurrent swapout, whereby we lose DMA reads to a page (modifying the
user page by writing to it).
O_DIRECT currently uses FOLL_GET for short-term (!FOLL_LONGTERM) DMA
from/to a user page. In the long run, we want to convert it to properly
use FOLL_PIN, and John is working on it, but that might take a while and
might not be easy to backport. In the meantime, let's restore what used
to work before we started modifying our COW logic: make R/W FOLL_GET
references reliable as long as there is no fork() after GUP involved.
This is just the natural follow-up of part 2, that will also further
reduce "wrong COW" on the swapin path, for example, when we cannot remove
a page from the swapcache due to concurrent writeback, or if we have two
threads faulting on the same swapped-out page. Fixing O_DIRECT is just a
nice side-product
This issue, including other related COW issues, has been summarized in [3]
under 2):
"
2. Intra Process Memory Corruptions due to Wrong COW (FOLL_GET)
It was discovered that we can create a memory corruption by reading a
file via O_DIRECT to a part (e.g., first 512 bytes) of a page,
concurrently writing to an unrelated part (e.g., last byte) of the same
page, and concurrently write-protecting the page via clear_refs
SOFTDIRTY tracking [6].
For the reproducer, the issue is that O_DIRECT grabs a reference of the
target page (via FOLL_GET) and clear_refs write-protects the relevant
page table entry. On successive write access to the page from the
process itself, we wrongly COW the page when resolving the write fault,
resulting in a loss of synchronicity and consequently a memory corruption.
While some people might think that using clear_refs in this combination
is a corner cases, it turns out to be a more generic problem unfortunately.
For example, it was just recently discovered that we can similarly
create a memory corruption without clear_refs, simply by concurrently
swapping out the buffer pages [7]. Note that we nowadays even use the
swap infrastructure in Linux without an actual swap disk/partition: the
prime example is zram which is enabled as default under Fedora [10].
The root issue is that a write-fault on a page that has additional
references results in a COW and thereby a loss of synchronicity
and consequently a memory corruption if two parties believe they are
referencing the same page.
"
We don't particularly care about R/O FOLL_GET references: they were never
reliable and O_DIRECT doesn't expect to observe modifications from a page
after DMA was started.
Note that:
* this only fixes the issue on x86, arm64, s390x and ppc64/book3s
("enterprise architectures"). Other architectures have to implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE to achieve the same.
* this does *not * consider any kind of fork() after taking the reference:
fork() after GUP never worked reliably with FOLL_GET.
* Not losing PG_anon_exclusive during swapout was the last remaining
piece. KSM already makes sure that there are no other references on
a page before considering it for sharing. Page migration maintains
PG_anon_exclusive and simply fails when there are additional references
(freezing the refcount fails). Only swapout code dropped the
PG_anon_exclusive flag because it requires more work to remember +
restore it.
With this series in place, most COW issues of [3] are fixed on said
architectures. Other architectures can implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE fairly easily.
[1] https://lkml.kernel.org/r/20220329160440.193848-1-david@redhat.com
[2] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[3] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 8):
Currently, we clear PG_anon_exclusive in try_to_unmap() and forget about
it. We do this, to keep fork() logic on swap entries easy and efficient:
for example, if we wouldn't clear it when unmapping, we'd have to lookup
the page in the swapcache for each and every swap entry during fork() and
clear PG_anon_exclusive if set.
Instead, we want to store that information directly in the swap pte,
protected by the page table lock, similarly to how we handle
SWP_MIGRATION_READ_EXCLUSIVE for migration entries. However, for actual
swap entries, we don't want to mess with the swap type (e.g., still one
bit) because it overcomplicates swap code.
In try_to_unmap(), we already reject to unmap in case the page might be
pinned, because we must not lose PG_anon_exclusive on pinned pages ever.
Checking if there are other unexpected references reliably *before*
completely unmapping a page is unfortunately not really possible: THP
heavily overcomplicate the situation. Once fully unmapped it's easier --
we, for example, make sure that there are no unexpected references *after*
unmapping a page before starting writeback on that page.
So, we currently might end up unmapping a page and clearing
PG_anon_exclusive if that page has additional references, for example, due
to a FOLL_GET.
do_swap_page() has to re-determine if a page is exclusive, which will
easily fail if there are other references on a page, most prominently GUP
references via FOLL_GET. This can currently result in memory corruptions
when taking a FOLL_GET | FOLL_WRITE reference on a page even when fork()
is never involved: try_to_unmap() will succeed, and when refaulting the
page, it cannot be marked exclusive and will get replaced by a copy in the
page tables on the next write access, resulting in writes via the GUP
reference to the page being lost.
In an ideal world, everybody that uses GUP and wants to modify page
content, such as O_DIRECT, would properly use FOLL_PIN. However, that
conversion will take a while. It's easier to fix what used to work in the
past (FOLL_GET | FOLL_WRITE) remembering PG_anon_exclusive. In addition,
by remembering PG_anon_exclusive we can further reduce unnecessary COW in
some cases, so it's the natural thing to do.
So let's transfer the PG_anon_exclusive information to the swap pte and
store it via an architecture-dependant pte bit; use that information when
restoring the swap pte in do_swap_page() and unuse_pte(). During fork(),
we simply have to clear the pte bit and are done.
Of course, there is one corner case to handle: swap backends that don't
support concurrent page modifications while the page is under writeback.
Special case these, and drop the exclusive marker. Add a comment why that
is just fine (also, reuse_swap_page() would have done the same in the
past).
In the future, we'll hopefully have all architectures support
__HAVE_ARCH_PTE_SWP_EXCLUSIVE, such that we can get rid of the empty stubs
and the define completely. Then, we can also convert
SWP_MIGRATION_READ_EXCLUSIVE. For architectures it's fairly easy to
support: either simply use a yet unused pte bit that can be used for swap
entries, steal one from the arch type bits if they exceed 5, or steal one
from the offset bits.
Note: R/O FOLL_GET references were never really reliable, especially when
taking one on a shared page and then writing to the page (e.g., GUP after
fork()). FOLL_GET, including R/W references, were never really reliable
once fork was involved (e.g., GUP before fork(), GUP during fork()). KSM
steps back in case it stumbles over unexpected references and is,
therefore, fine.
[david@redhat.com: fix SWP_STABLE_WRITES test]
Link: https://lkml.kernel.org/r/ac725bcb-313a-4fff-250a-68ba9a8f85fb@redhat.comLink: https://lkml.kernel.org/r/20220329164329.208407-1-david@redhat.com
Link: https://lkml.kernel.org/r/20220329164329.208407-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
exclusive = pte_swp_exclusive(vmf->orig_pte);
|
2022-09-02 19:46:11 +00:00
|
|
|
if (folio != swapcache) {
|
mm/swap: remember PG_anon_exclusive via a swp pte bit
Patch series "mm: COW fixes part 3: reliable GUP R/W FOLL_GET of anonymous pages", v2.
This series fixes memory corruptions when a GUP R/W reference (FOLL_WRITE
| FOLL_GET) was taken on an anonymous page and COW logic fails to detect
exclusivity of the page to then replacing the anonymous page by a copy in
the page table: The GUP reference lost synchronicity with the pages mapped
into the page tables. This series focuses on x86, arm64, s390x and
ppc64/book3s -- other architectures are fairly easy to support by
implementing __HAVE_ARCH_PTE_SWP_EXCLUSIVE.
This primarily fixes the O_DIRECT memory corruptions that can happen on
concurrent swapout, whereby we lose DMA reads to a page (modifying the
user page by writing to it).
O_DIRECT currently uses FOLL_GET for short-term (!FOLL_LONGTERM) DMA
from/to a user page. In the long run, we want to convert it to properly
use FOLL_PIN, and John is working on it, but that might take a while and
might not be easy to backport. In the meantime, let's restore what used
to work before we started modifying our COW logic: make R/W FOLL_GET
references reliable as long as there is no fork() after GUP involved.
This is just the natural follow-up of part 2, that will also further
reduce "wrong COW" on the swapin path, for example, when we cannot remove
a page from the swapcache due to concurrent writeback, or if we have two
threads faulting on the same swapped-out page. Fixing O_DIRECT is just a
nice side-product
This issue, including other related COW issues, has been summarized in [3]
under 2):
"
2. Intra Process Memory Corruptions due to Wrong COW (FOLL_GET)
It was discovered that we can create a memory corruption by reading a
file via O_DIRECT to a part (e.g., first 512 bytes) of a page,
concurrently writing to an unrelated part (e.g., last byte) of the same
page, and concurrently write-protecting the page via clear_refs
SOFTDIRTY tracking [6].
For the reproducer, the issue is that O_DIRECT grabs a reference of the
target page (via FOLL_GET) and clear_refs write-protects the relevant
page table entry. On successive write access to the page from the
process itself, we wrongly COW the page when resolving the write fault,
resulting in a loss of synchronicity and consequently a memory corruption.
While some people might think that using clear_refs in this combination
is a corner cases, it turns out to be a more generic problem unfortunately.
For example, it was just recently discovered that we can similarly
create a memory corruption without clear_refs, simply by concurrently
swapping out the buffer pages [7]. Note that we nowadays even use the
swap infrastructure in Linux without an actual swap disk/partition: the
prime example is zram which is enabled as default under Fedora [10].
The root issue is that a write-fault on a page that has additional
references results in a COW and thereby a loss of synchronicity
and consequently a memory corruption if two parties believe they are
referencing the same page.
"
We don't particularly care about R/O FOLL_GET references: they were never
reliable and O_DIRECT doesn't expect to observe modifications from a page
after DMA was started.
Note that:
* this only fixes the issue on x86, arm64, s390x and ppc64/book3s
("enterprise architectures"). Other architectures have to implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE to achieve the same.
* this does *not * consider any kind of fork() after taking the reference:
fork() after GUP never worked reliably with FOLL_GET.
* Not losing PG_anon_exclusive during swapout was the last remaining
piece. KSM already makes sure that there are no other references on
a page before considering it for sharing. Page migration maintains
PG_anon_exclusive and simply fails when there are additional references
(freezing the refcount fails). Only swapout code dropped the
PG_anon_exclusive flag because it requires more work to remember +
restore it.
With this series in place, most COW issues of [3] are fixed on said
architectures. Other architectures can implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE fairly easily.
[1] https://lkml.kernel.org/r/20220329160440.193848-1-david@redhat.com
[2] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[3] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 8):
Currently, we clear PG_anon_exclusive in try_to_unmap() and forget about
it. We do this, to keep fork() logic on swap entries easy and efficient:
for example, if we wouldn't clear it when unmapping, we'd have to lookup
the page in the swapcache for each and every swap entry during fork() and
clear PG_anon_exclusive if set.
Instead, we want to store that information directly in the swap pte,
protected by the page table lock, similarly to how we handle
SWP_MIGRATION_READ_EXCLUSIVE for migration entries. However, for actual
swap entries, we don't want to mess with the swap type (e.g., still one
bit) because it overcomplicates swap code.
In try_to_unmap(), we already reject to unmap in case the page might be
pinned, because we must not lose PG_anon_exclusive on pinned pages ever.
Checking if there are other unexpected references reliably *before*
completely unmapping a page is unfortunately not really possible: THP
heavily overcomplicate the situation. Once fully unmapped it's easier --
we, for example, make sure that there are no unexpected references *after*
unmapping a page before starting writeback on that page.
So, we currently might end up unmapping a page and clearing
PG_anon_exclusive if that page has additional references, for example, due
to a FOLL_GET.
do_swap_page() has to re-determine if a page is exclusive, which will
easily fail if there are other references on a page, most prominently GUP
references via FOLL_GET. This can currently result in memory corruptions
when taking a FOLL_GET | FOLL_WRITE reference on a page even when fork()
is never involved: try_to_unmap() will succeed, and when refaulting the
page, it cannot be marked exclusive and will get replaced by a copy in the
page tables on the next write access, resulting in writes via the GUP
reference to the page being lost.
In an ideal world, everybody that uses GUP and wants to modify page
content, such as O_DIRECT, would properly use FOLL_PIN. However, that
conversion will take a while. It's easier to fix what used to work in the
past (FOLL_GET | FOLL_WRITE) remembering PG_anon_exclusive. In addition,
by remembering PG_anon_exclusive we can further reduce unnecessary COW in
some cases, so it's the natural thing to do.
So let's transfer the PG_anon_exclusive information to the swap pte and
store it via an architecture-dependant pte bit; use that information when
restoring the swap pte in do_swap_page() and unuse_pte(). During fork(),
we simply have to clear the pte bit and are done.
Of course, there is one corner case to handle: swap backends that don't
support concurrent page modifications while the page is under writeback.
Special case these, and drop the exclusive marker. Add a comment why that
is just fine (also, reuse_swap_page() would have done the same in the
past).
In the future, we'll hopefully have all architectures support
__HAVE_ARCH_PTE_SWP_EXCLUSIVE, such that we can get rid of the empty stubs
and the define completely. Then, we can also convert
SWP_MIGRATION_READ_EXCLUSIVE. For architectures it's fairly easy to
support: either simply use a yet unused pte bit that can be used for swap
entries, steal one from the arch type bits if they exceed 5, or steal one
from the offset bits.
Note: R/O FOLL_GET references were never really reliable, especially when
taking one on a shared page and then writing to the page (e.g., GUP after
fork()). FOLL_GET, including R/W references, were never really reliable
once fork was involved (e.g., GUP before fork(), GUP during fork()). KSM
steps back in case it stumbles over unexpected references and is,
therefore, fine.
[david@redhat.com: fix SWP_STABLE_WRITES test]
Link: https://lkml.kernel.org/r/ac725bcb-313a-4fff-250a-68ba9a8f85fb@redhat.comLink: https://lkml.kernel.org/r/20220329164329.208407-1-david@redhat.com
Link: https://lkml.kernel.org/r/20220329164329.208407-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
/*
|
|
|
|
* We have a fresh page that is not exposed to the
|
|
|
|
* swapcache -> certainly exclusive.
|
|
|
|
*/
|
|
|
|
exclusive = true;
|
2022-09-02 19:46:10 +00:00
|
|
|
} else if (exclusive && folio_test_writeback(folio) &&
|
2022-05-19 21:08:51 +00:00
|
|
|
data_race(si->flags & SWP_STABLE_WRITES)) {
|
mm/swap: remember PG_anon_exclusive via a swp pte bit
Patch series "mm: COW fixes part 3: reliable GUP R/W FOLL_GET of anonymous pages", v2.
This series fixes memory corruptions when a GUP R/W reference (FOLL_WRITE
| FOLL_GET) was taken on an anonymous page and COW logic fails to detect
exclusivity of the page to then replacing the anonymous page by a copy in
the page table: The GUP reference lost synchronicity with the pages mapped
into the page tables. This series focuses on x86, arm64, s390x and
ppc64/book3s -- other architectures are fairly easy to support by
implementing __HAVE_ARCH_PTE_SWP_EXCLUSIVE.
This primarily fixes the O_DIRECT memory corruptions that can happen on
concurrent swapout, whereby we lose DMA reads to a page (modifying the
user page by writing to it).
O_DIRECT currently uses FOLL_GET for short-term (!FOLL_LONGTERM) DMA
from/to a user page. In the long run, we want to convert it to properly
use FOLL_PIN, and John is working on it, but that might take a while and
might not be easy to backport. In the meantime, let's restore what used
to work before we started modifying our COW logic: make R/W FOLL_GET
references reliable as long as there is no fork() after GUP involved.
This is just the natural follow-up of part 2, that will also further
reduce "wrong COW" on the swapin path, for example, when we cannot remove
a page from the swapcache due to concurrent writeback, or if we have two
threads faulting on the same swapped-out page. Fixing O_DIRECT is just a
nice side-product
This issue, including other related COW issues, has been summarized in [3]
under 2):
"
2. Intra Process Memory Corruptions due to Wrong COW (FOLL_GET)
It was discovered that we can create a memory corruption by reading a
file via O_DIRECT to a part (e.g., first 512 bytes) of a page,
concurrently writing to an unrelated part (e.g., last byte) of the same
page, and concurrently write-protecting the page via clear_refs
SOFTDIRTY tracking [6].
For the reproducer, the issue is that O_DIRECT grabs a reference of the
target page (via FOLL_GET) and clear_refs write-protects the relevant
page table entry. On successive write access to the page from the
process itself, we wrongly COW the page when resolving the write fault,
resulting in a loss of synchronicity and consequently a memory corruption.
While some people might think that using clear_refs in this combination
is a corner cases, it turns out to be a more generic problem unfortunately.
For example, it was just recently discovered that we can similarly
create a memory corruption without clear_refs, simply by concurrently
swapping out the buffer pages [7]. Note that we nowadays even use the
swap infrastructure in Linux without an actual swap disk/partition: the
prime example is zram which is enabled as default under Fedora [10].
The root issue is that a write-fault on a page that has additional
references results in a COW and thereby a loss of synchronicity
and consequently a memory corruption if two parties believe they are
referencing the same page.
"
We don't particularly care about R/O FOLL_GET references: they were never
reliable and O_DIRECT doesn't expect to observe modifications from a page
after DMA was started.
Note that:
* this only fixes the issue on x86, arm64, s390x and ppc64/book3s
("enterprise architectures"). Other architectures have to implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE to achieve the same.
* this does *not * consider any kind of fork() after taking the reference:
fork() after GUP never worked reliably with FOLL_GET.
* Not losing PG_anon_exclusive during swapout was the last remaining
piece. KSM already makes sure that there are no other references on
a page before considering it for sharing. Page migration maintains
PG_anon_exclusive and simply fails when there are additional references
(freezing the refcount fails). Only swapout code dropped the
PG_anon_exclusive flag because it requires more work to remember +
restore it.
With this series in place, most COW issues of [3] are fixed on said
architectures. Other architectures can implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE fairly easily.
[1] https://lkml.kernel.org/r/20220329160440.193848-1-david@redhat.com
[2] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[3] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 8):
Currently, we clear PG_anon_exclusive in try_to_unmap() and forget about
it. We do this, to keep fork() logic on swap entries easy and efficient:
for example, if we wouldn't clear it when unmapping, we'd have to lookup
the page in the swapcache for each and every swap entry during fork() and
clear PG_anon_exclusive if set.
Instead, we want to store that information directly in the swap pte,
protected by the page table lock, similarly to how we handle
SWP_MIGRATION_READ_EXCLUSIVE for migration entries. However, for actual
swap entries, we don't want to mess with the swap type (e.g., still one
bit) because it overcomplicates swap code.
In try_to_unmap(), we already reject to unmap in case the page might be
pinned, because we must not lose PG_anon_exclusive on pinned pages ever.
Checking if there are other unexpected references reliably *before*
completely unmapping a page is unfortunately not really possible: THP
heavily overcomplicate the situation. Once fully unmapped it's easier --
we, for example, make sure that there are no unexpected references *after*
unmapping a page before starting writeback on that page.
So, we currently might end up unmapping a page and clearing
PG_anon_exclusive if that page has additional references, for example, due
to a FOLL_GET.
do_swap_page() has to re-determine if a page is exclusive, which will
easily fail if there are other references on a page, most prominently GUP
references via FOLL_GET. This can currently result in memory corruptions
when taking a FOLL_GET | FOLL_WRITE reference on a page even when fork()
is never involved: try_to_unmap() will succeed, and when refaulting the
page, it cannot be marked exclusive and will get replaced by a copy in the
page tables on the next write access, resulting in writes via the GUP
reference to the page being lost.
In an ideal world, everybody that uses GUP and wants to modify page
content, such as O_DIRECT, would properly use FOLL_PIN. However, that
conversion will take a while. It's easier to fix what used to work in the
past (FOLL_GET | FOLL_WRITE) remembering PG_anon_exclusive. In addition,
by remembering PG_anon_exclusive we can further reduce unnecessary COW in
some cases, so it's the natural thing to do.
So let's transfer the PG_anon_exclusive information to the swap pte and
store it via an architecture-dependant pte bit; use that information when
restoring the swap pte in do_swap_page() and unuse_pte(). During fork(),
we simply have to clear the pte bit and are done.
Of course, there is one corner case to handle: swap backends that don't
support concurrent page modifications while the page is under writeback.
Special case these, and drop the exclusive marker. Add a comment why that
is just fine (also, reuse_swap_page() would have done the same in the
past).
In the future, we'll hopefully have all architectures support
__HAVE_ARCH_PTE_SWP_EXCLUSIVE, such that we can get rid of the empty stubs
and the define completely. Then, we can also convert
SWP_MIGRATION_READ_EXCLUSIVE. For architectures it's fairly easy to
support: either simply use a yet unused pte bit that can be used for swap
entries, steal one from the arch type bits if they exceed 5, or steal one
from the offset bits.
Note: R/O FOLL_GET references were never really reliable, especially when
taking one on a shared page and then writing to the page (e.g., GUP after
fork()). FOLL_GET, including R/W references, were never really reliable
once fork was involved (e.g., GUP before fork(), GUP during fork()). KSM
steps back in case it stumbles over unexpected references and is,
therefore, fine.
[david@redhat.com: fix SWP_STABLE_WRITES test]
Link: https://lkml.kernel.org/r/ac725bcb-313a-4fff-250a-68ba9a8f85fb@redhat.comLink: https://lkml.kernel.org/r/20220329164329.208407-1-david@redhat.com
Link: https://lkml.kernel.org/r/20220329164329.208407-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
/*
|
|
|
|
* This is tricky: not all swap backends support
|
|
|
|
* concurrent page modifications while under writeback.
|
|
|
|
*
|
|
|
|
* So if we stumble over such a page in the swapcache
|
|
|
|
* we must not set the page exclusive, otherwise we can
|
|
|
|
* map it writable without further checks and modify it
|
|
|
|
* while still under writeback.
|
|
|
|
*
|
|
|
|
* For these problematic swap backends, simply drop the
|
|
|
|
* exclusive marker: this is perfectly fine as we start
|
|
|
|
* writeback only if we fully unmapped the page and
|
|
|
|
* there are no unexpected references on the page after
|
|
|
|
* unmapping succeeded. After fully unmapped, no
|
|
|
|
* further GUP references (FOLL_GET and FOLL_PIN) can
|
|
|
|
* appear, so dropping the exclusive marker and mapping
|
|
|
|
* it only R/O is fine.
|
|
|
|
*/
|
|
|
|
exclusive = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2023-05-23 00:43:08 +00:00
|
|
|
/*
|
|
|
|
* Some architectures may have to restore extra metadata to the page
|
|
|
|
* when reading from swap. This metadata may be indexed by swap entry
|
|
|
|
* so this must be called before swap_free().
|
|
|
|
*/
|
|
|
|
arch_swap_restore(entry, folio);
|
|
|
|
|
2009-01-08 02:08:00 +00:00
|
|
|
/*
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
* Remove the swap entry and conditionally try to free up the swapcache.
|
|
|
|
* We're already holding a reference on the page but haven't mapped it
|
|
|
|
* yet.
|
2009-01-08 02:08:00 +00:00
|
|
|
*/
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
swap_free(entry);
|
2022-09-02 19:46:42 +00:00
|
|
|
if (should_try_to_free_swap(folio, vma, vmf->flags))
|
|
|
|
folio_free_swap(folio);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2022-10-24 05:28:41 +00:00
|
|
|
inc_mm_counter(vma->vm_mm, MM_ANONPAGES);
|
|
|
|
dec_mm_counter(vma->vm_mm, MM_SWAPENTS);
|
2005-04-16 22:20:36 +00:00
|
|
|
pte = mk_pte(page, vma->vm_page_prot);
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
|
|
|
|
/*
|
mm/swap: remember PG_anon_exclusive via a swp pte bit
Patch series "mm: COW fixes part 3: reliable GUP R/W FOLL_GET of anonymous pages", v2.
This series fixes memory corruptions when a GUP R/W reference (FOLL_WRITE
| FOLL_GET) was taken on an anonymous page and COW logic fails to detect
exclusivity of the page to then replacing the anonymous page by a copy in
the page table: The GUP reference lost synchronicity with the pages mapped
into the page tables. This series focuses on x86, arm64, s390x and
ppc64/book3s -- other architectures are fairly easy to support by
implementing __HAVE_ARCH_PTE_SWP_EXCLUSIVE.
This primarily fixes the O_DIRECT memory corruptions that can happen on
concurrent swapout, whereby we lose DMA reads to a page (modifying the
user page by writing to it).
O_DIRECT currently uses FOLL_GET for short-term (!FOLL_LONGTERM) DMA
from/to a user page. In the long run, we want to convert it to properly
use FOLL_PIN, and John is working on it, but that might take a while and
might not be easy to backport. In the meantime, let's restore what used
to work before we started modifying our COW logic: make R/W FOLL_GET
references reliable as long as there is no fork() after GUP involved.
This is just the natural follow-up of part 2, that will also further
reduce "wrong COW" on the swapin path, for example, when we cannot remove
a page from the swapcache due to concurrent writeback, or if we have two
threads faulting on the same swapped-out page. Fixing O_DIRECT is just a
nice side-product
This issue, including other related COW issues, has been summarized in [3]
under 2):
"
2. Intra Process Memory Corruptions due to Wrong COW (FOLL_GET)
It was discovered that we can create a memory corruption by reading a
file via O_DIRECT to a part (e.g., first 512 bytes) of a page,
concurrently writing to an unrelated part (e.g., last byte) of the same
page, and concurrently write-protecting the page via clear_refs
SOFTDIRTY tracking [6].
For the reproducer, the issue is that O_DIRECT grabs a reference of the
target page (via FOLL_GET) and clear_refs write-protects the relevant
page table entry. On successive write access to the page from the
process itself, we wrongly COW the page when resolving the write fault,
resulting in a loss of synchronicity and consequently a memory corruption.
While some people might think that using clear_refs in this combination
is a corner cases, it turns out to be a more generic problem unfortunately.
For example, it was just recently discovered that we can similarly
create a memory corruption without clear_refs, simply by concurrently
swapping out the buffer pages [7]. Note that we nowadays even use the
swap infrastructure in Linux without an actual swap disk/partition: the
prime example is zram which is enabled as default under Fedora [10].
The root issue is that a write-fault on a page that has additional
references results in a COW and thereby a loss of synchronicity
and consequently a memory corruption if two parties believe they are
referencing the same page.
"
We don't particularly care about R/O FOLL_GET references: they were never
reliable and O_DIRECT doesn't expect to observe modifications from a page
after DMA was started.
Note that:
* this only fixes the issue on x86, arm64, s390x and ppc64/book3s
("enterprise architectures"). Other architectures have to implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE to achieve the same.
* this does *not * consider any kind of fork() after taking the reference:
fork() after GUP never worked reliably with FOLL_GET.
* Not losing PG_anon_exclusive during swapout was the last remaining
piece. KSM already makes sure that there are no other references on
a page before considering it for sharing. Page migration maintains
PG_anon_exclusive and simply fails when there are additional references
(freezing the refcount fails). Only swapout code dropped the
PG_anon_exclusive flag because it requires more work to remember +
restore it.
With this series in place, most COW issues of [3] are fixed on said
architectures. Other architectures can implement
__HAVE_ARCH_PTE_SWP_EXCLUSIVE fairly easily.
[1] https://lkml.kernel.org/r/20220329160440.193848-1-david@redhat.com
[2] https://lkml.kernel.org/r/20211217113049.23850-1-david@redhat.com
[3] https://lore.kernel.org/r/3ae33b08-d9ef-f846-56fb-645e3b9b4c66@redhat.com
This patch (of 8):
Currently, we clear PG_anon_exclusive in try_to_unmap() and forget about
it. We do this, to keep fork() logic on swap entries easy and efficient:
for example, if we wouldn't clear it when unmapping, we'd have to lookup
the page in the swapcache for each and every swap entry during fork() and
clear PG_anon_exclusive if set.
Instead, we want to store that information directly in the swap pte,
protected by the page table lock, similarly to how we handle
SWP_MIGRATION_READ_EXCLUSIVE for migration entries. However, for actual
swap entries, we don't want to mess with the swap type (e.g., still one
bit) because it overcomplicates swap code.
In try_to_unmap(), we already reject to unmap in case the page might be
pinned, because we must not lose PG_anon_exclusive on pinned pages ever.
Checking if there are other unexpected references reliably *before*
completely unmapping a page is unfortunately not really possible: THP
heavily overcomplicate the situation. Once fully unmapped it's easier --
we, for example, make sure that there are no unexpected references *after*
unmapping a page before starting writeback on that page.
So, we currently might end up unmapping a page and clearing
PG_anon_exclusive if that page has additional references, for example, due
to a FOLL_GET.
do_swap_page() has to re-determine if a page is exclusive, which will
easily fail if there are other references on a page, most prominently GUP
references via FOLL_GET. This can currently result in memory corruptions
when taking a FOLL_GET | FOLL_WRITE reference on a page even when fork()
is never involved: try_to_unmap() will succeed, and when refaulting the
page, it cannot be marked exclusive and will get replaced by a copy in the
page tables on the next write access, resulting in writes via the GUP
reference to the page being lost.
In an ideal world, everybody that uses GUP and wants to modify page
content, such as O_DIRECT, would properly use FOLL_PIN. However, that
conversion will take a while. It's easier to fix what used to work in the
past (FOLL_GET | FOLL_WRITE) remembering PG_anon_exclusive. In addition,
by remembering PG_anon_exclusive we can further reduce unnecessary COW in
some cases, so it's the natural thing to do.
So let's transfer the PG_anon_exclusive information to the swap pte and
store it via an architecture-dependant pte bit; use that information when
restoring the swap pte in do_swap_page() and unuse_pte(). During fork(),
we simply have to clear the pte bit and are done.
Of course, there is one corner case to handle: swap backends that don't
support concurrent page modifications while the page is under writeback.
Special case these, and drop the exclusive marker. Add a comment why that
is just fine (also, reuse_swap_page() would have done the same in the
past).
In the future, we'll hopefully have all architectures support
__HAVE_ARCH_PTE_SWP_EXCLUSIVE, such that we can get rid of the empty stubs
and the define completely. Then, we can also convert
SWP_MIGRATION_READ_EXCLUSIVE. For architectures it's fairly easy to
support: either simply use a yet unused pte bit that can be used for swap
entries, steal one from the arch type bits if they exceed 5, or steal one
from the offset bits.
Note: R/O FOLL_GET references were never really reliable, especially when
taking one on a shared page and then writing to the page (e.g., GUP after
fork()). FOLL_GET, including R/W references, were never really reliable
once fork was involved (e.g., GUP before fork(), GUP during fork()). KSM
steps back in case it stumbles over unexpected references and is,
therefore, fine.
[david@redhat.com: fix SWP_STABLE_WRITES test]
Link: https://lkml.kernel.org/r/ac725bcb-313a-4fff-250a-68ba9a8f85fb@redhat.comLink: https://lkml.kernel.org/r/20220329164329.208407-1-david@redhat.com
Link: https://lkml.kernel.org/r/20220329164329.208407-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
* Same logic as in do_wp_page(); however, optimize for pages that are
|
|
|
|
* certainly not shared either because we just allocated them without
|
|
|
|
* exposing them to the swapcache or because the swap entry indicates
|
|
|
|
* exclusivity.
|
mm: streamline COW logic in do_swap_page()
Currently we have a different COW logic when:
* triggering a read-fault to swapin first and then trigger a write-fault
-> do_swap_page() + do_wp_page()
* triggering a write-fault to swapin
-> do_swap_page() + do_wp_page() only if we fail reuse in do_swap_page()
The COW logic in do_swap_page() is different than our reuse logic in
do_wp_page(). The COW logic in do_wp_page() -- page_count() == 1 -- makes
currently sure that we certainly don't have a remaining reference, e.g.,
via GUP, on the target page we want to reuse: if there is any unexpected
reference, we have to copy to avoid information leaks.
As do_swap_page() behaves differently, in environments with swap enabled
we can currently have an unintended information leak from the parent to
the child, similar as known from CVE-2020-29374:
1. Parent writes to anonymous page
-> Page is mapped writable and modified
2. Page is swapped out
-> Page is unmapped and replaced by swap entry
3. fork()
-> Swap entries are copied to child
4. Child pins page R/O
-> Page is mapped R/O into child
5. Child unmaps page
-> Child still holds GUP reference
6. Parent writes to page
-> Page is reused in do_swap_page()
-> Child can observe changes
Exchanging 2. and 3. should have the same effect.
Let's apply the same COW logic as in do_wp_page(), conditionally trying to
remove the page from the swapcache after freeing the swap entry, however,
before actually mapping our page. We can change the order now that we use
try_to_free_swap(), which doesn't care about the mapcount, instead of
reuse_swap_page().
To handle references from the LRU pagevecs, conditionally drain the local
LRU pagevecs when required, however, don't consider the page_count() when
deciding whether to drain to keep it simple for now.
Link: https://lkml.kernel.org/r/20220131162940.210846-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-25 01:13:40 +00:00
|
|
|
*/
|
2022-09-02 19:46:10 +00:00
|
|
|
if (!folio_test_ksm(folio) &&
|
|
|
|
(exclusive || folio_ref_count(folio) == 1)) {
|
mm: remember exclusively mapped anonymous pages with PG_anon_exclusive
Let's mark exclusively mapped anonymous pages with PG_anon_exclusive as
exclusive, and use that information to make GUP pins reliable and stay
consistent with the page mapped into the page table even if the page table
entry gets write-protected.
With that information at hand, we can extend our COW logic to always reuse
anonymous pages that are exclusive. For anonymous pages that might be
shared, the existing logic applies.
As already documented, PG_anon_exclusive is usually only expressive in
combination with a page table entry. Especially PTE vs. PMD-mapped
anonymous pages require more thought, some examples: due to mremap() we
can easily have a single compound page PTE-mapped into multiple page
tables exclusively in a single process -- multiple page table locks apply.
Further, due to MADV_WIPEONFORK we might not necessarily write-protect
all PTEs, and only some subpages might be pinned. Long story short: once
PTE-mapped, we have to track information about exclusivity per sub-page,
but until then, we can just track it for the compound page in the head
page and not having to update a whole bunch of subpages all of the time
for a simple PMD mapping of a THP.
For simplicity, this commit mostly talks about "anonymous pages", while
it's for THP actually "the part of an anonymous folio referenced via a
page table entry".
To not spill PG_anon_exclusive code all over the mm code-base, we let the
anon rmap code to handle all PG_anon_exclusive logic it can easily handle.
If a writable, present page table entry points at an anonymous (sub)page,
that (sub)page must be PG_anon_exclusive. If GUP wants to take a reliably
pin (FOLL_PIN) on an anonymous page references via a present page table
entry, it must only pin if PG_anon_exclusive is set for the mapped
(sub)page.
This commit doesn't adjust GUP, so this is only implicitly handled for
FOLL_WRITE, follow-up commits will teach GUP to also respect it for
FOLL_PIN without FOLL_WRITE, to make all GUP pins of anonymous pages fully
reliable.
Whenever an anonymous page is to be shared (fork(), KSM), or when
temporarily unmapping an anonymous page (swap, migration), the relevant
PG_anon_exclusive bit has to be cleared to mark the anonymous page
possibly shared. Clearing will fail if there are GUP pins on the page:
* For fork(), this means having to copy the page and not being able to
share it. fork() protects against concurrent GUP using the PT lock and
the src_mm->write_protect_seq.
* For KSM, this means sharing will fail. For swap this means, unmapping
will fail, For migration this means, migration will fail early. All
three cases protect against concurrent GUP using the PT lock and a
proper clear/invalidate+flush of the relevant page table entry.
This fixes memory corruptions reported for FOLL_PIN | FOLL_WRITE, when a
pinned page gets mapped R/O and the successive write fault ends up
replacing the page instead of reusing it. It improves the situation for
O_DIRECT/vmsplice/... that still use FOLL_GET instead of FOLL_PIN, if
fork() is *not* involved, however swapout and fork() are still
problematic. Properly using FOLL_PIN instead of FOLL_GET for these GUP
users will fix the issue for them.
I. Details about basic handling
I.1. Fresh anonymous pages
page_add_new_anon_rmap() and hugepage_add_new_anon_rmap() will mark the
given page exclusive via __page_set_anon_rmap(exclusive=1). As that is
the mechanism fresh anonymous pages come into life (besides migration code
where we copy the page->mapping), all fresh anonymous pages will start out
as exclusive.
I.2. COW reuse handling of anonymous pages
When a COW handler stumbles over a (sub)page that's marked exclusive, it
simply reuses it. Otherwise, the handler tries harder under page lock to
detect if the (sub)page is exclusive and can be reused. If exclusive,
page_move_anon_rmap() will mark the given (sub)page exclusive.
Note that hugetlb code does not yet check for PageAnonExclusive(), as it
still uses the old COW logic that is prone to the COW security issue
because hugetlb code cannot really tolerate unnecessary/wrong COW as huge
pages are a scarce resource.
I.3. Migration handling
try_to_migrate() has to try marking an exclusive anonymous page shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. migrate_vma_collect_pmd() and
__split_huge_pmd_locked() are handled similarly.
Writable migration entries implicitly point at shared anonymous pages.
For readable migration entries that information is stored via a new
"readable-exclusive" migration entry, specific to anonymous pages.
When restoring a migration entry in remove_migration_pte(), information
about exlusivity is detected via the migration entry type, and
RMAP_EXCLUSIVE is set accordingly for
page_add_anon_rmap()/hugepage_add_anon_rmap() to restore that information.
I.4. Swapout handling
try_to_unmap() has to try marking the mapped page possibly shared via
page_try_share_anon_rmap(). If it fails because there are GUP pins on the
page, unmap fails. For now, information about exclusivity is lost. In
the future, we might want to remember that information in the swap entry
in some cases, however, it requires more thought, care, and a way to store
that information in swap entries.
I.5. Swapin handling
do_swap_page() will never stumble over exclusive anonymous pages in the
swap cache, as try_to_migrate() prohibits that. do_swap_page() always has
to detect manually if an anonymous page is exclusive and has to set
RMAP_EXCLUSIVE for page_add_anon_rmap() accordingly.
I.6. THP handling
__split_huge_pmd_locked() has to move the information about exclusivity
from the PMD to the PTEs.
a) In case we have a readable-exclusive PMD migration entry, simply
insert readable-exclusive PTE migration entries.
b) In case we have a present PMD entry and we don't want to freeze
("convert to migration entries"), simply forward PG_anon_exclusive to
all sub-pages, no need to temporarily clear the bit.
c) In case we have a present PMD entry and want to freeze, handle it
similar to try_to_migrate(): try marking the page shared first. In
case we fail, we ignore the "freeze" instruction and simply split
ordinarily. try_to_migrate() will properly fail because the THP is
still mapped via PTEs.
When splitting a compound anonymous folio (THP), the information about
exclusivity is implicitly handled via the migration entries: no need to
replicate PG_anon_exclusive manually.
I.7. fork() handling fork() handling is relatively easy, because
PG_anon_exclusive is only expressive for some page table entry types.
a) Present anonymous pages
page_try_dup_anon_rmap() will mark the given subpage shared -- which will
fail if the page is pinned. If it failed, we have to copy (or PTE-map a
PMD to handle it on the PTE level).
Note that device exclusive entries are just a pointer at a PageAnon()
page. fork() will first convert a device exclusive entry to a present
page table and handle it just like present anonymous pages.
b) Device private entry
Device private entries point at PageAnon() pages that cannot be mapped
directly and, therefore, cannot get pinned.
page_try_dup_anon_rmap() will mark the given subpage shared, which cannot
fail because they cannot get pinned.
c) HW poison entries
PG_anon_exclusive will remain untouched and is stale -- the page table
entry is just a placeholder after all.
d) Migration entries
Writable and readable-exclusive entries are converted to readable entries:
possibly shared.
I.8. mprotect() handling
mprotect() only has to properly handle the new readable-exclusive
migration entry:
When write-protecting a migration entry that points at an anonymous page,
remember the information about exclusivity via the "readable-exclusive"
migration entry type.
II. Migration and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a migration entry, we have to mark the page possibly
shared and synchronize against GUP-fast by a proper clear/invalidate+flush
to make the following scenario impossible:
1. try_to_migrate() places a migration entry after checking for GUP pins
and marks the page possibly shared.
2. GUP-fast pins the page due to lack of synchronization
3. fork() converts the "writable/readable-exclusive" migration entry into a
readable migration entry
4. Migration fails due to the GUP pin (failing to freeze the refcount)
5. Migration entries are restored. PG_anon_exclusive is lost
-> We have a pinned page that is not marked exclusive anymore.
Note that we move information about exclusivity from the page to the
migration entry as it otherwise highly overcomplicates fork() and
PTE-mapping a THP.
III. Swapout and GUP-fast
Whenever replacing a present page table entry that maps an exclusive
anonymous page by a swap entry, we have to mark the page possibly shared
and synchronize against GUP-fast by a proper clear/invalidate+flush to
make the following scenario impossible:
1. try_to_unmap() places a swap entry after checking for GUP pins and
clears exclusivity information on the page.
2. GUP-fast pins the page due to lack of synchronization.
-> We have a pinned page that is not marked exclusive anymore.
If we'd ever store information about exclusivity in the swap entry,
similar to migration handling, the same considerations as in II would
apply. This is future work.
Link: https://lkml.kernel.org/r/20220428083441.37290-13-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:44 +00:00
|
|
|
if (vmf->flags & FAULT_FLAG_WRITE) {
|
|
|
|
pte = maybe_mkwrite(pte_mkdirty(pte), vma);
|
|
|
|
vmf->flags &= ~FAULT_FLAG_WRITE;
|
|
|
|
}
|
2022-05-10 01:20:43 +00:00
|
|
|
rmap_flags |= RMAP_EXCLUSIVE;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
flush_icache_page(vma, page);
|
2016-12-14 23:07:16 +00:00
|
|
|
if (pte_swp_soft_dirty(vmf->orig_pte))
|
2013-08-13 23:00:49 +00:00
|
|
|
pte = pte_mksoft_dirty(pte);
|
2022-12-14 20:15:33 +00:00
|
|
|
if (pte_swp_uffd_wp(vmf->orig_pte))
|
2020-04-07 03:06:01 +00:00
|
|
|
pte = pte_mkuffd_wp(pte);
|
2016-12-14 23:07:16 +00:00
|
|
|
vmf->orig_pte = pte;
|
2017-11-16 01:33:07 +00:00
|
|
|
|
|
|
|
/* ksm created a completely new copy */
|
2022-09-02 19:46:11 +00:00
|
|
|
if (unlikely(folio != swapcache && swapcache)) {
|
2022-05-10 01:20:43 +00:00
|
|
|
page_add_new_anon_rmap(page, vma, vmf->address);
|
2022-09-02 19:46:10 +00:00
|
|
|
folio_add_lru_vma(folio, vma);
|
2017-11-16 01:33:07 +00:00
|
|
|
} else {
|
2022-05-10 01:20:43 +00:00
|
|
|
page_add_anon_rmap(page, vma, vmf->address, rmap_flags);
|
mm: memcontrol: rewrite charge API
These patches rework memcg charge lifetime to integrate more naturally
with the lifetime of user pages. This drastically simplifies the code and
reduces charging and uncharging overhead. The most expensive part of
charging and uncharging is the page_cgroup bit spinlock, which is removed
entirely after this series.
Here are the top-10 profile entries of a stress test that reads a 128G
sparse file on a freshly booted box, without even a dedicated cgroup (i.e.
executing in the root memcg). Before:
15.36% cat [kernel.kallsyms] [k] copy_user_generic_string
13.31% cat [kernel.kallsyms] [k] memset
11.48% cat [kernel.kallsyms] [k] do_mpage_readpage
4.23% cat [kernel.kallsyms] [k] get_page_from_freelist
2.38% cat [kernel.kallsyms] [k] put_page
2.32% cat [kernel.kallsyms] [k] __mem_cgroup_commit_charge
2.18% kswapd0 [kernel.kallsyms] [k] __mem_cgroup_uncharge_common
1.92% kswapd0 [kernel.kallsyms] [k] shrink_page_list
1.86% cat [kernel.kallsyms] [k] __radix_tree_lookup
1.62% cat [kernel.kallsyms] [k] __pagevec_lru_add_fn
After:
15.67% cat [kernel.kallsyms] [k] copy_user_generic_string
13.48% cat [kernel.kallsyms] [k] memset
11.42% cat [kernel.kallsyms] [k] do_mpage_readpage
3.98% cat [kernel.kallsyms] [k] get_page_from_freelist
2.46% cat [kernel.kallsyms] [k] put_page
2.13% kswapd0 [kernel.kallsyms] [k] shrink_page_list
1.88% cat [kernel.kallsyms] [k] __radix_tree_lookup
1.67% cat [kernel.kallsyms] [k] __pagevec_lru_add_fn
1.39% kswapd0 [kernel.kallsyms] [k] free_pcppages_bulk
1.30% cat [kernel.kallsyms] [k] kfree
As you can see, the memcg footprint has shrunk quite a bit.
text data bss dec hex filename
37970 9892 400 48262 bc86 mm/memcontrol.o.old
35239 9892 400 45531 b1db mm/memcontrol.o
This patch (of 4):
The memcg charge API charges pages before they are rmapped - i.e. have an
actual "type" - and so every callsite needs its own set of charge and
uncharge functions to know what type is being operated on. Worse,
uncharge has to happen from a context that is still type-specific, rather
than at the end of the page's lifetime with exclusive access, and so
requires a lot of synchronization.
Rewrite the charge API to provide a generic set of try_charge(),
commit_charge() and cancel_charge() transaction operations, much like
what's currently done for swap-in:
mem_cgroup_try_charge() attempts to reserve a charge, reclaiming
pages from the memcg if necessary.
mem_cgroup_commit_charge() commits the page to the charge once it
has a valid page->mapping and PageAnon() reliably tells the type.
mem_cgroup_cancel_charge() aborts the transaction.
This reduces the charge API and enables subsequent patches to
drastically simplify uncharging.
As pages need to be committed after rmap is established but before they
are added to the LRU, page_add_new_anon_rmap() must stop doing LRU
additions again. Revive lru_cache_add_active_or_unevictable().
[hughd@google.com: fix shmem_unuse]
[hughd@google.com: Add comments on the private use of -EAGAIN]
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vladimir Davydov <vdavydov@parallels.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-08-08 21:19:20 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2022-09-02 19:46:10 +00:00
|
|
|
VM_BUG_ON(!folio_test_anon(folio) ||
|
|
|
|
(pte_write(pte) && !PageAnonExclusive(page)));
|
2022-01-14 22:06:29 +00:00
|
|
|
set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
|
|
|
|
arch_do_swap_page(vma->vm_mm, vma, vmf->address, pte, vmf->orig_pte);
|
|
|
|
|
2022-09-02 19:46:10 +00:00
|
|
|
folio_unlock(folio);
|
2022-09-02 19:46:11 +00:00
|
|
|
if (folio != swapcache && swapcache) {
|
2010-09-09 23:37:52 +00:00
|
|
|
/*
|
|
|
|
* Hold the lock to avoid the swap entry to be reused
|
|
|
|
* until we take the PT lock for the pte_same() check
|
|
|
|
* (to avoid false positives from pte_same). For
|
|
|
|
* further safety release the lock after the swap_free
|
|
|
|
* so that the swap count won't change under a
|
|
|
|
* parallel locked swapcache.
|
|
|
|
*/
|
2022-09-02 19:46:11 +00:00
|
|
|
folio_unlock(swapcache);
|
|
|
|
folio_put(swapcache);
|
2010-09-09 23:37:52 +00:00
|
|
|
}
|
[PATCH] can_share_swap_page: use page_mapcount
Remember that ironic get_user_pages race? when the raised page_count on a
page swapped out led do_wp_page to decide that it had to copy on write, so
substituted a different page into userspace. 2.6.7 onwards have Andrea's
solution, where try_to_unmap_one backs out if it finds page_count raised.
Which works, but is unsatisfying (rmap.c has no other page_count heuristics),
and was found a few months ago to hang an intensive page migration test. A
year ago I was hesitant to engage page_mapcount, now it seems the right fix.
So remove the page_count hack from try_to_unmap_one; and use activate_page in
unuse_mm when dropping lock, to replace its secondary effect of helping
swapoff to make progress in that case.
Simplify can_share_swap_page (now called only on anonymous pages) to check
page_mapcount + page_swapcount == 1: still needs the page lock to stabilize
their (pessimistic) sum, but does not need swapper_space.tree_lock for that.
In do_swap_page, move swap_free and unlock_page below page_add_anon_rmap, to
keep sum on the high side, and correct when can_share_swap_page called.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:15:12 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
if (vmf->flags & FAULT_FLAG_WRITE) {
|
2016-12-14 23:07:16 +00:00
|
|
|
ret |= do_wp_page(vmf);
|
2008-03-04 22:29:04 +00:00
|
|
|
if (ret & VM_FAULT_ERROR)
|
|
|
|
ret &= VM_FAULT_ERROR;
|
2005-04-16 22:20:36 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* No need to invalidate - it was non-present before */
|
2016-12-14 23:06:58 +00:00
|
|
|
update_mmu_cache(vma, vmf->address, vmf->pte);
|
[PATCH] mm: page fault handlers tidyup
Impose a little more consistency on the page fault handlers do_wp_page,
do_swap_page, do_anonymous_page, do_no_page, do_file_page: why not pass their
arguments in the same order, called the same names?
break_cow is all very well, but what it did was inlined elsewhere: easier to
compare if it's brought back into do_wp_page.
do_file_page's fallback to do_no_page dates from a time when we were testing
pte_file by using it wherever possible: currently it's peculiar to nonlinear
vmas, so just check that. BUG_ON if not? Better not, it's probably page
table corruption, so just show the pte: hmm, there's a pte_ERROR macro, let's
use that for do_wp_page's invalid pfn too.
Hah! Someone in the ppc64 world noticed pte_ERROR was unused so removed it:
restored (and say "pud" not "pmd" in its pud_ERROR).
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 01:15:59 +00:00
|
|
|
unlock:
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2005-04-16 22:20:36 +00:00
|
|
|
out:
|
2021-06-29 02:36:50 +00:00
|
|
|
if (si)
|
|
|
|
put_swap_device(si);
|
2005-04-16 22:20:36 +00:00
|
|
|
return ret;
|
2005-05-17 04:53:50 +00:00
|
|
|
out_nomap:
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2009-04-30 22:08:08 +00:00
|
|
|
out_page:
|
2022-09-02 19:46:10 +00:00
|
|
|
folio_unlock(folio);
|
2009-10-13 23:51:41 +00:00
|
|
|
out_release:
|
2022-09-02 19:46:10 +00:00
|
|
|
folio_put(folio);
|
2022-09-02 19:46:11 +00:00
|
|
|
if (folio != swapcache && swapcache) {
|
|
|
|
folio_unlock(swapcache);
|
|
|
|
folio_put(swapcache);
|
2010-09-09 23:37:52 +00:00
|
|
|
}
|
2021-06-29 02:36:50 +00:00
|
|
|
if (si)
|
|
|
|
put_swap_device(si);
|
[PATCH] mm: page fault handlers tidyup
Impose a little more consistency on the page fault handlers do_wp_page,
do_swap_page, do_anonymous_page, do_no_page, do_file_page: why not pass their
arguments in the same order, called the same names?
break_cow is all very well, but what it did was inlined elsewhere: easier to
compare if it's brought back into do_wp_page.
do_file_page's fallback to do_no_page dates from a time when we were testing
pte_file by using it wherever possible: currently it's peculiar to nonlinear
vmas, so just check that. BUG_ON if not? Better not, it's probably page
table corruption, so just show the pte: hmm, there's a pte_ERROR macro, let's
use that for do_wp_page's invalid pfn too.
Hah! Someone in the ppc64 world noticed pte_ERROR was unused so removed it:
restored (and say "pud" not "pmd" in its pud_ERROR).
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 01:15:59 +00:00
|
|
|
return ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2020-06-09 04:33:54 +00:00
|
|
|
* We enter with non-exclusive mmap_lock (to exclude vma changes,
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
* but allow concurrent faults), and pte mapped but not yet locked.
|
2020-06-09 04:33:54 +00:00
|
|
|
* We return with mmap_lock still held, but pte unmapped and unlocked.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_anonymous_page(struct vm_fault *vmf)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
bool uffd_wp = vmf_orig_pte_uffd_wp(vmf);
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2023-01-16 19:18:09 +00:00
|
|
|
struct folio *folio;
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
pte_t entry;
|
|
|
|
|
2015-07-06 20:18:37 +00:00
|
|
|
/* File mapping without ->vm_ops ? */
|
|
|
|
if (vma->vm_flags & VM_SHARED)
|
|
|
|
return VM_FAULT_SIGBUS;
|
|
|
|
|
2016-07-26 22:25:23 +00:00
|
|
|
/*
|
|
|
|
* Use pte_alloc() instead of pte_alloc_map(). We can't run
|
|
|
|
* pte_offset_map() on pmds where a huge pmd might be created
|
|
|
|
* from a different thread.
|
|
|
|
*
|
2020-06-09 04:33:51 +00:00
|
|
|
* pte_alloc_map() is safe to use under mmap_write_lock(mm) or when
|
2016-07-26 22:25:23 +00:00
|
|
|
* parallel threads are excluded by other means.
|
|
|
|
*
|
2020-06-09 04:33:51 +00:00
|
|
|
* Here we only have mmap_read_lock(mm).
|
2016-07-26 22:25:23 +00:00
|
|
|
*/
|
mm: treewide: remove unused address argument from pte_alloc functions
Patch series "Add support for fast mremap".
This series speeds up the mremap(2) syscall by copying page tables at
the PMD level even for non-THP systems. There is concern that the extra
'address' argument that mremap passes to pte_alloc may do something
subtle architecture related in the future that may make the scheme not
work. Also we find that there is no point in passing the 'address' to
pte_alloc since its unused. This patch therefore removes this argument
tree-wide resulting in a nice negative diff as well. Also ensuring
along the way that the enabled architectures do not do anything funky
with the 'address' argument that goes unnoticed by the optimization.
Build and boot tested on x86-64. Build tested on arm64. The config
enablement patch for arm64 will be posted in the future after more
testing.
The changes were obtained by applying the following Coccinelle script.
(thanks Julia for answering all Coccinelle questions!).
Following fix ups were done manually:
* Removal of address argument from pte_fragment_alloc
* Removal of pte_alloc_one_fast definitions from m68k and microblaze.
// Options: --include-headers --no-includes
// Note: I split the 'identifier fn' line, so if you are manually
// running it, please unsplit it so it runs for you.
virtual patch
@pte_alloc_func_def depends on patch exists@
identifier E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
type T2;
@@
fn(...
- , T2 E2
)
{ ... }
@pte_alloc_func_proto_noarg depends on patch exists@
type T1, T2, T3, T4;
identifier fn =~ "^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1, T2);
+ T3 fn(T1);
|
- T3 fn(T1, T2, T4);
+ T3 fn(T1, T2);
)
@pte_alloc_func_proto depends on patch exists@
identifier E1, E2, E4;
type T1, T2, T3, T4;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1 E1, T2 E2);
+ T3 fn(T1 E1);
|
- T3 fn(T1 E1, T2 E2, T4 E4);
+ T3 fn(T1 E1, T2 E2);
)
@pte_alloc_func_call depends on patch exists@
expression E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
fn(...
-, E2
)
@pte_alloc_macro depends on patch exists@
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
identifier a, b, c;
expression e;
position p;
@@
(
- #define fn(a, b, c) e
+ #define fn(a, b) e
|
- #define fn(a, b) e
+ #define fn(a) e
)
Link: http://lkml.kernel.org/r/20181108181201.88826-2-joelaf@google.com
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Suggested-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: William Kucharski <william.kucharski@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-03 23:28:34 +00:00
|
|
|
if (pte_alloc(vma->vm_mm, vmf->pmd))
|
2016-07-26 22:25:23 +00:00
|
|
|
return VM_FAULT_OOM;
|
|
|
|
|
2020-12-19 12:19:23 +00:00
|
|
|
/* See comment in handle_pte_fault() */
|
2016-12-14 23:06:58 +00:00
|
|
|
if (unlikely(pmd_trans_unstable(vmf->pmd)))
|
2016-07-26 22:25:23 +00:00
|
|
|
return 0;
|
|
|
|
|
2010-08-14 18:44:56 +00:00
|
|
|
/* Use the zero-page for reads */
|
2016-12-14 23:06:58 +00:00
|
|
|
if (!(vmf->flags & FAULT_FLAG_WRITE) &&
|
2016-07-26 22:25:20 +00:00
|
|
|
!mm_forbids_zeropage(vma->vm_mm)) {
|
2016-12-14 23:06:58 +00:00
|
|
|
entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
|
2009-09-22 00:03:34 +00:00
|
|
|
vma->vm_page_prot));
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
|
|
|
|
vmf->address, &vmf->ptl);
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
if (vmf_pte_changed(vmf)) {
|
2020-05-27 02:25:18 +00:00
|
|
|
update_mmu_tlb(vma, vmf->address, vmf->pte);
|
2009-09-22 00:03:30 +00:00
|
|
|
goto unlock;
|
2020-05-27 02:25:18 +00:00
|
|
|
}
|
2017-08-18 22:16:15 +00:00
|
|
|
ret = check_stable_address_space(vma->vm_mm);
|
|
|
|
if (ret)
|
|
|
|
goto unlock;
|
2015-09-04 22:46:20 +00:00
|
|
|
/* Deliver the page fault to userland, check inside PT lock */
|
|
|
|
if (userfaultfd_missing(vma)) {
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
|
|
|
return handle_userfault(vmf, VM_UFFD_MISSING);
|
2015-09-04 22:46:20 +00:00
|
|
|
}
|
2009-09-22 00:03:30 +00:00
|
|
|
goto setpte;
|
|
|
|
}
|
|
|
|
|
remove ZERO_PAGE
The commit b5810039a54e5babf428e9a1e89fc1940fabff11 contains the note
A last caveat: the ZERO_PAGE is now refcounted and managed with rmap
(and thus mapcounted and count towards shared rss). These writes to
the struct page could cause excessive cacheline bouncing on big
systems. There are a number of ways this could be addressed if it is
an issue.
And indeed this cacheline bouncing has shown up on large SGI systems.
There was a situation where an Altix system was essentially livelocked
tearing down ZERO_PAGE pagetables when an HPC app aborted during startup.
This situation can be avoided in userspace, but it does highlight the
potential scalability problem with refcounting ZERO_PAGE, and corner
cases where it can really hurt (we don't want the system to livelock!).
There are several broad ways to fix this problem:
1. add back some special casing to avoid refcounting ZERO_PAGE
2. per-node or per-cpu ZERO_PAGES
3. remove the ZERO_PAGE completely
I will argue for 3. The others should also fix the problem, but they
result in more complex code than does 3, with little or no real benefit
that I can see.
Why? Inserting a ZERO_PAGE for anonymous read faults appears to be a
false optimisation: if an application is performance critical, it would
not be doing many read faults of new memory, or at least it could be
expected to write to that memory soon afterwards. If cache or memory use
is critical, it should not be working with a significant number of
ZERO_PAGEs anyway (a more compact representation of zeroes should be
used).
As a sanity check -- mesuring on my desktop system, there are never many
mappings to the ZERO_PAGE (eg. 2 or 3), thus memory usage here should not
increase much without it.
When running a make -j4 kernel compile on my dual core system, there are
about 1,000 mappings to the ZERO_PAGE created per second, but about 1,000
ZERO_PAGE COW faults per second (less than 1 ZERO_PAGE mapping per second
is torn down without being COWed). So removing ZERO_PAGE will save 1,000
page faults per second when running kbuild, while keeping it only saves
less than 1 page clearing operation per second. 1 page clear is cheaper
than a thousand faults, presumably, so there isn't an obvious loss.
Neither the logical argument nor these basic tests give a guarantee of no
regressions. However, this is a reasonable opportunity to try to remove
the ZERO_PAGE from the pagefault path. If it is found to cause regressions,
we can reintroduce it and just avoid refcounting it.
The /dev/zero ZERO_PAGE usage and TLB tricks also get nuked. I don't see
much use to them except on benchmarks. All other users of ZERO_PAGE are
converted just to use ZERO_PAGE(0) for simplicity. We can look at
replacing them all and maybe ripping out ZERO_PAGE completely when we are
more satisfied with this solution.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus "snif" Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:24:40 +00:00
|
|
|
/* Allocate our own private page. */
|
|
|
|
if (unlikely(anon_vma_prepare(vma)))
|
|
|
|
goto oom;
|
2023-01-16 19:18:09 +00:00
|
|
|
folio = vma_alloc_zeroed_movable_folio(vma, vmf->address);
|
|
|
|
if (!folio)
|
remove ZERO_PAGE
The commit b5810039a54e5babf428e9a1e89fc1940fabff11 contains the note
A last caveat: the ZERO_PAGE is now refcounted and managed with rmap
(and thus mapcounted and count towards shared rss). These writes to
the struct page could cause excessive cacheline bouncing on big
systems. There are a number of ways this could be addressed if it is
an issue.
And indeed this cacheline bouncing has shown up on large SGI systems.
There was a situation where an Altix system was essentially livelocked
tearing down ZERO_PAGE pagetables when an HPC app aborted during startup.
This situation can be avoided in userspace, but it does highlight the
potential scalability problem with refcounting ZERO_PAGE, and corner
cases where it can really hurt (we don't want the system to livelock!).
There are several broad ways to fix this problem:
1. add back some special casing to avoid refcounting ZERO_PAGE
2. per-node or per-cpu ZERO_PAGES
3. remove the ZERO_PAGE completely
I will argue for 3. The others should also fix the problem, but they
result in more complex code than does 3, with little or no real benefit
that I can see.
Why? Inserting a ZERO_PAGE for anonymous read faults appears to be a
false optimisation: if an application is performance critical, it would
not be doing many read faults of new memory, or at least it could be
expected to write to that memory soon afterwards. If cache or memory use
is critical, it should not be working with a significant number of
ZERO_PAGEs anyway (a more compact representation of zeroes should be
used).
As a sanity check -- mesuring on my desktop system, there are never many
mappings to the ZERO_PAGE (eg. 2 or 3), thus memory usage here should not
increase much without it.
When running a make -j4 kernel compile on my dual core system, there are
about 1,000 mappings to the ZERO_PAGE created per second, but about 1,000
ZERO_PAGE COW faults per second (less than 1 ZERO_PAGE mapping per second
is torn down without being COWed). So removing ZERO_PAGE will save 1,000
page faults per second when running kbuild, while keeping it only saves
less than 1 page clearing operation per second. 1 page clear is cheaper
than a thousand faults, presumably, so there isn't an obvious loss.
Neither the logical argument nor these basic tests give a guarantee of no
regressions. However, this is a reasonable opportunity to try to remove
the ZERO_PAGE from the pagefault path. If it is found to cause regressions,
we can reintroduce it and just avoid refcounting it.
The /dev/zero ZERO_PAGE usage and TLB tricks also get nuked. I don't see
much use to them except on benchmarks. All other users of ZERO_PAGE are
converted just to use ZERO_PAGE(0) for simplicity. We can look at
replacing them all and maybe ripping out ZERO_PAGE completely when we are
more satisfied with this solution.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus "snif" Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:24:40 +00:00
|
|
|
goto oom;
|
2015-06-24 23:57:27 +00:00
|
|
|
|
2023-01-16 19:18:09 +00:00
|
|
|
if (mem_cgroup_charge(folio, vma->vm_mm, GFP_KERNEL))
|
2015-06-24 23:57:27 +00:00
|
|
|
goto oom_free_page;
|
2023-03-02 11:58:33 +00:00
|
|
|
folio_throttle_swaprate(folio, GFP_KERNEL);
|
2015-06-24 23:57:27 +00:00
|
|
|
|
2013-04-29 22:08:15 +00:00
|
|
|
/*
|
2023-01-16 19:18:10 +00:00
|
|
|
* The memory barrier inside __folio_mark_uptodate makes sure that
|
2019-12-01 01:58:17 +00:00
|
|
|
* preceding stores to the page contents become visible before
|
2013-04-29 22:08:15 +00:00
|
|
|
* the set_pte_at() write.
|
|
|
|
*/
|
2023-01-16 19:18:10 +00:00
|
|
|
__folio_mark_uptodate(folio);
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
|
2023-01-16 19:18:10 +00:00
|
|
|
entry = mk_pte(&folio->page, vma->vm_page_prot);
|
2021-06-05 03:01:08 +00:00
|
|
|
entry = pte_sw_mkyoung(entry);
|
2009-09-22 00:03:29 +00:00
|
|
|
if (vma->vm_flags & VM_WRITE)
|
|
|
|
entry = pte_mkwrite(pte_mkdirty(entry));
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
|
|
|
|
&vmf->ptl);
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
if (vmf_pte_changed(vmf)) {
|
2022-09-29 11:23:17 +00:00
|
|
|
update_mmu_tlb(vma, vmf->address, vmf->pte);
|
remove ZERO_PAGE
The commit b5810039a54e5babf428e9a1e89fc1940fabff11 contains the note
A last caveat: the ZERO_PAGE is now refcounted and managed with rmap
(and thus mapcounted and count towards shared rss). These writes to
the struct page could cause excessive cacheline bouncing on big
systems. There are a number of ways this could be addressed if it is
an issue.
And indeed this cacheline bouncing has shown up on large SGI systems.
There was a situation where an Altix system was essentially livelocked
tearing down ZERO_PAGE pagetables when an HPC app aborted during startup.
This situation can be avoided in userspace, but it does highlight the
potential scalability problem with refcounting ZERO_PAGE, and corner
cases where it can really hurt (we don't want the system to livelock!).
There are several broad ways to fix this problem:
1. add back some special casing to avoid refcounting ZERO_PAGE
2. per-node or per-cpu ZERO_PAGES
3. remove the ZERO_PAGE completely
I will argue for 3. The others should also fix the problem, but they
result in more complex code than does 3, with little or no real benefit
that I can see.
Why? Inserting a ZERO_PAGE for anonymous read faults appears to be a
false optimisation: if an application is performance critical, it would
not be doing many read faults of new memory, or at least it could be
expected to write to that memory soon afterwards. If cache or memory use
is critical, it should not be working with a significant number of
ZERO_PAGEs anyway (a more compact representation of zeroes should be
used).
As a sanity check -- mesuring on my desktop system, there are never many
mappings to the ZERO_PAGE (eg. 2 or 3), thus memory usage here should not
increase much without it.
When running a make -j4 kernel compile on my dual core system, there are
about 1,000 mappings to the ZERO_PAGE created per second, but about 1,000
ZERO_PAGE COW faults per second (less than 1 ZERO_PAGE mapping per second
is torn down without being COWed). So removing ZERO_PAGE will save 1,000
page faults per second when running kbuild, while keeping it only saves
less than 1 page clearing operation per second. 1 page clear is cheaper
than a thousand faults, presumably, so there isn't an obvious loss.
Neither the logical argument nor these basic tests give a guarantee of no
regressions. However, this is a reasonable opportunity to try to remove
the ZERO_PAGE from the pagefault path. If it is found to cause regressions,
we can reintroduce it and just avoid refcounting it.
The /dev/zero ZERO_PAGE usage and TLB tricks also get nuked. I don't see
much use to them except on benchmarks. All other users of ZERO_PAGE are
converted just to use ZERO_PAGE(0) for simplicity. We can look at
replacing them all and maybe ripping out ZERO_PAGE completely when we are
more satisfied with this solution.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus "snif" Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:24:40 +00:00
|
|
|
goto release;
|
2020-05-27 02:25:18 +00:00
|
|
|
}
|
ksm: fix oom deadlock
There's a now-obvious deadlock in KSM's out-of-memory handling:
imagine ksmd or KSM_RUN_UNMERGE handling, holding ksm_thread_mutex,
trying to allocate a page to break KSM in an mm which becomes the
OOM victim (quite likely in the unmerge case): it's killed and goes
to exit, and hangs there waiting to acquire ksm_thread_mutex.
Clearly we must not require ksm_thread_mutex in __ksm_exit, simple
though that made everything else: perhaps use mmap_sem somehow?
And part of the answer lies in the comments on unmerge_ksm_pages:
__ksm_exit should also leave all the rmap_item removal to ksmd.
But there's a fundamental problem, that KSM relies upon mmap_sem to
guarantee the consistency of the mm it's dealing with, yet exit_mmap
tears down an mm without taking mmap_sem. And bumping mm_users won't
help at all, that just ensures that the pages the OOM killer assumes
are on their way to being freed will not be freed.
The best answer seems to be, to move the ksm_exit callout from just
before exit_mmap, to the middle of exit_mmap: after the mm's pages
have been freed (if the mmu_gather is flushed), but before its page
tables and vma structures have been freed; and down_write,up_write
mmap_sem there to serialize with KSM's own reliance on mmap_sem.
But KSM then needs to be careful, whenever it downs mmap_sem, to
check that the mm is not already exiting: there's a danger of using
find_vma on a layout that's being torn apart, or writing into page
tables which have been freed for reuse; and even do_anonymous_page
and __do_fault need to check they're not being called by break_ksm
to reinstate a pte after zap_pte_range has zapped that page table.
Though it might be clearer to add an exiting flag, set while holding
mmap_sem in __ksm_exit, that wouldn't cover the issue of reinstating
a zapped pte. All we need is to check whether mm_users is 0 - but
must remember that ksmd may detect that before __ksm_exit is reached.
So, ksm_test_exit(mm) added to comment such checks on mm->mm_users.
__ksm_exit now has to leave clearing up the rmap_items to ksmd,
that needs ksm_thread_mutex; but shift the exiting mm just after the
ksm_scan cursor so that it will soon be dealt with. __ksm_enter raise
mm_count to hold the mm_struct, ksmd's exit processing (exactly like
its processing when it finds all VM_MERGEABLEs unmapped) mmdrop it,
similar procedure for KSM_RUN_UNMERGE (which has stopped ksmd).
But also give __ksm_exit a fast path: when there's no complication
(no rmap_items attached to mm and it's not at the ksm_scan cursor),
it can safely do all the exiting work itself. This is not just an
optimization: when ksmd is not running, the raised mm_count would
otherwise leak mm_structs.
Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Acked-by: Izik Eidus <ieidus@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-09-22 00:02:20 +00:00
|
|
|
|
2017-08-18 22:16:15 +00:00
|
|
|
ret = check_stable_address_space(vma->vm_mm);
|
|
|
|
if (ret)
|
|
|
|
goto release;
|
|
|
|
|
2015-09-04 22:46:20 +00:00
|
|
|
/* Deliver the page fault to userland, check inside PT lock */
|
|
|
|
if (userfaultfd_missing(vma)) {
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2023-01-16 19:18:10 +00:00
|
|
|
folio_put(folio);
|
2016-12-14 23:06:58 +00:00
|
|
|
return handle_userfault(vmf, VM_UFFD_MISSING);
|
2015-09-04 22:46:20 +00:00
|
|
|
}
|
|
|
|
|
2022-10-24 05:28:41 +00:00
|
|
|
inc_mm_counter(vma->vm_mm, MM_ANONPAGES);
|
2023-01-16 19:18:10 +00:00
|
|
|
folio_add_new_anon_rmap(folio, vma, vmf->address);
|
|
|
|
folio_add_lru_vma(folio, vma);
|
2009-09-22 00:03:30 +00:00
|
|
|
setpte:
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
if (uffd_wp)
|
|
|
|
entry = pte_mkuffd_wp(entry);
|
2016-12-14 23:06:58 +00:00
|
|
|
set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* No need to invalidate - it was non-present before */
|
2016-12-14 23:06:58 +00:00
|
|
|
update_mmu_cache(vma, vmf->address, vmf->pte);
|
[PATCH] mm: page fault handlers tidyup
Impose a little more consistency on the page fault handlers do_wp_page,
do_swap_page, do_anonymous_page, do_no_page, do_file_page: why not pass their
arguments in the same order, called the same names?
break_cow is all very well, but what it did was inlined elsewhere: easier to
compare if it's brought back into do_wp_page.
do_file_page's fallback to do_no_page dates from a time when we were testing
pte_file by using it wherever possible: currently it's peculiar to nonlinear
vmas, so just check that. BUG_ON if not? Better not, it's probably page
table corruption, so just show the pte: hmm, there's a pte_ERROR macro, let's
use that for do_wp_page's invalid pfn too.
Hah! Someone in the ppc64 world noticed pte_ERROR was unused so removed it:
restored (and say "pud" not "pmd" in its pud_ERROR).
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 01:15:59 +00:00
|
|
|
unlock:
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2017-08-18 22:16:15 +00:00
|
|
|
return ret;
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
release:
|
2023-01-16 19:18:10 +00:00
|
|
|
folio_put(folio);
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
goto unlock;
|
2008-02-07 08:13:53 +00:00
|
|
|
oom_free_page:
|
2023-01-16 19:18:10 +00:00
|
|
|
folio_put(folio);
|
[PATCH] mm: page fault handlers tidyup
Impose a little more consistency on the page fault handlers do_wp_page,
do_swap_page, do_anonymous_page, do_no_page, do_file_page: why not pass their
arguments in the same order, called the same names?
break_cow is all very well, but what it did was inlined elsewhere: easier to
compare if it's brought back into do_wp_page.
do_file_page's fallback to do_no_page dates from a time when we were testing
pte_file by using it wherever possible: currently it's peculiar to nonlinear
vmas, so just check that. BUG_ON if not? Better not, it's probably page
table corruption, so just show the pte: hmm, there's a pte_ERROR macro, let's
use that for do_wp_page's invalid pfn too.
Hah! Someone in the ppc64 world noticed pte_ERROR was unused so removed it:
restored (and say "pud" not "pmd" in its pud_ERROR).
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 01:15:59 +00:00
|
|
|
oom:
|
2005-04-16 22:20:36 +00:00
|
|
|
return VM_FAULT_OOM;
|
|
|
|
}
|
|
|
|
|
2014-08-06 23:07:24 +00:00
|
|
|
/*
|
2020-06-09 04:33:54 +00:00
|
|
|
* The mmap_lock must have been held on entry, and may have been
|
2014-08-06 23:07:24 +00:00
|
|
|
* released depending on flags and vma->vm_ops->fault() return value.
|
|
|
|
* See filemap_fault() and __lock_page_retry().
|
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t __do_fault(struct vm_fault *vmf)
|
2014-04-03 21:48:10 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret;
|
2014-04-03 21:48:10 +00:00
|
|
|
|
2019-01-08 23:23:07 +00:00
|
|
|
/*
|
|
|
|
* Preallocate pte before we take page_lock because this might lead to
|
|
|
|
* deadlocks for memcg reclaim which waits for pages under writeback:
|
|
|
|
* lock_page(A)
|
|
|
|
* SetPageWriteback(A)
|
|
|
|
* unlock_page(A)
|
|
|
|
* lock_page(B)
|
|
|
|
* lock_page(B)
|
2020-10-13 23:53:26 +00:00
|
|
|
* pte_alloc_one
|
2019-01-08 23:23:07 +00:00
|
|
|
* shrink_page_list
|
|
|
|
* wait_on_page_writeback(A)
|
|
|
|
* SetPageWriteback(B)
|
|
|
|
* unlock_page(B)
|
|
|
|
* # flush A, B to clear the writeback
|
|
|
|
*/
|
|
|
|
if (pmd_none(*vmf->pmd) && !vmf->prealloc_pte) {
|
2020-10-13 23:53:29 +00:00
|
|
|
vmf->prealloc_pte = pte_alloc_one(vma->vm_mm);
|
2019-01-08 23:23:07 +00:00
|
|
|
if (!vmf->prealloc_pte)
|
|
|
|
return VM_FAULT_OOM;
|
|
|
|
}
|
|
|
|
|
2017-02-24 22:56:41 +00:00
|
|
|
ret = vma->vm_ops->fault(vmf);
|
2016-12-14 23:07:18 +00:00
|
|
|
if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY |
|
2016-12-14 23:07:24 +00:00
|
|
|
VM_FAULT_DONE_COW)))
|
2016-05-12 16:29:19 +00:00
|
|
|
return ret;
|
2014-04-03 21:48:10 +00:00
|
|
|
|
2016-12-14 23:07:07 +00:00
|
|
|
if (unlikely(PageHWPoison(vmf->page))) {
|
2022-04-01 18:28:42 +00:00
|
|
|
struct page *page = vmf->page;
|
2022-03-22 21:44:09 +00:00
|
|
|
vm_fault_t poisonret = VM_FAULT_HWPOISON;
|
|
|
|
if (ret & VM_FAULT_LOCKED) {
|
2022-04-01 18:28:42 +00:00
|
|
|
if (page_mapped(page))
|
|
|
|
unmap_mapping_pages(page_mapping(page),
|
|
|
|
page->index, 1, false);
|
2022-03-22 21:44:09 +00:00
|
|
|
/* Retry if a clean page was removed from the cache. */
|
2022-04-01 18:28:42 +00:00
|
|
|
if (invalidate_inode_page(page))
|
|
|
|
poisonret = VM_FAULT_NOPAGE;
|
|
|
|
unlock_page(page);
|
2022-03-22 21:44:09 +00:00
|
|
|
}
|
2022-04-01 18:28:42 +00:00
|
|
|
put_page(page);
|
2016-12-14 23:07:10 +00:00
|
|
|
vmf->page = NULL;
|
2022-03-22 21:44:09 +00:00
|
|
|
return poisonret;
|
2014-04-03 21:48:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (unlikely(!(ret & VM_FAULT_LOCKED)))
|
2016-12-14 23:07:07 +00:00
|
|
|
lock_page(vmf->page);
|
2014-04-03 21:48:10 +00:00
|
|
|
else
|
2016-12-14 23:07:07 +00:00
|
|
|
VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
|
2014-04-03 21:48:10 +00:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2020-04-07 03:04:35 +00:00
|
|
|
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
|
2016-12-14 23:06:58 +00:00
|
|
|
static void deposit_prealloc_pte(struct vm_fault *vmf)
|
2016-12-13 00:44:32 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2016-12-13 00:44:32 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
|
2016-12-13 00:44:32 +00:00
|
|
|
/*
|
|
|
|
* We are going to consume the prealloc table,
|
|
|
|
* count that as nr_ptes.
|
|
|
|
*/
|
2017-11-16 01:35:37 +00:00
|
|
|
mm_inc_nr_ptes(vma->vm_mm);
|
2017-02-24 22:58:59 +00:00
|
|
|
vmf->prealloc_pte = NULL;
|
2016-12-13 00:44:32 +00:00
|
|
|
}
|
|
|
|
|
2020-12-19 12:19:23 +00:00
|
|
|
vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
|
2016-07-26 22:25:29 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
|
|
|
bool write = vmf->flags & FAULT_FLAG_WRITE;
|
|
|
|
unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
|
2016-07-26 22:25:29 +00:00
|
|
|
pmd_t entry;
|
2018-08-24 00:01:36 +00:00
|
|
|
int i;
|
2020-10-16 03:05:26 +00:00
|
|
|
vm_fault_t ret = VM_FAULT_FALLBACK;
|
2016-07-26 22:25:29 +00:00
|
|
|
|
|
|
|
if (!transhuge_vma_suitable(vma, haddr))
|
2020-10-16 03:05:26 +00:00
|
|
|
return ret;
|
2016-07-26 22:25:29 +00:00
|
|
|
|
|
|
|
page = compound_head(page);
|
2020-10-16 03:05:26 +00:00
|
|
|
if (compound_order(page) != HPAGE_PMD_ORDER)
|
|
|
|
return ret;
|
2016-07-26 22:25:29 +00:00
|
|
|
|
2021-10-28 21:36:11 +00:00
|
|
|
/*
|
|
|
|
* Just backoff if any subpage of a THP is corrupted otherwise
|
|
|
|
* the corrupted page may mapped by PMD silently to escape the
|
|
|
|
* check. This kind of THP just can be PTE mapped. Access to
|
|
|
|
* the corrupted subpage should trigger SIGBUS as expected.
|
|
|
|
*/
|
|
|
|
if (unlikely(PageHasHWPoisoned(page)))
|
|
|
|
return ret;
|
|
|
|
|
2016-12-13 00:44:32 +00:00
|
|
|
/*
|
2021-05-07 01:06:47 +00:00
|
|
|
* Archs like ppc64 need additional space to store information
|
2016-12-13 00:44:32 +00:00
|
|
|
* related to pte entry. Use the preallocated table for that.
|
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
|
mm: treewide: remove unused address argument from pte_alloc functions
Patch series "Add support for fast mremap".
This series speeds up the mremap(2) syscall by copying page tables at
the PMD level even for non-THP systems. There is concern that the extra
'address' argument that mremap passes to pte_alloc may do something
subtle architecture related in the future that may make the scheme not
work. Also we find that there is no point in passing the 'address' to
pte_alloc since its unused. This patch therefore removes this argument
tree-wide resulting in a nice negative diff as well. Also ensuring
along the way that the enabled architectures do not do anything funky
with the 'address' argument that goes unnoticed by the optimization.
Build and boot tested on x86-64. Build tested on arm64. The config
enablement patch for arm64 will be posted in the future after more
testing.
The changes were obtained by applying the following Coccinelle script.
(thanks Julia for answering all Coccinelle questions!).
Following fix ups were done manually:
* Removal of address argument from pte_fragment_alloc
* Removal of pte_alloc_one_fast definitions from m68k and microblaze.
// Options: --include-headers --no-includes
// Note: I split the 'identifier fn' line, so if you are manually
// running it, please unsplit it so it runs for you.
virtual patch
@pte_alloc_func_def depends on patch exists@
identifier E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
type T2;
@@
fn(...
- , T2 E2
)
{ ... }
@pte_alloc_func_proto_noarg depends on patch exists@
type T1, T2, T3, T4;
identifier fn =~ "^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1, T2);
+ T3 fn(T1);
|
- T3 fn(T1, T2, T4);
+ T3 fn(T1, T2);
)
@pte_alloc_func_proto depends on patch exists@
identifier E1, E2, E4;
type T1, T2, T3, T4;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1 E1, T2 E2);
+ T3 fn(T1 E1);
|
- T3 fn(T1 E1, T2 E2, T4 E4);
+ T3 fn(T1 E1, T2 E2);
)
@pte_alloc_func_call depends on patch exists@
expression E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
fn(...
-, E2
)
@pte_alloc_macro depends on patch exists@
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
identifier a, b, c;
expression e;
position p;
@@
(
- #define fn(a, b, c) e
+ #define fn(a, b) e
|
- #define fn(a, b) e
+ #define fn(a) e
)
Link: http://lkml.kernel.org/r/20181108181201.88826-2-joelaf@google.com
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Suggested-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: William Kucharski <william.kucharski@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-03 23:28:34 +00:00
|
|
|
vmf->prealloc_pte = pte_alloc_one(vma->vm_mm);
|
2016-12-14 23:06:58 +00:00
|
|
|
if (!vmf->prealloc_pte)
|
2016-12-13 00:44:32 +00:00
|
|
|
return VM_FAULT_OOM;
|
|
|
|
}
|
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
|
|
|
|
if (unlikely(!pmd_none(*vmf->pmd)))
|
2016-07-26 22:25:29 +00:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
for (i = 0; i < HPAGE_PMD_NR; i++)
|
|
|
|
flush_icache_page(vma, page + i);
|
|
|
|
|
|
|
|
entry = mk_huge_pmd(page, vma->vm_page_prot);
|
|
|
|
if (write)
|
2017-11-29 17:01:01 +00:00
|
|
|
entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);
|
2016-07-26 22:25:29 +00:00
|
|
|
|
2018-08-17 22:44:55 +00:00
|
|
|
add_mm_counter(vma->vm_mm, mm_counter_file(page), HPAGE_PMD_NR);
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
page_add_file_rmap(page, vma, true);
|
|
|
|
|
2016-12-13 00:44:32 +00:00
|
|
|
/*
|
|
|
|
* deposit and withdraw with pmd lock held
|
|
|
|
*/
|
|
|
|
if (arch_needs_pgtable_deposit())
|
2016-12-14 23:06:58 +00:00
|
|
|
deposit_prealloc_pte(vmf);
|
2016-07-26 22:25:29 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
|
2016-07-26 22:25:29 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
update_mmu_cache_pmd(vma, haddr, vmf->pmd);
|
2016-07-26 22:25:29 +00:00
|
|
|
|
|
|
|
/* fault is handled */
|
|
|
|
ret = 0;
|
2016-07-26 22:25:31 +00:00
|
|
|
count_vm_event(THP_FILE_MAPPED);
|
2016-07-26 22:25:29 +00:00
|
|
|
out:
|
2016-12-14 23:06:58 +00:00
|
|
|
spin_unlock(vmf->ptl);
|
2016-07-26 22:25:29 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
#else
|
2020-12-19 12:19:23 +00:00
|
|
|
vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
|
2016-07-26 22:25:29 +00:00
|
|
|
{
|
2020-12-19 12:19:23 +00:00
|
|
|
return VM_FAULT_FALLBACK;
|
2016-07-26 22:25:29 +00:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2021-01-14 15:24:19 +00:00
|
|
|
void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr)
|
2014-04-03 21:48:16 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
bool uffd_wp = vmf_orig_pte_uffd_wp(vmf);
|
2016-12-14 23:06:58 +00:00
|
|
|
bool write = vmf->flags & FAULT_FLAG_WRITE;
|
2021-01-14 15:24:19 +00:00
|
|
|
bool prefault = vmf->address != addr;
|
2014-04-03 21:48:16 +00:00
|
|
|
pte_t entry;
|
2016-07-26 22:25:23 +00:00
|
|
|
|
2014-04-03 21:48:16 +00:00
|
|
|
flush_icache_page(vma, page);
|
|
|
|
entry = mk_pte(page, vma->vm_page_prot);
|
2020-11-24 18:48:26 +00:00
|
|
|
|
|
|
|
if (prefault && arch_wants_old_prefaulted_pte())
|
|
|
|
entry = pte_mkold(entry);
|
2021-06-05 03:01:08 +00:00
|
|
|
else
|
|
|
|
entry = pte_sw_mkyoung(entry);
|
2020-11-24 18:48:26 +00:00
|
|
|
|
2014-04-03 21:48:16 +00:00
|
|
|
if (write)
|
|
|
|
entry = maybe_mkwrite(pte_mkdirty(entry), vma);
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
if (unlikely(uffd_wp))
|
2022-12-14 20:15:33 +00:00
|
|
|
entry = pte_mkuffd_wp(entry);
|
2016-07-26 22:25:20 +00:00
|
|
|
/* copy-on-write page */
|
|
|
|
if (write && !(vma->vm_flags & VM_SHARED)) {
|
2022-10-24 05:28:41 +00:00
|
|
|
inc_mm_counter(vma->vm_mm, MM_ANONPAGES);
|
2022-05-10 01:20:43 +00:00
|
|
|
page_add_new_anon_rmap(page, vma, addr);
|
2020-08-12 01:30:40 +00:00
|
|
|
lru_cache_add_inactive_or_unevictable(page, vma);
|
2014-04-03 21:48:16 +00:00
|
|
|
} else {
|
2022-10-24 05:28:41 +00:00
|
|
|
inc_mm_counter(vma->vm_mm, mm_counter_file(page));
|
mm/munlock: rmap call mlock_vma_page() munlock_vma_page()
Add vma argument to mlock_vma_page() and munlock_vma_page(), make them
inline functions which check (vma->vm_flags & VM_LOCKED) before calling
mlock_page() and munlock_page() in mm/mlock.c.
Add bool compound to mlock_vma_page() and munlock_vma_page(): this is
because we have understandable difficulty in accounting pte maps of THPs,
and if passed a PageHead page, mlock_page() and munlock_page() cannot
tell whether it's a pmd map to be counted or a pte map to be ignored.
Add vma arg to page_add_file_rmap() and page_remove_rmap(), like the
others, and use that to call mlock_vma_page() at the end of the page
adds, and munlock_vma_page() at the end of page_remove_rmap() (end or
beginning? unimportant, but end was easier for assertions in testing).
No page lock is required (although almost all adds happen to hold it):
delete the "Serialize with page migration" BUG_ON(!PageLocked(page))s.
Certainly page lock did serialize with page migration, but I'm having
difficulty explaining why that was ever important.
Mlock accounting on THPs has been hard to define, differed between anon
and file, involved PageDoubleMap in some places and not others, required
clear_page_mlock() at some points. Keep it simple now: just count the
pmds and ignore the ptes, there is no reason for ptes to undo pmd mlocks.
page_add_new_anon_rmap() callers unchanged: they have long been calling
lru_cache_add_inactive_or_unevictable(), which does its own VM_LOCKED
handling (it also checks for not VM_SPECIAL: I think that's overcautious,
and inconsistent with other checks, that mmap_region() already prevents
VM_LOCKED on VM_SPECIAL; but haven't quite convinced myself to change it).
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2022-02-15 02:26:39 +00:00
|
|
|
page_add_file_rmap(page, vma, false);
|
2014-04-03 21:48:16 +00:00
|
|
|
}
|
2021-01-14 15:24:19 +00:00
|
|
|
set_pte_at(vma->vm_mm, addr, vmf->pte, entry);
|
2014-04-03 21:48:16 +00:00
|
|
|
}
|
|
|
|
|
2022-05-13 03:22:52 +00:00
|
|
|
static bool vmf_pte_changed(struct vm_fault *vmf)
|
|
|
|
{
|
|
|
|
if (vmf->flags & FAULT_FLAG_ORIG_PTE_VALID)
|
|
|
|
return !pte_same(*vmf->pte, vmf->orig_pte);
|
|
|
|
|
|
|
|
return !pte_none(*vmf->pte);
|
|
|
|
}
|
|
|
|
|
2016-12-14 23:07:21 +00:00
|
|
|
/**
|
|
|
|
* finish_fault - finish page fault once we have prepared the page to fault
|
|
|
|
*
|
|
|
|
* @vmf: structure describing the fault
|
|
|
|
*
|
|
|
|
* This function handles all that is needed to finish a page fault once the
|
|
|
|
* page to fault in is prepared. It handles locking of PTEs, inserts PTE for
|
|
|
|
* given page, adds reverse page mapping, handles memcg charges and LRU
|
2019-03-05 23:48:42 +00:00
|
|
|
* addition.
|
2016-12-14 23:07:21 +00:00
|
|
|
*
|
|
|
|
* The function expects the page to be locked and on success it consumes a
|
|
|
|
* reference of a page being mapped (for the PTE which maps it).
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: %0 on success, %VM_FAULT_ code in case of error.
|
2016-12-14 23:07:21 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t finish_fault(struct vm_fault *vmf)
|
2016-12-14 23:07:21 +00:00
|
|
|
{
|
2020-12-19 12:19:23 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2016-12-14 23:07:21 +00:00
|
|
|
struct page *page;
|
2020-12-19 12:19:23 +00:00
|
|
|
vm_fault_t ret;
|
2016-12-14 23:07:21 +00:00
|
|
|
|
|
|
|
/* Did we COW the page? */
|
2020-12-19 12:19:23 +00:00
|
|
|
if ((vmf->flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
|
2016-12-14 23:07:21 +00:00
|
|
|
page = vmf->cow_page;
|
|
|
|
else
|
|
|
|
page = vmf->page;
|
2017-08-18 22:16:15 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* check even for read faults because we might have lost our CoWed
|
|
|
|
* page
|
|
|
|
*/
|
2020-12-19 12:19:23 +00:00
|
|
|
if (!(vma->vm_flags & VM_SHARED)) {
|
|
|
|
ret = check_stable_address_space(vma->vm_mm);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (pmd_none(*vmf->pmd)) {
|
|
|
|
if (PageTransCompound(page)) {
|
|
|
|
ret = do_set_pmd(vmf, page);
|
|
|
|
if (ret != VM_FAULT_FALLBACK)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2021-11-05 20:38:38 +00:00
|
|
|
if (vmf->prealloc_pte)
|
|
|
|
pmd_install(vma->vm_mm, vmf->pmd, &vmf->prealloc_pte);
|
|
|
|
else if (unlikely(pte_alloc(vma->vm_mm, vmf->pmd)))
|
2020-12-19 12:19:23 +00:00
|
|
|
return VM_FAULT_OOM;
|
|
|
|
}
|
|
|
|
|
mm: fix page leak with multiple threads mapping the same page
We have an application with a lot of threads that use a shared mmap backed
by tmpfs mounted with -o huge=within_size. This application started
leaking loads of huge pages when we upgraded to a recent kernel.
Using the page ref tracepoints and a BPF program written by Tejun Heo we
were able to determine that these pages would have multiple refcounts from
the page fault path, but when it came to unmap time we wouldn't drop the
number of refs we had added from the faults.
I wrote a reproducer that mmap'ed a file backed by tmpfs with -o
huge=always, and then spawned 20 threads all looping faulting random
offsets in this map, while using madvise(MADV_DONTNEED) randomly for huge
page aligned ranges. This very quickly reproduced the problem.
The problem here is that we check for the case that we have multiple
threads faulting in a range that was previously unmapped. One thread maps
the PMD, the other thread loses the race and then returns 0. However at
this point we already have the page, and we are no longer putting this
page into the processes address space, and so we leak the page. We
actually did the correct thing prior to f9ce0be71d1f, however it looks
like Kirill copied what we do in the anonymous page case. In the
anonymous page case we don't yet have a page, so we don't have to drop a
reference on anything. Previously we did the correct thing for file based
faults by returning VM_FAULT_NOPAGE so we correctly drop the reference on
the page we faulted in.
Fix this by returning VM_FAULT_NOPAGE in the pmd_devmap_trans_unstable()
case, this makes us drop the ref on the page properly, and now my
reproducer no longer leaks the huge pages.
[josef@toxicpanda.com: v2]
Link: https://lkml.kernel.org/r/e90c8f0dbae836632b669c2afc434006a00d4a67.1657721478.git.josef@toxicpanda.com
Link: https://lkml.kernel.org/r/2b798acfd95c9ab9395fe85e8d5a835e2e10a920.1657051137.git.josef@toxicpanda.com
Fixes: f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Rik van Riel <riel@surriel.com>
Signed-off-by: Chris Mason <clm@fb.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-07-05 20:00:36 +00:00
|
|
|
/*
|
|
|
|
* See comment in handle_pte_fault() for how this scenario happens, we
|
|
|
|
* need to return NOPAGE so that we drop this page.
|
|
|
|
*/
|
2020-12-19 12:19:23 +00:00
|
|
|
if (pmd_devmap_trans_unstable(vmf->pmd))
|
mm: fix page leak with multiple threads mapping the same page
We have an application with a lot of threads that use a shared mmap backed
by tmpfs mounted with -o huge=within_size. This application started
leaking loads of huge pages when we upgraded to a recent kernel.
Using the page ref tracepoints and a BPF program written by Tejun Heo we
were able to determine that these pages would have multiple refcounts from
the page fault path, but when it came to unmap time we wouldn't drop the
number of refs we had added from the faults.
I wrote a reproducer that mmap'ed a file backed by tmpfs with -o
huge=always, and then spawned 20 threads all looping faulting random
offsets in this map, while using madvise(MADV_DONTNEED) randomly for huge
page aligned ranges. This very quickly reproduced the problem.
The problem here is that we check for the case that we have multiple
threads faulting in a range that was previously unmapped. One thread maps
the PMD, the other thread loses the race and then returns 0. However at
this point we already have the page, and we are no longer putting this
page into the processes address space, and so we leak the page. We
actually did the correct thing prior to f9ce0be71d1f, however it looks
like Kirill copied what we do in the anonymous page case. In the
anonymous page case we don't yet have a page, so we don't have to drop a
reference on anything. Previously we did the correct thing for file based
faults by returning VM_FAULT_NOPAGE so we correctly drop the reference on
the page we faulted in.
Fix this by returning VM_FAULT_NOPAGE in the pmd_devmap_trans_unstable()
case, this makes us drop the ref on the page properly, and now my
reproducer no longer leaks the huge pages.
[josef@toxicpanda.com: v2]
Link: https://lkml.kernel.org/r/e90c8f0dbae836632b669c2afc434006a00d4a67.1657721478.git.josef@toxicpanda.com
Link: https://lkml.kernel.org/r/2b798acfd95c9ab9395fe85e8d5a835e2e10a920.1657051137.git.josef@toxicpanda.com
Fixes: f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Rik van Riel <riel@surriel.com>
Signed-off-by: Chris Mason <clm@fb.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-07-05 20:00:36 +00:00
|
|
|
return VM_FAULT_NOPAGE;
|
2020-12-19 12:19:23 +00:00
|
|
|
|
|
|
|
vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
|
|
|
|
vmf->address, &vmf->ptl);
|
mm: bring back update_mmu_cache() to finish_fault()
Running this test program on ARMv4 a few times (sometimes just once)
reproduces the bug.
int main()
{
unsigned i;
char paragon[SIZE];
void* ptr;
memset(paragon, 0xAA, SIZE);
ptr = mmap(NULL, SIZE, PROT_READ | PROT_WRITE,
MAP_ANON | MAP_SHARED, -1, 0);
if (ptr == MAP_FAILED) return 1;
printf("ptr = %p\n", ptr);
for (i=0;i<10000;i++){
memset(ptr, 0xAA, SIZE);
if (memcmp(ptr, paragon, SIZE)) {
printf("Unexpected bytes on iteration %u!!!\n", i);
break;
}
}
munmap(ptr, SIZE);
}
In the "ptr" buffer there appear runs of zero bytes which are aligned
by 16 and their lengths are multiple of 16.
Linux v5.11 does not have the bug, "git bisect" finds the first bad commit:
f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Before the commit update_mmu_cache() was called during a call to
filemap_map_pages() as well as finish_fault(). After the commit
finish_fault() lacks it.
Bring back update_mmu_cache() to finish_fault() to fix the bug.
Also call update_mmu_tlb() only when returning VM_FAULT_NOPAGE to more
closely reproduce the code of alloc_set_pte() function that existed before
the commit.
On many platforms update_mmu_cache() is nop:
x86, see arch/x86/include/asm/pgtable
ARMv6+, see arch/arm/include/asm/tlbflush.h
So, it seems, few users ran into this bug.
Link: https://lkml.kernel.org/r/20220908204809.2012451-1-saproj@gmail.com
Fixes: f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Signed-off-by: Sergei Antonov <saproj@gmail.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Will Deacon <will@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-08 20:48:09 +00:00
|
|
|
|
2020-12-19 12:19:23 +00:00
|
|
|
/* Re-check under ptl */
|
mm: bring back update_mmu_cache() to finish_fault()
Running this test program on ARMv4 a few times (sometimes just once)
reproduces the bug.
int main()
{
unsigned i;
char paragon[SIZE];
void* ptr;
memset(paragon, 0xAA, SIZE);
ptr = mmap(NULL, SIZE, PROT_READ | PROT_WRITE,
MAP_ANON | MAP_SHARED, -1, 0);
if (ptr == MAP_FAILED) return 1;
printf("ptr = %p\n", ptr);
for (i=0;i<10000;i++){
memset(ptr, 0xAA, SIZE);
if (memcmp(ptr, paragon, SIZE)) {
printf("Unexpected bytes on iteration %u!!!\n", i);
break;
}
}
munmap(ptr, SIZE);
}
In the "ptr" buffer there appear runs of zero bytes which are aligned
by 16 and their lengths are multiple of 16.
Linux v5.11 does not have the bug, "git bisect" finds the first bad commit:
f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Before the commit update_mmu_cache() was called during a call to
filemap_map_pages() as well as finish_fault(). After the commit
finish_fault() lacks it.
Bring back update_mmu_cache() to finish_fault() to fix the bug.
Also call update_mmu_tlb() only when returning VM_FAULT_NOPAGE to more
closely reproduce the code of alloc_set_pte() function that existed before
the commit.
On many platforms update_mmu_cache() is nop:
x86, see arch/x86/include/asm/pgtable
ARMv6+, see arch/arm/include/asm/tlbflush.h
So, it seems, few users ran into this bug.
Link: https://lkml.kernel.org/r/20220908204809.2012451-1-saproj@gmail.com
Fixes: f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Signed-off-by: Sergei Antonov <saproj@gmail.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Will Deacon <will@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-08 20:48:09 +00:00
|
|
|
if (likely(!vmf_pte_changed(vmf))) {
|
2021-01-14 15:24:19 +00:00
|
|
|
do_set_pte(vmf, page, vmf->address);
|
mm: bring back update_mmu_cache() to finish_fault()
Running this test program on ARMv4 a few times (sometimes just once)
reproduces the bug.
int main()
{
unsigned i;
char paragon[SIZE];
void* ptr;
memset(paragon, 0xAA, SIZE);
ptr = mmap(NULL, SIZE, PROT_READ | PROT_WRITE,
MAP_ANON | MAP_SHARED, -1, 0);
if (ptr == MAP_FAILED) return 1;
printf("ptr = %p\n", ptr);
for (i=0;i<10000;i++){
memset(ptr, 0xAA, SIZE);
if (memcmp(ptr, paragon, SIZE)) {
printf("Unexpected bytes on iteration %u!!!\n", i);
break;
}
}
munmap(ptr, SIZE);
}
In the "ptr" buffer there appear runs of zero bytes which are aligned
by 16 and their lengths are multiple of 16.
Linux v5.11 does not have the bug, "git bisect" finds the first bad commit:
f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Before the commit update_mmu_cache() was called during a call to
filemap_map_pages() as well as finish_fault(). After the commit
finish_fault() lacks it.
Bring back update_mmu_cache() to finish_fault() to fix the bug.
Also call update_mmu_tlb() only when returning VM_FAULT_NOPAGE to more
closely reproduce the code of alloc_set_pte() function that existed before
the commit.
On many platforms update_mmu_cache() is nop:
x86, see arch/x86/include/asm/pgtable
ARMv6+, see arch/arm/include/asm/tlbflush.h
So, it seems, few users ran into this bug.
Link: https://lkml.kernel.org/r/20220908204809.2012451-1-saproj@gmail.com
Fixes: f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Signed-off-by: Sergei Antonov <saproj@gmail.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Will Deacon <will@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-08 20:48:09 +00:00
|
|
|
|
|
|
|
/* no need to invalidate: a not-present page won't be cached */
|
|
|
|
update_mmu_cache(vma, vmf->address, vmf->pte);
|
|
|
|
|
|
|
|
ret = 0;
|
|
|
|
} else {
|
|
|
|
update_mmu_tlb(vma, vmf->address, vmf->pte);
|
2020-12-19 12:19:23 +00:00
|
|
|
ret = VM_FAULT_NOPAGE;
|
mm: bring back update_mmu_cache() to finish_fault()
Running this test program on ARMv4 a few times (sometimes just once)
reproduces the bug.
int main()
{
unsigned i;
char paragon[SIZE];
void* ptr;
memset(paragon, 0xAA, SIZE);
ptr = mmap(NULL, SIZE, PROT_READ | PROT_WRITE,
MAP_ANON | MAP_SHARED, -1, 0);
if (ptr == MAP_FAILED) return 1;
printf("ptr = %p\n", ptr);
for (i=0;i<10000;i++){
memset(ptr, 0xAA, SIZE);
if (memcmp(ptr, paragon, SIZE)) {
printf("Unexpected bytes on iteration %u!!!\n", i);
break;
}
}
munmap(ptr, SIZE);
}
In the "ptr" buffer there appear runs of zero bytes which are aligned
by 16 and their lengths are multiple of 16.
Linux v5.11 does not have the bug, "git bisect" finds the first bad commit:
f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Before the commit update_mmu_cache() was called during a call to
filemap_map_pages() as well as finish_fault(). After the commit
finish_fault() lacks it.
Bring back update_mmu_cache() to finish_fault() to fix the bug.
Also call update_mmu_tlb() only when returning VM_FAULT_NOPAGE to more
closely reproduce the code of alloc_set_pte() function that existed before
the commit.
On many platforms update_mmu_cache() is nop:
x86, see arch/x86/include/asm/pgtable
ARMv6+, see arch/arm/include/asm/tlbflush.h
So, it seems, few users ran into this bug.
Link: https://lkml.kernel.org/r/20220908204809.2012451-1-saproj@gmail.com
Fixes: f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths")
Signed-off-by: Sergei Antonov <saproj@gmail.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Will Deacon <will@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-08 20:48:09 +00:00
|
|
|
}
|
2020-12-19 12:19:23 +00:00
|
|
|
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2016-12-14 23:07:21 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2023-03-17 21:58:26 +00:00
|
|
|
static unsigned long fault_around_pages __read_mostly =
|
|
|
|
65536 >> PAGE_SHIFT;
|
2014-06-04 23:10:54 +00:00
|
|
|
|
|
|
|
#ifdef CONFIG_DEBUG_FS
|
|
|
|
static int fault_around_bytes_get(void *data, u64 *val)
|
2014-04-07 22:37:22 +00:00
|
|
|
{
|
2023-03-17 21:58:26 +00:00
|
|
|
*val = fault_around_pages << PAGE_SHIFT;
|
2014-04-07 22:37:22 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-07-30 23:08:35 +00:00
|
|
|
/*
|
2018-02-01 00:21:11 +00:00
|
|
|
* fault_around_bytes must be rounded down to the nearest page order as it's
|
|
|
|
* what do_fault_around() expects to see.
|
2014-07-30 23:08:35 +00:00
|
|
|
*/
|
2014-06-04 23:10:54 +00:00
|
|
|
static int fault_around_bytes_set(void *data, u64 val)
|
2014-04-07 22:37:22 +00:00
|
|
|
{
|
2014-06-04 23:10:54 +00:00
|
|
|
if (val / PAGE_SIZE > PTRS_PER_PTE)
|
2014-04-07 22:37:22 +00:00
|
|
|
return -EINVAL;
|
2023-03-17 21:58:26 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The minimum value is 1 page, however this results in no fault-around
|
|
|
|
* at all. See should_fault_around().
|
|
|
|
*/
|
|
|
|
fault_around_pages = max(rounddown_pow_of_two(val) >> PAGE_SHIFT, 1UL);
|
|
|
|
|
2014-04-07 22:37:22 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2017-07-10 22:47:17 +00:00
|
|
|
DEFINE_DEBUGFS_ATTRIBUTE(fault_around_bytes_fops,
|
2014-06-04 23:10:54 +00:00
|
|
|
fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
|
2014-04-07 22:37:22 +00:00
|
|
|
|
|
|
|
static int __init fault_around_debugfs(void)
|
|
|
|
{
|
2019-03-05 23:46:09 +00:00
|
|
|
debugfs_create_file_unsafe("fault_around_bytes", 0644, NULL, NULL,
|
|
|
|
&fault_around_bytes_fops);
|
2014-04-07 22:37:22 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
late_initcall(fault_around_debugfs);
|
|
|
|
#endif
|
mm: introduce vm_ops->map_pages()
Here's new version of faultaround patchset. It took a while to tune it
and collect performance data.
First patch adds new callback ->map_pages to vm_operations_struct.
->map_pages() is called when VM asks to map easy accessible pages.
Filesystem should find and map pages associated with offsets from
"pgoff" till "max_pgoff". ->map_pages() is called with page table
locked and must not block. If it's not possible to reach a page without
blocking, filesystem should skip it. Filesystem should use do_set_pte()
to setup page table entry. Pointer to entry associated with offset
"pgoff" is passed in "pte" field in vm_fault structure. Pointers to
entries for other offsets should be calculated relative to "pte".
Currently VM use ->map_pages only on read page fault path. We try to
map FAULT_AROUND_PAGES a time. FAULT_AROUND_PAGES is 16 for now.
Performance data for different FAULT_AROUND_ORDER is below.
TODO:
- implement ->map_pages() for shmem/tmpfs;
- modify get_user_pages() to be able to use ->map_pages() and implement
mmap(MAP_POPULATE|MAP_NONBLOCK) on top.
=========================================================================
Tested on 4-socket machine (120 threads) with 128GiB of RAM.
Few real-world workloads. The sweet spot for FAULT_AROUND_ORDER here is
somewhere between 3 and 5. Let's say 4 :)
Linux build (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 283,301,572 247,151,987 212,215,789 204,772,882 199,568,944 194,703,779 193,381,485
time, seconds 151.227629483 153.920996480 151.356125472 150.863792049 150.879207877 151.150764954 151.450962358
Linux rebuild (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 5,396,854 4,148,444 2,855,286 2,577,282 2,361,957 2,169,573 2,112,643
time, seconds 27.404543757 27.559725591 27.030057426 26.855045126 26.678618635 26.974523490 26.761320095
Git test suite (make -j60 test)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 129,591,823 99,200,751 66,106,718 57,606,410 51,510,808 45,776,813 44,085,515
time, seconds 66.087215026 64.784546905 64.401156567 65.282708668 66.034016829 66.793780811 67.237810413
Two synthetic tests: access every word in file in sequential/random order.
It doesn't improve much after FAULT_AROUND_ORDER == 4.
Sequential access 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 4,195,437 2,098,275 525,068 262,251 131,170 32,856 8,282
time, seconds 7.250461742 6.461711074 5.493859139 5.488488147 5.707213983 5.898510832 5.109232856
8 threads
minor-faults 33,557,540 16,892,728 4,515,848 2,366,999 1,423,382 442,732 142,339
time, seconds 16.649304881 9.312555263 6.612490639 6.394316732 6.669827501 6.75078944 6.371900528
32 threads
minor-faults 134,228,222 67,526,810 17,725,386 9,716,537 4,763,731 1,668,921 537,200
time, seconds 49.164430543 29.712060103 12.938649729 10.175151004 11.840094583 9.594081325 9.928461797
60 threads
minor-faults 251,687,988 126,146,952 32,919,406 18,208,804 10,458,947 2,733,907 928,217
time, seconds 86.260656897 49.626551828 22.335007632 17.608243696 16.523119035 16.339489186 16.326390902
120 threads
minor-faults 503,352,863 252,939,677 67,039,168 35,191,827 19,170,091 4,688,357 1,471,862
time, seconds 124.589206333 79.757867787 39.508707872 32.167281632 29.972989292 28.729834575 28.042251622
Random access 1GiB file
1 thread
minor-faults 262,636 132,743 34,369 17,299 8,527 3,451 1,222
time, seconds 15.351890914 16.613802482 16.569227308 15.179220992 16.557356122 16.578247824 15.365266994
8 threads
minor-faults 2,098,948 1,061,871 273,690 154,501 87,110 25,663 7,384
time, seconds 15.040026343 15.096933500 14.474757288 14.289129964 14.411537468 14.296316837 14.395635804
32 threads
minor-faults 8,390,734 4,231,023 1,054,432 528,847 269,242 97,746 26,881
time, seconds 20.430433109 21.585235358 22.115062928 14.872878951 14.880856305 14.883370649 14.821261690
60 threads
minor-faults 15,733,258 7,892,809 1,973,393 988,266 594,789 164,994 51,691
time, seconds 26.577302548 25.692397770 18.728863715 20.153026398 21.619101933 17.745086260 17.613215273
120 threads
minor-faults 31,471,111 15,816,616 3,959,209 1,978,685 1,008,299 264,635 96,010
time, seconds 41.835322703 40.459786095 36.085306105 35.313894834 35.814445675 36.552633793 34.289210594
Touch only one page in page table in 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 8,372 8,324 8,270 8,260 8,249 8,239 8,237
time, seconds 0.039892712 0.045369149 0.051846126 0.063681685 0.079095975 0.17652406 0.541213386
8 threads
minor-faults 65,731 65,681 65,628 65,620 65,608 65,599 65,596
time, seconds 0.124159196 0.488600638 0.156854426 0.191901957 0.242631486 0.543569456 1.677303984
32 threads
minor-faults 262,388 262,341 262,285 262,276 262,266 262,257 263,183
time, seconds 0.452421421 0.488600638 0.565020946 0.648229739 0.789850823 1.651584361 5.000361559
60 threads
minor-faults 491,822 491,792 491,723 491,711 491,701 491,691 491,825
time, seconds 0.763288616 0.869620515 0.980727360 1.161732354 1.466915814 3.04041448 9.308612938
120 threads
minor-faults 983,466 983,655 983,366 983,372 983,363 984,083 984,164
time, seconds 1.595846553 1.667902182 2.008959376 2.425380942 2.941368804 5.977807890 18.401846125
This patch (of 2):
Introduce new vm_ops callback ->map_pages() and uses it for mapping easy
accessible pages around fault address.
On read page fault, if filesystem provides ->map_pages(), we try to map up
to FAULT_AROUND_PAGES pages around page fault address in hope to reduce
number of minor page faults.
We call ->map_pages first and use ->fault() as fallback if page by the
offset is not ready to be mapped (cold page cache or something).
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ning Qu <quning@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:18 +00:00
|
|
|
|
2014-06-04 23:10:55 +00:00
|
|
|
/*
|
|
|
|
* do_fault_around() tries to map few pages around the fault address. The hope
|
|
|
|
* is that the pages will be needed soon and this will lower the number of
|
|
|
|
* faults to handle.
|
|
|
|
*
|
|
|
|
* It uses vm_ops->map_pages() to map the pages, which skips the page if it's
|
|
|
|
* not ready to be mapped: not up-to-date, locked, etc.
|
|
|
|
*
|
2023-03-17 21:58:25 +00:00
|
|
|
* This function doesn't cross VMA or page table boundaries, in order to call
|
|
|
|
* map_pages() and acquire a PTE lock only once.
|
2014-06-04 23:10:55 +00:00
|
|
|
*
|
2023-03-17 21:58:26 +00:00
|
|
|
* fault_around_pages defines how many pages we'll try to map.
|
2018-02-01 00:21:11 +00:00
|
|
|
* do_fault_around() expects it to be set to a power of two less than or equal
|
|
|
|
* to PTRS_PER_PTE.
|
2014-06-04 23:10:55 +00:00
|
|
|
*
|
2018-02-01 00:21:11 +00:00
|
|
|
* The virtual address of the area that we map is naturally aligned to
|
2023-03-17 21:58:26 +00:00
|
|
|
* fault_around_pages * PAGE_SIZE rounded down to the machine page size
|
2018-02-01 00:21:11 +00:00
|
|
|
* (and therefore to page order). This way it's easier to guarantee
|
|
|
|
* that we don't cross page table boundaries.
|
2014-06-04 23:10:55 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_fault_around(struct vm_fault *vmf)
|
mm: introduce vm_ops->map_pages()
Here's new version of faultaround patchset. It took a while to tune it
and collect performance data.
First patch adds new callback ->map_pages to vm_operations_struct.
->map_pages() is called when VM asks to map easy accessible pages.
Filesystem should find and map pages associated with offsets from
"pgoff" till "max_pgoff". ->map_pages() is called with page table
locked and must not block. If it's not possible to reach a page without
blocking, filesystem should skip it. Filesystem should use do_set_pte()
to setup page table entry. Pointer to entry associated with offset
"pgoff" is passed in "pte" field in vm_fault structure. Pointers to
entries for other offsets should be calculated relative to "pte".
Currently VM use ->map_pages only on read page fault path. We try to
map FAULT_AROUND_PAGES a time. FAULT_AROUND_PAGES is 16 for now.
Performance data for different FAULT_AROUND_ORDER is below.
TODO:
- implement ->map_pages() for shmem/tmpfs;
- modify get_user_pages() to be able to use ->map_pages() and implement
mmap(MAP_POPULATE|MAP_NONBLOCK) on top.
=========================================================================
Tested on 4-socket machine (120 threads) with 128GiB of RAM.
Few real-world workloads. The sweet spot for FAULT_AROUND_ORDER here is
somewhere between 3 and 5. Let's say 4 :)
Linux build (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 283,301,572 247,151,987 212,215,789 204,772,882 199,568,944 194,703,779 193,381,485
time, seconds 151.227629483 153.920996480 151.356125472 150.863792049 150.879207877 151.150764954 151.450962358
Linux rebuild (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 5,396,854 4,148,444 2,855,286 2,577,282 2,361,957 2,169,573 2,112,643
time, seconds 27.404543757 27.559725591 27.030057426 26.855045126 26.678618635 26.974523490 26.761320095
Git test suite (make -j60 test)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 129,591,823 99,200,751 66,106,718 57,606,410 51,510,808 45,776,813 44,085,515
time, seconds 66.087215026 64.784546905 64.401156567 65.282708668 66.034016829 66.793780811 67.237810413
Two synthetic tests: access every word in file in sequential/random order.
It doesn't improve much after FAULT_AROUND_ORDER == 4.
Sequential access 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 4,195,437 2,098,275 525,068 262,251 131,170 32,856 8,282
time, seconds 7.250461742 6.461711074 5.493859139 5.488488147 5.707213983 5.898510832 5.109232856
8 threads
minor-faults 33,557,540 16,892,728 4,515,848 2,366,999 1,423,382 442,732 142,339
time, seconds 16.649304881 9.312555263 6.612490639 6.394316732 6.669827501 6.75078944 6.371900528
32 threads
minor-faults 134,228,222 67,526,810 17,725,386 9,716,537 4,763,731 1,668,921 537,200
time, seconds 49.164430543 29.712060103 12.938649729 10.175151004 11.840094583 9.594081325 9.928461797
60 threads
minor-faults 251,687,988 126,146,952 32,919,406 18,208,804 10,458,947 2,733,907 928,217
time, seconds 86.260656897 49.626551828 22.335007632 17.608243696 16.523119035 16.339489186 16.326390902
120 threads
minor-faults 503,352,863 252,939,677 67,039,168 35,191,827 19,170,091 4,688,357 1,471,862
time, seconds 124.589206333 79.757867787 39.508707872 32.167281632 29.972989292 28.729834575 28.042251622
Random access 1GiB file
1 thread
minor-faults 262,636 132,743 34,369 17,299 8,527 3,451 1,222
time, seconds 15.351890914 16.613802482 16.569227308 15.179220992 16.557356122 16.578247824 15.365266994
8 threads
minor-faults 2,098,948 1,061,871 273,690 154,501 87,110 25,663 7,384
time, seconds 15.040026343 15.096933500 14.474757288 14.289129964 14.411537468 14.296316837 14.395635804
32 threads
minor-faults 8,390,734 4,231,023 1,054,432 528,847 269,242 97,746 26,881
time, seconds 20.430433109 21.585235358 22.115062928 14.872878951 14.880856305 14.883370649 14.821261690
60 threads
minor-faults 15,733,258 7,892,809 1,973,393 988,266 594,789 164,994 51,691
time, seconds 26.577302548 25.692397770 18.728863715 20.153026398 21.619101933 17.745086260 17.613215273
120 threads
minor-faults 31,471,111 15,816,616 3,959,209 1,978,685 1,008,299 264,635 96,010
time, seconds 41.835322703 40.459786095 36.085306105 35.313894834 35.814445675 36.552633793 34.289210594
Touch only one page in page table in 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 8,372 8,324 8,270 8,260 8,249 8,239 8,237
time, seconds 0.039892712 0.045369149 0.051846126 0.063681685 0.079095975 0.17652406 0.541213386
8 threads
minor-faults 65,731 65,681 65,628 65,620 65,608 65,599 65,596
time, seconds 0.124159196 0.488600638 0.156854426 0.191901957 0.242631486 0.543569456 1.677303984
32 threads
minor-faults 262,388 262,341 262,285 262,276 262,266 262,257 263,183
time, seconds 0.452421421 0.488600638 0.565020946 0.648229739 0.789850823 1.651584361 5.000361559
60 threads
minor-faults 491,822 491,792 491,723 491,711 491,701 491,691 491,825
time, seconds 0.763288616 0.869620515 0.980727360 1.161732354 1.466915814 3.04041448 9.308612938
120 threads
minor-faults 983,466 983,655 983,366 983,372 983,363 984,083 984,164
time, seconds 1.595846553 1.667902182 2.008959376 2.425380942 2.941368804 5.977807890 18.401846125
This patch (of 2):
Introduce new vm_ops callback ->map_pages() and uses it for mapping easy
accessible pages around fault address.
On read page fault, if filesystem provides ->map_pages(), we try to map up
to FAULT_AROUND_PAGES pages around page fault address in hope to reduce
number of minor page faults.
We call ->map_pages first and use ->fault() as fallback if page by the
offset is not ready to be mapped (cold page cache or something).
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ning Qu <quning@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:18 +00:00
|
|
|
{
|
2023-03-17 21:58:26 +00:00
|
|
|
pgoff_t nr_pages = READ_ONCE(fault_around_pages);
|
2023-03-17 21:58:25 +00:00
|
|
|
pgoff_t pte_off = pte_index(vmf->address);
|
|
|
|
/* The page offset of vmf->address within the VMA. */
|
|
|
|
pgoff_t vma_off = vmf->pgoff - vmf->vma->vm_pgoff;
|
|
|
|
pgoff_t from_pte, to_pte;
|
2023-03-27 17:45:15 +00:00
|
|
|
vm_fault_t ret;
|
mm: introduce vm_ops->map_pages()
Here's new version of faultaround patchset. It took a while to tune it
and collect performance data.
First patch adds new callback ->map_pages to vm_operations_struct.
->map_pages() is called when VM asks to map easy accessible pages.
Filesystem should find and map pages associated with offsets from
"pgoff" till "max_pgoff". ->map_pages() is called with page table
locked and must not block. If it's not possible to reach a page without
blocking, filesystem should skip it. Filesystem should use do_set_pte()
to setup page table entry. Pointer to entry associated with offset
"pgoff" is passed in "pte" field in vm_fault structure. Pointers to
entries for other offsets should be calculated relative to "pte".
Currently VM use ->map_pages only on read page fault path. We try to
map FAULT_AROUND_PAGES a time. FAULT_AROUND_PAGES is 16 for now.
Performance data for different FAULT_AROUND_ORDER is below.
TODO:
- implement ->map_pages() for shmem/tmpfs;
- modify get_user_pages() to be able to use ->map_pages() and implement
mmap(MAP_POPULATE|MAP_NONBLOCK) on top.
=========================================================================
Tested on 4-socket machine (120 threads) with 128GiB of RAM.
Few real-world workloads. The sweet spot for FAULT_AROUND_ORDER here is
somewhere between 3 and 5. Let's say 4 :)
Linux build (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 283,301,572 247,151,987 212,215,789 204,772,882 199,568,944 194,703,779 193,381,485
time, seconds 151.227629483 153.920996480 151.356125472 150.863792049 150.879207877 151.150764954 151.450962358
Linux rebuild (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 5,396,854 4,148,444 2,855,286 2,577,282 2,361,957 2,169,573 2,112,643
time, seconds 27.404543757 27.559725591 27.030057426 26.855045126 26.678618635 26.974523490 26.761320095
Git test suite (make -j60 test)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 129,591,823 99,200,751 66,106,718 57,606,410 51,510,808 45,776,813 44,085,515
time, seconds 66.087215026 64.784546905 64.401156567 65.282708668 66.034016829 66.793780811 67.237810413
Two synthetic tests: access every word in file in sequential/random order.
It doesn't improve much after FAULT_AROUND_ORDER == 4.
Sequential access 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 4,195,437 2,098,275 525,068 262,251 131,170 32,856 8,282
time, seconds 7.250461742 6.461711074 5.493859139 5.488488147 5.707213983 5.898510832 5.109232856
8 threads
minor-faults 33,557,540 16,892,728 4,515,848 2,366,999 1,423,382 442,732 142,339
time, seconds 16.649304881 9.312555263 6.612490639 6.394316732 6.669827501 6.75078944 6.371900528
32 threads
minor-faults 134,228,222 67,526,810 17,725,386 9,716,537 4,763,731 1,668,921 537,200
time, seconds 49.164430543 29.712060103 12.938649729 10.175151004 11.840094583 9.594081325 9.928461797
60 threads
minor-faults 251,687,988 126,146,952 32,919,406 18,208,804 10,458,947 2,733,907 928,217
time, seconds 86.260656897 49.626551828 22.335007632 17.608243696 16.523119035 16.339489186 16.326390902
120 threads
minor-faults 503,352,863 252,939,677 67,039,168 35,191,827 19,170,091 4,688,357 1,471,862
time, seconds 124.589206333 79.757867787 39.508707872 32.167281632 29.972989292 28.729834575 28.042251622
Random access 1GiB file
1 thread
minor-faults 262,636 132,743 34,369 17,299 8,527 3,451 1,222
time, seconds 15.351890914 16.613802482 16.569227308 15.179220992 16.557356122 16.578247824 15.365266994
8 threads
minor-faults 2,098,948 1,061,871 273,690 154,501 87,110 25,663 7,384
time, seconds 15.040026343 15.096933500 14.474757288 14.289129964 14.411537468 14.296316837 14.395635804
32 threads
minor-faults 8,390,734 4,231,023 1,054,432 528,847 269,242 97,746 26,881
time, seconds 20.430433109 21.585235358 22.115062928 14.872878951 14.880856305 14.883370649 14.821261690
60 threads
minor-faults 15,733,258 7,892,809 1,973,393 988,266 594,789 164,994 51,691
time, seconds 26.577302548 25.692397770 18.728863715 20.153026398 21.619101933 17.745086260 17.613215273
120 threads
minor-faults 31,471,111 15,816,616 3,959,209 1,978,685 1,008,299 264,635 96,010
time, seconds 41.835322703 40.459786095 36.085306105 35.313894834 35.814445675 36.552633793 34.289210594
Touch only one page in page table in 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 8,372 8,324 8,270 8,260 8,249 8,239 8,237
time, seconds 0.039892712 0.045369149 0.051846126 0.063681685 0.079095975 0.17652406 0.541213386
8 threads
minor-faults 65,731 65,681 65,628 65,620 65,608 65,599 65,596
time, seconds 0.124159196 0.488600638 0.156854426 0.191901957 0.242631486 0.543569456 1.677303984
32 threads
minor-faults 262,388 262,341 262,285 262,276 262,266 262,257 263,183
time, seconds 0.452421421 0.488600638 0.565020946 0.648229739 0.789850823 1.651584361 5.000361559
60 threads
minor-faults 491,822 491,792 491,723 491,711 491,701 491,691 491,825
time, seconds 0.763288616 0.869620515 0.980727360 1.161732354 1.466915814 3.04041448 9.308612938
120 threads
minor-faults 983,466 983,655 983,366 983,372 983,363 984,083 984,164
time, seconds 1.595846553 1.667902182 2.008959376 2.425380942 2.941368804 5.977807890 18.401846125
This patch (of 2):
Introduce new vm_ops callback ->map_pages() and uses it for mapping easy
accessible pages around fault address.
On read page fault, if filesystem provides ->map_pages(), we try to map up
to FAULT_AROUND_PAGES pages around page fault address in hope to reduce
number of minor page faults.
We call ->map_pages first and use ->fault() as fallback if page by the
offset is not ready to be mapped (cold page cache or something).
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ning Qu <quning@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:18 +00:00
|
|
|
|
2023-03-17 21:58:25 +00:00
|
|
|
/* The PTE offset of the start address, clamped to the VMA. */
|
|
|
|
from_pte = max(ALIGN_DOWN(pte_off, nr_pages),
|
|
|
|
pte_off - min(pte_off, vma_off));
|
2014-08-06 23:08:05 +00:00
|
|
|
|
2023-03-17 21:58:25 +00:00
|
|
|
/* The PTE offset of the end address, clamped to the VMA and PTE. */
|
|
|
|
to_pte = min3(from_pte + nr_pages, (pgoff_t)PTRS_PER_PTE,
|
|
|
|
pte_off + vma_pages(vmf->vma) - vma_off) - 1;
|
mm: introduce vm_ops->map_pages()
Here's new version of faultaround patchset. It took a while to tune it
and collect performance data.
First patch adds new callback ->map_pages to vm_operations_struct.
->map_pages() is called when VM asks to map easy accessible pages.
Filesystem should find and map pages associated with offsets from
"pgoff" till "max_pgoff". ->map_pages() is called with page table
locked and must not block. If it's not possible to reach a page without
blocking, filesystem should skip it. Filesystem should use do_set_pte()
to setup page table entry. Pointer to entry associated with offset
"pgoff" is passed in "pte" field in vm_fault structure. Pointers to
entries for other offsets should be calculated relative to "pte".
Currently VM use ->map_pages only on read page fault path. We try to
map FAULT_AROUND_PAGES a time. FAULT_AROUND_PAGES is 16 for now.
Performance data for different FAULT_AROUND_ORDER is below.
TODO:
- implement ->map_pages() for shmem/tmpfs;
- modify get_user_pages() to be able to use ->map_pages() and implement
mmap(MAP_POPULATE|MAP_NONBLOCK) on top.
=========================================================================
Tested on 4-socket machine (120 threads) with 128GiB of RAM.
Few real-world workloads. The sweet spot for FAULT_AROUND_ORDER here is
somewhere between 3 and 5. Let's say 4 :)
Linux build (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 283,301,572 247,151,987 212,215,789 204,772,882 199,568,944 194,703,779 193,381,485
time, seconds 151.227629483 153.920996480 151.356125472 150.863792049 150.879207877 151.150764954 151.450962358
Linux rebuild (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 5,396,854 4,148,444 2,855,286 2,577,282 2,361,957 2,169,573 2,112,643
time, seconds 27.404543757 27.559725591 27.030057426 26.855045126 26.678618635 26.974523490 26.761320095
Git test suite (make -j60 test)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 129,591,823 99,200,751 66,106,718 57,606,410 51,510,808 45,776,813 44,085,515
time, seconds 66.087215026 64.784546905 64.401156567 65.282708668 66.034016829 66.793780811 67.237810413
Two synthetic tests: access every word in file in sequential/random order.
It doesn't improve much after FAULT_AROUND_ORDER == 4.
Sequential access 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 4,195,437 2,098,275 525,068 262,251 131,170 32,856 8,282
time, seconds 7.250461742 6.461711074 5.493859139 5.488488147 5.707213983 5.898510832 5.109232856
8 threads
minor-faults 33,557,540 16,892,728 4,515,848 2,366,999 1,423,382 442,732 142,339
time, seconds 16.649304881 9.312555263 6.612490639 6.394316732 6.669827501 6.75078944 6.371900528
32 threads
minor-faults 134,228,222 67,526,810 17,725,386 9,716,537 4,763,731 1,668,921 537,200
time, seconds 49.164430543 29.712060103 12.938649729 10.175151004 11.840094583 9.594081325 9.928461797
60 threads
minor-faults 251,687,988 126,146,952 32,919,406 18,208,804 10,458,947 2,733,907 928,217
time, seconds 86.260656897 49.626551828 22.335007632 17.608243696 16.523119035 16.339489186 16.326390902
120 threads
minor-faults 503,352,863 252,939,677 67,039,168 35,191,827 19,170,091 4,688,357 1,471,862
time, seconds 124.589206333 79.757867787 39.508707872 32.167281632 29.972989292 28.729834575 28.042251622
Random access 1GiB file
1 thread
minor-faults 262,636 132,743 34,369 17,299 8,527 3,451 1,222
time, seconds 15.351890914 16.613802482 16.569227308 15.179220992 16.557356122 16.578247824 15.365266994
8 threads
minor-faults 2,098,948 1,061,871 273,690 154,501 87,110 25,663 7,384
time, seconds 15.040026343 15.096933500 14.474757288 14.289129964 14.411537468 14.296316837 14.395635804
32 threads
minor-faults 8,390,734 4,231,023 1,054,432 528,847 269,242 97,746 26,881
time, seconds 20.430433109 21.585235358 22.115062928 14.872878951 14.880856305 14.883370649 14.821261690
60 threads
minor-faults 15,733,258 7,892,809 1,973,393 988,266 594,789 164,994 51,691
time, seconds 26.577302548 25.692397770 18.728863715 20.153026398 21.619101933 17.745086260 17.613215273
120 threads
minor-faults 31,471,111 15,816,616 3,959,209 1,978,685 1,008,299 264,635 96,010
time, seconds 41.835322703 40.459786095 36.085306105 35.313894834 35.814445675 36.552633793 34.289210594
Touch only one page in page table in 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 8,372 8,324 8,270 8,260 8,249 8,239 8,237
time, seconds 0.039892712 0.045369149 0.051846126 0.063681685 0.079095975 0.17652406 0.541213386
8 threads
minor-faults 65,731 65,681 65,628 65,620 65,608 65,599 65,596
time, seconds 0.124159196 0.488600638 0.156854426 0.191901957 0.242631486 0.543569456 1.677303984
32 threads
minor-faults 262,388 262,341 262,285 262,276 262,266 262,257 263,183
time, seconds 0.452421421 0.488600638 0.565020946 0.648229739 0.789850823 1.651584361 5.000361559
60 threads
minor-faults 491,822 491,792 491,723 491,711 491,701 491,691 491,825
time, seconds 0.763288616 0.869620515 0.980727360 1.161732354 1.466915814 3.04041448 9.308612938
120 threads
minor-faults 983,466 983,655 983,366 983,372 983,363 984,083 984,164
time, seconds 1.595846553 1.667902182 2.008959376 2.425380942 2.941368804 5.977807890 18.401846125
This patch (of 2):
Introduce new vm_ops callback ->map_pages() and uses it for mapping easy
accessible pages around fault address.
On read page fault, if filesystem provides ->map_pages(), we try to map up
to FAULT_AROUND_PAGES pages around page fault address in hope to reduce
number of minor page faults.
We call ->map_pages first and use ->fault() as fallback if page by the
offset is not ready to be mapped (cold page cache or something).
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ning Qu <quning@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:18 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
if (pmd_none(*vmf->pmd)) {
|
mm: treewide: remove unused address argument from pte_alloc functions
Patch series "Add support for fast mremap".
This series speeds up the mremap(2) syscall by copying page tables at
the PMD level even for non-THP systems. There is concern that the extra
'address' argument that mremap passes to pte_alloc may do something
subtle architecture related in the future that may make the scheme not
work. Also we find that there is no point in passing the 'address' to
pte_alloc since its unused. This patch therefore removes this argument
tree-wide resulting in a nice negative diff as well. Also ensuring
along the way that the enabled architectures do not do anything funky
with the 'address' argument that goes unnoticed by the optimization.
Build and boot tested on x86-64. Build tested on arm64. The config
enablement patch for arm64 will be posted in the future after more
testing.
The changes were obtained by applying the following Coccinelle script.
(thanks Julia for answering all Coccinelle questions!).
Following fix ups were done manually:
* Removal of address argument from pte_fragment_alloc
* Removal of pte_alloc_one_fast definitions from m68k and microblaze.
// Options: --include-headers --no-includes
// Note: I split the 'identifier fn' line, so if you are manually
// running it, please unsplit it so it runs for you.
virtual patch
@pte_alloc_func_def depends on patch exists@
identifier E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
type T2;
@@
fn(...
- , T2 E2
)
{ ... }
@pte_alloc_func_proto_noarg depends on patch exists@
type T1, T2, T3, T4;
identifier fn =~ "^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1, T2);
+ T3 fn(T1);
|
- T3 fn(T1, T2, T4);
+ T3 fn(T1, T2);
)
@pte_alloc_func_proto depends on patch exists@
identifier E1, E2, E4;
type T1, T2, T3, T4;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
(
- T3 fn(T1 E1, T2 E2);
+ T3 fn(T1 E1);
|
- T3 fn(T1 E1, T2 E2, T4 E4);
+ T3 fn(T1 E1, T2 E2);
)
@pte_alloc_func_call depends on patch exists@
expression E2;
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
@@
fn(...
-, E2
)
@pte_alloc_macro depends on patch exists@
identifier fn =~
"^(__pte_alloc|pte_alloc_one|pte_alloc|__pte_alloc_kernel|pte_alloc_one_kernel)$";
identifier a, b, c;
expression e;
position p;
@@
(
- #define fn(a, b, c) e
+ #define fn(a, b) e
|
- #define fn(a, b) e
+ #define fn(a) e
)
Link: http://lkml.kernel.org/r/20181108181201.88826-2-joelaf@google.com
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Suggested-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: William Kucharski <william.kucharski@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-01-03 23:28:34 +00:00
|
|
|
vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm);
|
2016-12-14 23:06:58 +00:00
|
|
|
if (!vmf->prealloc_pte)
|
2020-12-19 12:19:23 +00:00
|
|
|
return VM_FAULT_OOM;
|
mm: introduce vm_ops->map_pages()
Here's new version of faultaround patchset. It took a while to tune it
and collect performance data.
First patch adds new callback ->map_pages to vm_operations_struct.
->map_pages() is called when VM asks to map easy accessible pages.
Filesystem should find and map pages associated with offsets from
"pgoff" till "max_pgoff". ->map_pages() is called with page table
locked and must not block. If it's not possible to reach a page without
blocking, filesystem should skip it. Filesystem should use do_set_pte()
to setup page table entry. Pointer to entry associated with offset
"pgoff" is passed in "pte" field in vm_fault structure. Pointers to
entries for other offsets should be calculated relative to "pte".
Currently VM use ->map_pages only on read page fault path. We try to
map FAULT_AROUND_PAGES a time. FAULT_AROUND_PAGES is 16 for now.
Performance data for different FAULT_AROUND_ORDER is below.
TODO:
- implement ->map_pages() for shmem/tmpfs;
- modify get_user_pages() to be able to use ->map_pages() and implement
mmap(MAP_POPULATE|MAP_NONBLOCK) on top.
=========================================================================
Tested on 4-socket machine (120 threads) with 128GiB of RAM.
Few real-world workloads. The sweet spot for FAULT_AROUND_ORDER here is
somewhere between 3 and 5. Let's say 4 :)
Linux build (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 283,301,572 247,151,987 212,215,789 204,772,882 199,568,944 194,703,779 193,381,485
time, seconds 151.227629483 153.920996480 151.356125472 150.863792049 150.879207877 151.150764954 151.450962358
Linux rebuild (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 5,396,854 4,148,444 2,855,286 2,577,282 2,361,957 2,169,573 2,112,643
time, seconds 27.404543757 27.559725591 27.030057426 26.855045126 26.678618635 26.974523490 26.761320095
Git test suite (make -j60 test)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 129,591,823 99,200,751 66,106,718 57,606,410 51,510,808 45,776,813 44,085,515
time, seconds 66.087215026 64.784546905 64.401156567 65.282708668 66.034016829 66.793780811 67.237810413
Two synthetic tests: access every word in file in sequential/random order.
It doesn't improve much after FAULT_AROUND_ORDER == 4.
Sequential access 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 4,195,437 2,098,275 525,068 262,251 131,170 32,856 8,282
time, seconds 7.250461742 6.461711074 5.493859139 5.488488147 5.707213983 5.898510832 5.109232856
8 threads
minor-faults 33,557,540 16,892,728 4,515,848 2,366,999 1,423,382 442,732 142,339
time, seconds 16.649304881 9.312555263 6.612490639 6.394316732 6.669827501 6.75078944 6.371900528
32 threads
minor-faults 134,228,222 67,526,810 17,725,386 9,716,537 4,763,731 1,668,921 537,200
time, seconds 49.164430543 29.712060103 12.938649729 10.175151004 11.840094583 9.594081325 9.928461797
60 threads
minor-faults 251,687,988 126,146,952 32,919,406 18,208,804 10,458,947 2,733,907 928,217
time, seconds 86.260656897 49.626551828 22.335007632 17.608243696 16.523119035 16.339489186 16.326390902
120 threads
minor-faults 503,352,863 252,939,677 67,039,168 35,191,827 19,170,091 4,688,357 1,471,862
time, seconds 124.589206333 79.757867787 39.508707872 32.167281632 29.972989292 28.729834575 28.042251622
Random access 1GiB file
1 thread
minor-faults 262,636 132,743 34,369 17,299 8,527 3,451 1,222
time, seconds 15.351890914 16.613802482 16.569227308 15.179220992 16.557356122 16.578247824 15.365266994
8 threads
minor-faults 2,098,948 1,061,871 273,690 154,501 87,110 25,663 7,384
time, seconds 15.040026343 15.096933500 14.474757288 14.289129964 14.411537468 14.296316837 14.395635804
32 threads
minor-faults 8,390,734 4,231,023 1,054,432 528,847 269,242 97,746 26,881
time, seconds 20.430433109 21.585235358 22.115062928 14.872878951 14.880856305 14.883370649 14.821261690
60 threads
minor-faults 15,733,258 7,892,809 1,973,393 988,266 594,789 164,994 51,691
time, seconds 26.577302548 25.692397770 18.728863715 20.153026398 21.619101933 17.745086260 17.613215273
120 threads
minor-faults 31,471,111 15,816,616 3,959,209 1,978,685 1,008,299 264,635 96,010
time, seconds 41.835322703 40.459786095 36.085306105 35.313894834 35.814445675 36.552633793 34.289210594
Touch only one page in page table in 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 8,372 8,324 8,270 8,260 8,249 8,239 8,237
time, seconds 0.039892712 0.045369149 0.051846126 0.063681685 0.079095975 0.17652406 0.541213386
8 threads
minor-faults 65,731 65,681 65,628 65,620 65,608 65,599 65,596
time, seconds 0.124159196 0.488600638 0.156854426 0.191901957 0.242631486 0.543569456 1.677303984
32 threads
minor-faults 262,388 262,341 262,285 262,276 262,266 262,257 263,183
time, seconds 0.452421421 0.488600638 0.565020946 0.648229739 0.789850823 1.651584361 5.000361559
60 threads
minor-faults 491,822 491,792 491,723 491,711 491,701 491,691 491,825
time, seconds 0.763288616 0.869620515 0.980727360 1.161732354 1.466915814 3.04041448 9.308612938
120 threads
minor-faults 983,466 983,655 983,366 983,372 983,363 984,083 984,164
time, seconds 1.595846553 1.667902182 2.008959376 2.425380942 2.941368804 5.977807890 18.401846125
This patch (of 2):
Introduce new vm_ops callback ->map_pages() and uses it for mapping easy
accessible pages around fault address.
On read page fault, if filesystem provides ->map_pages(), we try to map up
to FAULT_AROUND_PAGES pages around page fault address in hope to reduce
number of minor page faults.
We call ->map_pages first and use ->fault() as fallback if page by the
offset is not ready to be mapped (cold page cache or something).
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ning Qu <quning@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:18 +00:00
|
|
|
}
|
|
|
|
|
2023-03-27 17:45:15 +00:00
|
|
|
rcu_read_lock();
|
|
|
|
ret = vmf->vma->vm_ops->map_pages(vmf,
|
|
|
|
vmf->pgoff + from_pte - pte_off,
|
|
|
|
vmf->pgoff + to_pte - pte_off);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return ret;
|
mm: introduce vm_ops->map_pages()
Here's new version of faultaround patchset. It took a while to tune it
and collect performance data.
First patch adds new callback ->map_pages to vm_operations_struct.
->map_pages() is called when VM asks to map easy accessible pages.
Filesystem should find and map pages associated with offsets from
"pgoff" till "max_pgoff". ->map_pages() is called with page table
locked and must not block. If it's not possible to reach a page without
blocking, filesystem should skip it. Filesystem should use do_set_pte()
to setup page table entry. Pointer to entry associated with offset
"pgoff" is passed in "pte" field in vm_fault structure. Pointers to
entries for other offsets should be calculated relative to "pte".
Currently VM use ->map_pages only on read page fault path. We try to
map FAULT_AROUND_PAGES a time. FAULT_AROUND_PAGES is 16 for now.
Performance data for different FAULT_AROUND_ORDER is below.
TODO:
- implement ->map_pages() for shmem/tmpfs;
- modify get_user_pages() to be able to use ->map_pages() and implement
mmap(MAP_POPULATE|MAP_NONBLOCK) on top.
=========================================================================
Tested on 4-socket machine (120 threads) with 128GiB of RAM.
Few real-world workloads. The sweet spot for FAULT_AROUND_ORDER here is
somewhere between 3 and 5. Let's say 4 :)
Linux build (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 283,301,572 247,151,987 212,215,789 204,772,882 199,568,944 194,703,779 193,381,485
time, seconds 151.227629483 153.920996480 151.356125472 150.863792049 150.879207877 151.150764954 151.450962358
Linux rebuild (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 5,396,854 4,148,444 2,855,286 2,577,282 2,361,957 2,169,573 2,112,643
time, seconds 27.404543757 27.559725591 27.030057426 26.855045126 26.678618635 26.974523490 26.761320095
Git test suite (make -j60 test)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 129,591,823 99,200,751 66,106,718 57,606,410 51,510,808 45,776,813 44,085,515
time, seconds 66.087215026 64.784546905 64.401156567 65.282708668 66.034016829 66.793780811 67.237810413
Two synthetic tests: access every word in file in sequential/random order.
It doesn't improve much after FAULT_AROUND_ORDER == 4.
Sequential access 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 4,195,437 2,098,275 525,068 262,251 131,170 32,856 8,282
time, seconds 7.250461742 6.461711074 5.493859139 5.488488147 5.707213983 5.898510832 5.109232856
8 threads
minor-faults 33,557,540 16,892,728 4,515,848 2,366,999 1,423,382 442,732 142,339
time, seconds 16.649304881 9.312555263 6.612490639 6.394316732 6.669827501 6.75078944 6.371900528
32 threads
minor-faults 134,228,222 67,526,810 17,725,386 9,716,537 4,763,731 1,668,921 537,200
time, seconds 49.164430543 29.712060103 12.938649729 10.175151004 11.840094583 9.594081325 9.928461797
60 threads
minor-faults 251,687,988 126,146,952 32,919,406 18,208,804 10,458,947 2,733,907 928,217
time, seconds 86.260656897 49.626551828 22.335007632 17.608243696 16.523119035 16.339489186 16.326390902
120 threads
minor-faults 503,352,863 252,939,677 67,039,168 35,191,827 19,170,091 4,688,357 1,471,862
time, seconds 124.589206333 79.757867787 39.508707872 32.167281632 29.972989292 28.729834575 28.042251622
Random access 1GiB file
1 thread
minor-faults 262,636 132,743 34,369 17,299 8,527 3,451 1,222
time, seconds 15.351890914 16.613802482 16.569227308 15.179220992 16.557356122 16.578247824 15.365266994
8 threads
minor-faults 2,098,948 1,061,871 273,690 154,501 87,110 25,663 7,384
time, seconds 15.040026343 15.096933500 14.474757288 14.289129964 14.411537468 14.296316837 14.395635804
32 threads
minor-faults 8,390,734 4,231,023 1,054,432 528,847 269,242 97,746 26,881
time, seconds 20.430433109 21.585235358 22.115062928 14.872878951 14.880856305 14.883370649 14.821261690
60 threads
minor-faults 15,733,258 7,892,809 1,973,393 988,266 594,789 164,994 51,691
time, seconds 26.577302548 25.692397770 18.728863715 20.153026398 21.619101933 17.745086260 17.613215273
120 threads
minor-faults 31,471,111 15,816,616 3,959,209 1,978,685 1,008,299 264,635 96,010
time, seconds 41.835322703 40.459786095 36.085306105 35.313894834 35.814445675 36.552633793 34.289210594
Touch only one page in page table in 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 8,372 8,324 8,270 8,260 8,249 8,239 8,237
time, seconds 0.039892712 0.045369149 0.051846126 0.063681685 0.079095975 0.17652406 0.541213386
8 threads
minor-faults 65,731 65,681 65,628 65,620 65,608 65,599 65,596
time, seconds 0.124159196 0.488600638 0.156854426 0.191901957 0.242631486 0.543569456 1.677303984
32 threads
minor-faults 262,388 262,341 262,285 262,276 262,266 262,257 263,183
time, seconds 0.452421421 0.488600638 0.565020946 0.648229739 0.789850823 1.651584361 5.000361559
60 threads
minor-faults 491,822 491,792 491,723 491,711 491,701 491,691 491,825
time, seconds 0.763288616 0.869620515 0.980727360 1.161732354 1.466915814 3.04041448 9.308612938
120 threads
minor-faults 983,466 983,655 983,366 983,372 983,363 984,083 984,164
time, seconds 1.595846553 1.667902182 2.008959376 2.425380942 2.941368804 5.977807890 18.401846125
This patch (of 2):
Introduce new vm_ops callback ->map_pages() and uses it for mapping easy
accessible pages around fault address.
On read page fault, if filesystem provides ->map_pages(), we try to map up
to FAULT_AROUND_PAGES pages around page fault address in hope to reduce
number of minor page faults.
We call ->map_pages first and use ->fault() as fallback if page by the
offset is not ready to be mapped (cold page cache or something).
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ning Qu <quning@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:18 +00:00
|
|
|
}
|
|
|
|
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
/* Return true if we should do read fault-around, false otherwise */
|
|
|
|
static inline bool should_fault_around(struct vm_fault *vmf)
|
|
|
|
{
|
|
|
|
/* No ->map_pages? No way to fault around... */
|
|
|
|
if (!vmf->vma->vm_ops->map_pages)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (uffd_disable_fault_around(vmf->vma))
|
|
|
|
return false;
|
|
|
|
|
2023-03-17 21:58:26 +00:00
|
|
|
/* A single page implies no faulting 'around' at all. */
|
|
|
|
return fault_around_pages > 1;
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
}
|
|
|
|
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_read_fault(struct vm_fault *vmf)
|
2014-04-03 21:48:11 +00:00
|
|
|
{
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret = 0;
|
mm: introduce vm_ops->map_pages()
Here's new version of faultaround patchset. It took a while to tune it
and collect performance data.
First patch adds new callback ->map_pages to vm_operations_struct.
->map_pages() is called when VM asks to map easy accessible pages.
Filesystem should find and map pages associated with offsets from
"pgoff" till "max_pgoff". ->map_pages() is called with page table
locked and must not block. If it's not possible to reach a page without
blocking, filesystem should skip it. Filesystem should use do_set_pte()
to setup page table entry. Pointer to entry associated with offset
"pgoff" is passed in "pte" field in vm_fault structure. Pointers to
entries for other offsets should be calculated relative to "pte".
Currently VM use ->map_pages only on read page fault path. We try to
map FAULT_AROUND_PAGES a time. FAULT_AROUND_PAGES is 16 for now.
Performance data for different FAULT_AROUND_ORDER is below.
TODO:
- implement ->map_pages() for shmem/tmpfs;
- modify get_user_pages() to be able to use ->map_pages() and implement
mmap(MAP_POPULATE|MAP_NONBLOCK) on top.
=========================================================================
Tested on 4-socket machine (120 threads) with 128GiB of RAM.
Few real-world workloads. The sweet spot for FAULT_AROUND_ORDER here is
somewhere between 3 and 5. Let's say 4 :)
Linux build (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 283,301,572 247,151,987 212,215,789 204,772,882 199,568,944 194,703,779 193,381,485
time, seconds 151.227629483 153.920996480 151.356125472 150.863792049 150.879207877 151.150764954 151.450962358
Linux rebuild (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 5,396,854 4,148,444 2,855,286 2,577,282 2,361,957 2,169,573 2,112,643
time, seconds 27.404543757 27.559725591 27.030057426 26.855045126 26.678618635 26.974523490 26.761320095
Git test suite (make -j60 test)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 129,591,823 99,200,751 66,106,718 57,606,410 51,510,808 45,776,813 44,085,515
time, seconds 66.087215026 64.784546905 64.401156567 65.282708668 66.034016829 66.793780811 67.237810413
Two synthetic tests: access every word in file in sequential/random order.
It doesn't improve much after FAULT_AROUND_ORDER == 4.
Sequential access 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 4,195,437 2,098,275 525,068 262,251 131,170 32,856 8,282
time, seconds 7.250461742 6.461711074 5.493859139 5.488488147 5.707213983 5.898510832 5.109232856
8 threads
minor-faults 33,557,540 16,892,728 4,515,848 2,366,999 1,423,382 442,732 142,339
time, seconds 16.649304881 9.312555263 6.612490639 6.394316732 6.669827501 6.75078944 6.371900528
32 threads
minor-faults 134,228,222 67,526,810 17,725,386 9,716,537 4,763,731 1,668,921 537,200
time, seconds 49.164430543 29.712060103 12.938649729 10.175151004 11.840094583 9.594081325 9.928461797
60 threads
minor-faults 251,687,988 126,146,952 32,919,406 18,208,804 10,458,947 2,733,907 928,217
time, seconds 86.260656897 49.626551828 22.335007632 17.608243696 16.523119035 16.339489186 16.326390902
120 threads
minor-faults 503,352,863 252,939,677 67,039,168 35,191,827 19,170,091 4,688,357 1,471,862
time, seconds 124.589206333 79.757867787 39.508707872 32.167281632 29.972989292 28.729834575 28.042251622
Random access 1GiB file
1 thread
minor-faults 262,636 132,743 34,369 17,299 8,527 3,451 1,222
time, seconds 15.351890914 16.613802482 16.569227308 15.179220992 16.557356122 16.578247824 15.365266994
8 threads
minor-faults 2,098,948 1,061,871 273,690 154,501 87,110 25,663 7,384
time, seconds 15.040026343 15.096933500 14.474757288 14.289129964 14.411537468 14.296316837 14.395635804
32 threads
minor-faults 8,390,734 4,231,023 1,054,432 528,847 269,242 97,746 26,881
time, seconds 20.430433109 21.585235358 22.115062928 14.872878951 14.880856305 14.883370649 14.821261690
60 threads
minor-faults 15,733,258 7,892,809 1,973,393 988,266 594,789 164,994 51,691
time, seconds 26.577302548 25.692397770 18.728863715 20.153026398 21.619101933 17.745086260 17.613215273
120 threads
minor-faults 31,471,111 15,816,616 3,959,209 1,978,685 1,008,299 264,635 96,010
time, seconds 41.835322703 40.459786095 36.085306105 35.313894834 35.814445675 36.552633793 34.289210594
Touch only one page in page table in 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 8,372 8,324 8,270 8,260 8,249 8,239 8,237
time, seconds 0.039892712 0.045369149 0.051846126 0.063681685 0.079095975 0.17652406 0.541213386
8 threads
minor-faults 65,731 65,681 65,628 65,620 65,608 65,599 65,596
time, seconds 0.124159196 0.488600638 0.156854426 0.191901957 0.242631486 0.543569456 1.677303984
32 threads
minor-faults 262,388 262,341 262,285 262,276 262,266 262,257 263,183
time, seconds 0.452421421 0.488600638 0.565020946 0.648229739 0.789850823 1.651584361 5.000361559
60 threads
minor-faults 491,822 491,792 491,723 491,711 491,701 491,691 491,825
time, seconds 0.763288616 0.869620515 0.980727360 1.161732354 1.466915814 3.04041448 9.308612938
120 threads
minor-faults 983,466 983,655 983,366 983,372 983,363 984,083 984,164
time, seconds 1.595846553 1.667902182 2.008959376 2.425380942 2.941368804 5.977807890 18.401846125
This patch (of 2):
Introduce new vm_ops callback ->map_pages() and uses it for mapping easy
accessible pages around fault address.
On read page fault, if filesystem provides ->map_pages(), we try to map up
to FAULT_AROUND_PAGES pages around page fault address in hope to reduce
number of minor page faults.
We call ->map_pages first and use ->fault() as fallback if page by the
offset is not ready to be mapped (cold page cache or something).
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ning Qu <quning@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:18 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Let's call ->map_pages() first and use ->fault() as fallback
|
|
|
|
* if page by the offset is not ready to be mapped (cold cache or
|
|
|
|
* something).
|
|
|
|
*/
|
mm/shmem: handle uffd-wp special pte in page fault handler
File-backed memories are prone to unmap/swap so the ptes are always
unstable, because they can be easily faulted back later using the page
cache. This could lead to uffd-wp getting lost when unmapping or swapping
out such memory. One example is shmem. PTE markers are needed to store
those information.
This patch prepares it by handling uffd-wp pte markers first it is applied
elsewhere, so that the page fault handler can recognize uffd-wp pte
markers.
The handling of uffd-wp pte markers is similar to missing fault, it's just
that we'll handle this "missing fault" when we see the pte markers,
meanwhile we need to make sure the marker information is kept during
processing the fault.
This is a slow path of uffd-wp handling, because zapping of wr-protected
shmem ptes should be rare. So far it should only trigger in two
conditions:
(1) When trying to punch holes in shmem_fallocate(), there is an
optimization to zap the pgtables before evicting the page.
(2) When swapping out shmem pages.
Because of this, the page fault handling is simplifed too by not sending
the wr-protect message in the 1st page fault, instead the page will be
installed read-only, so the uffd-wp message will be generated in the next
fault, which will trigger the do_wp_page() path of general uffd-wp
handling.
Disable fault-around for all uffd-wp registered ranges for extra safety
just like uffd-minor fault, and clean the code up.
Link: https://lkml.kernel.org/r/20220405014844.14239-1-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-13 03:22:53 +00:00
|
|
|
if (should_fault_around(vmf)) {
|
|
|
|
ret = do_fault_around(vmf);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
mm: introduce vm_ops->map_pages()
Here's new version of faultaround patchset. It took a while to tune it
and collect performance data.
First patch adds new callback ->map_pages to vm_operations_struct.
->map_pages() is called when VM asks to map easy accessible pages.
Filesystem should find and map pages associated with offsets from
"pgoff" till "max_pgoff". ->map_pages() is called with page table
locked and must not block. If it's not possible to reach a page without
blocking, filesystem should skip it. Filesystem should use do_set_pte()
to setup page table entry. Pointer to entry associated with offset
"pgoff" is passed in "pte" field in vm_fault structure. Pointers to
entries for other offsets should be calculated relative to "pte".
Currently VM use ->map_pages only on read page fault path. We try to
map FAULT_AROUND_PAGES a time. FAULT_AROUND_PAGES is 16 for now.
Performance data for different FAULT_AROUND_ORDER is below.
TODO:
- implement ->map_pages() for shmem/tmpfs;
- modify get_user_pages() to be able to use ->map_pages() and implement
mmap(MAP_POPULATE|MAP_NONBLOCK) on top.
=========================================================================
Tested on 4-socket machine (120 threads) with 128GiB of RAM.
Few real-world workloads. The sweet spot for FAULT_AROUND_ORDER here is
somewhere between 3 and 5. Let's say 4 :)
Linux build (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 283,301,572 247,151,987 212,215,789 204,772,882 199,568,944 194,703,779 193,381,485
time, seconds 151.227629483 153.920996480 151.356125472 150.863792049 150.879207877 151.150764954 151.450962358
Linux rebuild (make -j60)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 5,396,854 4,148,444 2,855,286 2,577,282 2,361,957 2,169,573 2,112,643
time, seconds 27.404543757 27.559725591 27.030057426 26.855045126 26.678618635 26.974523490 26.761320095
Git test suite (make -j60 test)
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
minor-faults 129,591,823 99,200,751 66,106,718 57,606,410 51,510,808 45,776,813 44,085,515
time, seconds 66.087215026 64.784546905 64.401156567 65.282708668 66.034016829 66.793780811 67.237810413
Two synthetic tests: access every word in file in sequential/random order.
It doesn't improve much after FAULT_AROUND_ORDER == 4.
Sequential access 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 4,195,437 2,098,275 525,068 262,251 131,170 32,856 8,282
time, seconds 7.250461742 6.461711074 5.493859139 5.488488147 5.707213983 5.898510832 5.109232856
8 threads
minor-faults 33,557,540 16,892,728 4,515,848 2,366,999 1,423,382 442,732 142,339
time, seconds 16.649304881 9.312555263 6.612490639 6.394316732 6.669827501 6.75078944 6.371900528
32 threads
minor-faults 134,228,222 67,526,810 17,725,386 9,716,537 4,763,731 1,668,921 537,200
time, seconds 49.164430543 29.712060103 12.938649729 10.175151004 11.840094583 9.594081325 9.928461797
60 threads
minor-faults 251,687,988 126,146,952 32,919,406 18,208,804 10,458,947 2,733,907 928,217
time, seconds 86.260656897 49.626551828 22.335007632 17.608243696 16.523119035 16.339489186 16.326390902
120 threads
minor-faults 503,352,863 252,939,677 67,039,168 35,191,827 19,170,091 4,688,357 1,471,862
time, seconds 124.589206333 79.757867787 39.508707872 32.167281632 29.972989292 28.729834575 28.042251622
Random access 1GiB file
1 thread
minor-faults 262,636 132,743 34,369 17,299 8,527 3,451 1,222
time, seconds 15.351890914 16.613802482 16.569227308 15.179220992 16.557356122 16.578247824 15.365266994
8 threads
minor-faults 2,098,948 1,061,871 273,690 154,501 87,110 25,663 7,384
time, seconds 15.040026343 15.096933500 14.474757288 14.289129964 14.411537468 14.296316837 14.395635804
32 threads
minor-faults 8,390,734 4,231,023 1,054,432 528,847 269,242 97,746 26,881
time, seconds 20.430433109 21.585235358 22.115062928 14.872878951 14.880856305 14.883370649 14.821261690
60 threads
minor-faults 15,733,258 7,892,809 1,973,393 988,266 594,789 164,994 51,691
time, seconds 26.577302548 25.692397770 18.728863715 20.153026398 21.619101933 17.745086260 17.613215273
120 threads
minor-faults 31,471,111 15,816,616 3,959,209 1,978,685 1,008,299 264,635 96,010
time, seconds 41.835322703 40.459786095 36.085306105 35.313894834 35.814445675 36.552633793 34.289210594
Touch only one page in page table in 16GiB file
FAULT_AROUND_ORDER Baseline 1 3 4 5 7 9
1 thread
minor-faults 8,372 8,324 8,270 8,260 8,249 8,239 8,237
time, seconds 0.039892712 0.045369149 0.051846126 0.063681685 0.079095975 0.17652406 0.541213386
8 threads
minor-faults 65,731 65,681 65,628 65,620 65,608 65,599 65,596
time, seconds 0.124159196 0.488600638 0.156854426 0.191901957 0.242631486 0.543569456 1.677303984
32 threads
minor-faults 262,388 262,341 262,285 262,276 262,266 262,257 263,183
time, seconds 0.452421421 0.488600638 0.565020946 0.648229739 0.789850823 1.651584361 5.000361559
60 threads
minor-faults 491,822 491,792 491,723 491,711 491,701 491,691 491,825
time, seconds 0.763288616 0.869620515 0.980727360 1.161732354 1.466915814 3.04041448 9.308612938
120 threads
minor-faults 983,466 983,655 983,366 983,372 983,363 984,083 984,164
time, seconds 1.595846553 1.667902182 2.008959376 2.425380942 2.941368804 5.977807890 18.401846125
This patch (of 2):
Introduce new vm_ops callback ->map_pages() and uses it for mapping easy
accessible pages around fault address.
On read page fault, if filesystem provides ->map_pages(), we try to map up
to FAULT_AROUND_PAGES pages around page fault address in hope to reduce
number of minor page faults.
We call ->map_pages first and use ->fault() as fallback if page by the
offset is not ready to be mapped (cold page cache or something).
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ning Qu <quning@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-07 22:37:18 +00:00
|
|
|
}
|
2014-04-03 21:48:11 +00:00
|
|
|
|
2016-12-14 23:07:10 +00:00
|
|
|
ret = __do_fault(vmf);
|
2014-04-03 21:48:11 +00:00
|
|
|
if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
|
|
|
|
return ret;
|
|
|
|
|
2016-12-14 23:07:21 +00:00
|
|
|
ret |= finish_fault(vmf);
|
2016-12-14 23:07:10 +00:00
|
|
|
unlock_page(vmf->page);
|
2016-07-26 22:25:23 +00:00
|
|
|
if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
|
2016-12-14 23:07:10 +00:00
|
|
|
put_page(vmf->page);
|
2014-04-03 21:48:11 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_cow_fault(struct vm_fault *vmf)
|
2014-04-03 21:48:12 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret;
|
2014-04-03 21:48:12 +00:00
|
|
|
|
|
|
|
if (unlikely(anon_vma_prepare(vma)))
|
|
|
|
return VM_FAULT_OOM;
|
|
|
|
|
2016-12-14 23:07:10 +00:00
|
|
|
vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
|
|
|
|
if (!vmf->cow_page)
|
2014-04-03 21:48:12 +00:00
|
|
|
return VM_FAULT_OOM;
|
|
|
|
|
2021-06-25 13:27:04 +00:00
|
|
|
if (mem_cgroup_charge(page_folio(vmf->cow_page), vma->vm_mm,
|
|
|
|
GFP_KERNEL)) {
|
2016-12-14 23:07:10 +00:00
|
|
|
put_page(vmf->cow_page);
|
2014-04-03 21:48:12 +00:00
|
|
|
return VM_FAULT_OOM;
|
|
|
|
}
|
2023-03-02 11:58:34 +00:00
|
|
|
folio_throttle_swaprate(page_folio(vmf->cow_page), GFP_KERNEL);
|
2014-04-03 21:48:12 +00:00
|
|
|
|
2016-12-14 23:07:10 +00:00
|
|
|
ret = __do_fault(vmf);
|
2014-04-03 21:48:12 +00:00
|
|
|
if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
|
|
|
|
goto uncharge_out;
|
2016-12-14 23:07:18 +00:00
|
|
|
if (ret & VM_FAULT_DONE_COW)
|
|
|
|
return ret;
|
2014-04-03 21:48:12 +00:00
|
|
|
|
2016-12-14 23:07:24 +00:00
|
|
|
copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
|
2016-12-14 23:07:10 +00:00
|
|
|
__SetPageUptodate(vmf->cow_page);
|
2014-04-03 21:48:12 +00:00
|
|
|
|
2016-12-14 23:07:21 +00:00
|
|
|
ret |= finish_fault(vmf);
|
2016-12-14 23:07:24 +00:00
|
|
|
unlock_page(vmf->page);
|
|
|
|
put_page(vmf->page);
|
2016-07-26 22:25:23 +00:00
|
|
|
if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
|
|
|
|
goto uncharge_out;
|
2014-04-03 21:48:12 +00:00
|
|
|
return ret;
|
|
|
|
uncharge_out:
|
2016-12-14 23:07:10 +00:00
|
|
|
put_page(vmf->cow_page);
|
2014-04-03 21:48:12 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_shared_fault(struct vm_fault *vmf)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret, tmp;
|
2011-07-26 00:12:27 +00:00
|
|
|
|
2016-12-14 23:07:10 +00:00
|
|
|
ret = __do_fault(vmf);
|
2014-04-03 21:48:10 +00:00
|
|
|
if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
|
2014-04-03 21:48:13 +00:00
|
|
|
return ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/*
|
2014-04-03 21:48:13 +00:00
|
|
|
* Check if the backing address space wants to know that the page is
|
|
|
|
* about to become writable
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2014-04-03 21:48:15 +00:00
|
|
|
if (vma->vm_ops->page_mkwrite) {
|
2016-12-14 23:07:10 +00:00
|
|
|
unlock_page(vmf->page);
|
2016-12-14 23:07:30 +00:00
|
|
|
tmp = do_page_mkwrite(vmf);
|
2014-04-03 21:48:15 +00:00
|
|
|
if (unlikely(!tmp ||
|
|
|
|
(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
|
2016-12-14 23:07:10 +00:00
|
|
|
put_page(vmf->page);
|
2014-04-03 21:48:15 +00:00
|
|
|
return tmp;
|
2005-10-30 01:16:05 +00:00
|
|
|
}
|
2014-04-03 21:48:15 +00:00
|
|
|
}
|
|
|
|
|
2016-12-14 23:07:21 +00:00
|
|
|
ret |= finish_fault(vmf);
|
2016-07-26 22:25:23 +00:00
|
|
|
if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
|
|
|
|
VM_FAULT_RETRY))) {
|
2016-12-14 23:07:10 +00:00
|
|
|
unlock_page(vmf->page);
|
|
|
|
put_page(vmf->page);
|
2014-04-03 21:48:13 +00:00
|
|
|
return ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
mm: close page_mkwrite races
Change page_mkwrite to allow implementations to return with the page
locked, and also change it's callers (in page fault paths) to hold the
lock until the page is marked dirty. This allows the filesystem to have
full control of page dirtying events coming from the VM.
Rather than simply hold the page locked over the page_mkwrite call, we
call page_mkwrite with the page unlocked and allow callers to return with
it locked, so filesystems can avoid LOR conditions with page lock.
The problem with the current scheme is this: a filesystem that wants to
associate some metadata with a page as long as the page is dirty, will
perform this manipulation in its ->page_mkwrite. It currently then must
return with the page unlocked and may not hold any other locks (according
to existing page_mkwrite convention).
In this window, the VM could write out the page, clearing page-dirty. The
filesystem has no good way to detect that a dirty pte is about to be
attached, so it will happily write out the page, at which point, the
filesystem may manipulate the metadata to reflect that the page is no
longer dirty.
It is not always possible to perform the required metadata manipulation in
->set_page_dirty, because that function cannot block or fail. The
filesystem may need to allocate some data structure, for example.
And the VM cannot mark the pte dirty before page_mkwrite, because
page_mkwrite is allowed to fail, so we must not allow any window where the
page could be written to if page_mkwrite does fail.
This solution of holding the page locked over the 3 critical operations
(page_mkwrite, setting the pte dirty, and finally setting the page dirty)
closes out races nicely, preventing page cleaning for writeout being
initiated in that window. This provides the filesystem with a strong
synchronisation against the VM here.
- Sage needs this race closed for ceph filesystem.
- Trond for NFS (http://bugzilla.kernel.org/show_bug.cgi?id=12913).
- I need it for fsblock.
- I suspect other filesystems may need it too (eg. btrfs).
- I have converted buffer.c to the new locking. Even simple block allocation
under dirty pages might be susceptible to i_size changing under partial page
at the end of file (we also have a buffer.c-side problem here, but it cannot
be fixed properly without this patch).
- Other filesystems (eg. NFS, maybe btrfs) will need to change their
page_mkwrite functions themselves.
[ This also moves page_mkwrite another step closer to fault, which should
eventually allow page_mkwrite to be moved into ->fault, and thus avoiding a
filesystem calldown and page lock/unlock cycle in __do_fault. ]
[akpm@linux-foundation.org: fix derefs of NULL ->mapping]
Cc: Sage Weil <sage@newdream.net>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-04-30 22:08:16 +00:00
|
|
|
|
2019-12-01 01:50:22 +00:00
|
|
|
ret |= fault_dirty_shared_page(vmf);
|
2011-07-26 00:12:27 +00:00
|
|
|
return ret;
|
2007-07-19 08:46:59 +00:00
|
|
|
}
|
mm: fix fault vs invalidate race for linear mappings
Fix the race between invalidate_inode_pages and do_no_page.
Andrea Arcangeli identified a subtle race between invalidation of pages from
pagecache with userspace mappings, and do_no_page.
The issue is that invalidation has to shoot down all mappings to the page,
before it can be discarded from the pagecache. Between shooting down ptes to
a particular page, and actually dropping the struct page from the pagecache,
do_no_page from any process might fault on that page and establish a new
mapping to the page just before it gets discarded from the pagecache.
The most common case where such invalidation is used is in file truncation.
This case was catered for by doing a sort of open-coded seqlock between the
file's i_size, and its truncate_count.
Truncation will decrease i_size, then increment truncate_count before
unmapping userspace pages; do_no_page will read truncate_count, then find the
page if it is within i_size, and then check truncate_count under the page
table lock and back out and retry if it had subsequently been changed (ptl
will serialise against unmapping, and ensure a potentially updated
truncate_count is actually visible).
Complexity and documentation issues aside, the locking protocol fails in the
case where we would like to invalidate pagecache inside i_size. do_no_page
can come in anytime and filemap_nopage is not aware of the invalidation in
progress (as it is when it is outside i_size). The end result is that
dangling (->mapping == NULL) pages that appear to be from a particular file
may be mapped into userspace with nonsense data. Valid mappings to the same
place will see a different page.
Andrea implemented two working fixes, one using a real seqlock, another using
a page->flags bit. He also proposed using the page lock in do_no_page, but
that was initially considered too heavyweight. However, it is not a global or
per-file lock, and the page cacheline is modified in do_no_page to increment
_count and _mapcount anyway, so a further modification should not be a large
performance hit. Scalability is not an issue.
This patch implements this latter approach. ->nopage implementations return
with the page locked if it is possible for their underlying file to be
invalidated (in that case, they must set a special vm_flags bit to indicate
so). do_no_page only unlocks the page after setting up the mapping
completely. invalidation is excluded because it holds the page lock during
invalidation of each page (and ensures that the page is not mapped while
holding the lock).
This also allows significant simplifications in do_no_page, because we have
the page locked in the right place in the pagecache from the start.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 08:46:57 +00:00
|
|
|
|
2014-08-06 23:07:24 +00:00
|
|
|
/*
|
2020-06-09 04:33:54 +00:00
|
|
|
* We enter with non-exclusive mmap_lock (to exclude vma changes,
|
2014-08-06 23:07:24 +00:00
|
|
|
* but allow concurrent faults).
|
2020-06-09 04:33:54 +00:00
|
|
|
* The mmap_lock may have been released depending on flags and our
|
2021-03-19 01:39:45 +00:00
|
|
|
* return value. See filemap_fault() and __folio_lock_or_retry().
|
2020-06-09 04:33:54 +00:00
|
|
|
* If mmap_lock is released, vma may become invalid (for example
|
2019-03-05 23:50:08 +00:00
|
|
|
* by other thread calling munmap()).
|
2014-08-06 23:07:24 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_fault(struct vm_fault *vmf)
|
2007-07-19 08:46:59 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2019-03-05 23:50:08 +00:00
|
|
|
struct mm_struct *vm_mm = vma->vm_mm;
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret;
|
2007-07-19 08:46:59 +00:00
|
|
|
|
2018-10-26 22:09:01 +00:00
|
|
|
/*
|
|
|
|
* The VMA was not fully populated on mmap() or missing VM_DONTEXPAND
|
|
|
|
*/
|
|
|
|
if (!vma->vm_ops->fault) {
|
|
|
|
/*
|
|
|
|
* If we find a migration pmd entry or a none pmd entry, which
|
|
|
|
* should never happen, return SIGBUS
|
|
|
|
*/
|
|
|
|
if (unlikely(!pmd_present(*vmf->pmd)))
|
|
|
|
ret = VM_FAULT_SIGBUS;
|
|
|
|
else {
|
|
|
|
vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm,
|
|
|
|
vmf->pmd,
|
|
|
|
vmf->address,
|
|
|
|
&vmf->ptl);
|
|
|
|
/*
|
|
|
|
* Make sure this is not a temporary clearing of pte
|
|
|
|
* by holding ptl and checking again. A R/M/W update
|
|
|
|
* of pte involves: take ptl, clearing the pte so that
|
|
|
|
* we don't have concurrent modification by hardware
|
|
|
|
* followed by an update.
|
|
|
|
*/
|
|
|
|
if (unlikely(pte_none(*vmf->pte)))
|
|
|
|
ret = VM_FAULT_SIGBUS;
|
|
|
|
else
|
|
|
|
ret = VM_FAULT_NOPAGE;
|
|
|
|
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
|
|
|
}
|
|
|
|
} else if (!(vmf->flags & FAULT_FLAG_WRITE))
|
mm: stop leaking PageTables
4.10-rc loadtest (even on x86, and even without THPCache) fails with
"fork: Cannot allocate memory" or some such; and /proc/meminfo shows
PageTables growing.
Commit 953c66c2b22a ("mm: THP page cache support for ppc64") that got
merged in rc1 removed the freeing of an unused preallocated pagetable
after do_fault_around() has called map_pages().
This is usually a good optimization, so that the followup doesn't have
to reallocate one; but it's not sufficient to shift the freeing into
alloc_set_pte(), since there are failure cases (most commonly
VM_FAULT_RETRY) which never reach finish_fault().
Check and free it at the outer level in do_fault(), then we don't need
to worry in alloc_set_pte(), and can restore that to how it was (I
cannot find any reason to pte_free() under lock as it was doing).
And fix a separate pagetable leak, or crash, introduced by the same
change, that could only show up on some ppc64: why does do_set_pmd()'s
failure case attempt to withdraw a pagetable when it never deposited
one, at the same time overwriting (so leaking) the vmf->prealloc_pte?
Residue of an earlier implementation, perhaps? Delete it.
Fixes: 953c66c2b22a ("mm: THP page cache support for ppc64")
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Neuling <mikey@neuling.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-07 23:37:31 +00:00
|
|
|
ret = do_read_fault(vmf);
|
|
|
|
else if (!(vma->vm_flags & VM_SHARED))
|
|
|
|
ret = do_cow_fault(vmf);
|
|
|
|
else
|
|
|
|
ret = do_shared_fault(vmf);
|
|
|
|
|
|
|
|
/* preallocated pagetable is unused: free it */
|
|
|
|
if (vmf->prealloc_pte) {
|
2019-03-05 23:50:08 +00:00
|
|
|
pte_free(vm_mm, vmf->prealloc_pte);
|
2017-02-24 22:58:59 +00:00
|
|
|
vmf->prealloc_pte = NULL;
|
mm: stop leaking PageTables
4.10-rc loadtest (even on x86, and even without THPCache) fails with
"fork: Cannot allocate memory" or some such; and /proc/meminfo shows
PageTables growing.
Commit 953c66c2b22a ("mm: THP page cache support for ppc64") that got
merged in rc1 removed the freeing of an unused preallocated pagetable
after do_fault_around() has called map_pages().
This is usually a good optimization, so that the followup doesn't have
to reallocate one; but it's not sufficient to shift the freeing into
alloc_set_pte(), since there are failure cases (most commonly
VM_FAULT_RETRY) which never reach finish_fault().
Check and free it at the outer level in do_fault(), then we don't need
to worry in alloc_set_pte(), and can restore that to how it was (I
cannot find any reason to pte_free() under lock as it was doing).
And fix a separate pagetable leak, or crash, introduced by the same
change, that could only show up on some ppc64: why does do_set_pmd()'s
failure case attempt to withdraw a pagetable when it never deposited
one, at the same time overwriting (so leaking) the vmf->prealloc_pte?
Residue of an earlier implementation, perhaps? Delete it.
Fixes: 953c66c2b22a ("mm: THP page cache support for ppc64")
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Neuling <mikey@neuling.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-07 23:37:31 +00:00
|
|
|
}
|
|
|
|
return ret;
|
2007-07-19 08:46:59 +00:00
|
|
|
}
|
|
|
|
|
2021-07-01 01:51:39 +00:00
|
|
|
int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
|
|
|
|
unsigned long addr, int page_nid, int *flags)
|
2012-11-15 01:24:32 +00:00
|
|
|
{
|
|
|
|
get_page(page);
|
|
|
|
|
2023-03-01 12:19:01 +00:00
|
|
|
/* Record the current PID acceesing VMA */
|
|
|
|
vma_set_access_pid_bit(vma);
|
|
|
|
|
2012-11-15 01:24:32 +00:00
|
|
|
count_vm_numa_event(NUMA_HINT_FAULTS);
|
2013-10-07 10:29:36 +00:00
|
|
|
if (page_nid == numa_node_id()) {
|
2012-11-15 01:24:32 +00:00
|
|
|
count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
|
2013-10-07 10:29:36 +00:00
|
|
|
*flags |= TNF_FAULT_LOCAL;
|
|
|
|
}
|
2012-11-15 01:24:32 +00:00
|
|
|
|
|
|
|
return mpol_misplaced(page, vma, addr);
|
|
|
|
}
|
|
|
|
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t do_numa_page(struct vm_fault *vmf)
|
2012-10-25 12:16:31 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_area_struct *vma = vmf->vma;
|
2012-11-02 11:33:45 +00:00
|
|
|
struct page *page = NULL;
|
2019-03-05 23:42:58 +00:00
|
|
|
int page_nid = NUMA_NO_NODE;
|
2022-11-08 17:46:50 +00:00
|
|
|
bool writable = false;
|
2013-10-07 10:29:20 +00:00
|
|
|
int last_cpupid;
|
2012-10-25 12:16:43 +00:00
|
|
|
int target_nid;
|
2019-03-05 23:46:29 +00:00
|
|
|
pte_t pte, old_pte;
|
2013-10-07 10:29:24 +00:00
|
|
|
int flags = 0;
|
2012-10-25 12:16:31 +00:00
|
|
|
|
|
|
|
/*
|
2017-02-24 22:59:01 +00:00
|
|
|
* The "pte" at this point cannot be used safely without
|
|
|
|
* validation through pte_unmap_same(). It's of NUMA type but
|
|
|
|
* the pfn may be screwed if the read is non atomic.
|
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->ptl = pte_lockptr(vma->vm_mm, vmf->pmd);
|
|
|
|
spin_lock(vmf->ptl);
|
2017-02-24 22:59:13 +00:00
|
|
|
if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2012-11-02 11:33:45 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
/* Get the normal PTE */
|
|
|
|
old_pte = ptep_get(vmf->pte);
|
2019-03-05 23:46:29 +00:00
|
|
|
pte = pte_modify(old_pte, vma->vm_page_prot);
|
2012-10-25 12:16:31 +00:00
|
|
|
|
2022-11-08 17:46:50 +00:00
|
|
|
/*
|
|
|
|
* Detect now whether the PTE could be writable; this information
|
|
|
|
* is only valid while holding the PT lock.
|
|
|
|
*/
|
|
|
|
writable = pte_write(pte);
|
|
|
|
if (!writable && vma_wants_manual_pte_write_upgrade(vma) &&
|
|
|
|
can_change_pte_writable(vma, vmf->address, pte))
|
|
|
|
writable = true;
|
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
page = vm_normal_page(vma, vmf->address, pte);
|
2022-07-15 15:05:11 +00:00
|
|
|
if (!page || is_zone_device_page(page))
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
goto out_map;
|
2012-10-25 12:16:31 +00:00
|
|
|
|
2016-01-16 00:53:49 +00:00
|
|
|
/* TODO: handle PTE-mapped THP */
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
if (PageCompound(page))
|
|
|
|
goto out_map;
|
2016-01-16 00:53:49 +00:00
|
|
|
|
2013-10-07 10:29:24 +00:00
|
|
|
/*
|
2015-03-25 22:55:37 +00:00
|
|
|
* Avoid grouping on RO pages in general. RO pages shouldn't hurt as
|
|
|
|
* much anyway since they can be in shared cache state. This misses
|
|
|
|
* the case where a mapping is writable but the process never writes
|
|
|
|
* to it but pte_write gets cleared during protection updates and
|
|
|
|
* pte_dirty has unpredictable behaviour between PTE scan updates,
|
|
|
|
* background writeback, dirty balancing and application behaviour.
|
2013-10-07 10:29:24 +00:00
|
|
|
*/
|
2022-11-08 17:46:50 +00:00
|
|
|
if (!writable)
|
2013-10-07 10:29:24 +00:00
|
|
|
flags |= TNF_NO_GROUP;
|
|
|
|
|
2013-10-07 10:29:34 +00:00
|
|
|
/*
|
|
|
|
* Flag if the page is shared between multiple address spaces. This
|
|
|
|
* is later used when determining whether to group tasks together
|
|
|
|
*/
|
|
|
|
if (page_mapcount(page) > 1 && (vma->vm_flags & VM_SHARED))
|
|
|
|
flags |= TNF_SHARED;
|
|
|
|
|
mm: numa: Sanitize task_numa_fault() callsites
There are three callers of task_numa_fault():
- do_huge_pmd_numa_page():
Accounts against the current node, not the node where the
page resides, unless we migrated, in which case it accounts
against the node we migrated to.
- do_numa_page():
Accounts against the current node, not the node where the
page resides, unless we migrated, in which case it accounts
against the node we migrated to.
- do_pmd_numa_page():
Accounts not at all when the page isn't migrated, otherwise
accounts against the node we migrated towards.
This seems wrong to me; all three sites should have the same
sementaics, furthermore we should accounts against where the page
really is, we already know where the task is.
So modify all three sites to always account; we did after all receive
the fault; and always account to where the page is after migration,
regardless of success.
They all still differ on when they clear the PTE/PMD; ideally that
would get sorted too.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-8-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2013-10-07 10:28:45 +00:00
|
|
|
page_nid = page_to_nid(page);
|
memory tiering: hot page selection with hint page fault latency
Patch series "memory tiering: hot page selection", v4.
To optimize page placement in a memory tiering system with NUMA balancing,
the hot pages in the slow memory nodes need to be identified.
Essentially, the original NUMA balancing implementation selects the mostly
recently accessed (MRU) pages to promote. But this isn't a perfect
algorithm to identify the hot pages. Because the pages with quite low
access frequency may be accessed eventually given the NUMA balancing page
table scanning period could be quite long (e.g. 60 seconds). So in this
patchset, we implement a new hot page identification algorithm based on
the latency between NUMA balancing page table scanning and hint page
fault. Which is a kind of mostly frequently accessed (MFU) algorithm.
In NUMA balancing memory tiering mode, if there are hot pages in slow
memory node and cold pages in fast memory node, we need to promote/demote
hot/cold pages between the fast and cold memory nodes.
A choice is to promote/demote as fast as possible. But the CPU cycles and
memory bandwidth consumed by the high promoting/demoting throughput will
hurt the latency of some workload because of accessing inflating and slow
memory bandwidth contention.
A way to resolve this issue is to restrict the max promoting/demoting
throughput. It will take longer to finish the promoting/demoting. But
the workload latency will be better. This is implemented in this patchset
as the page promotion rate limit mechanism.
The promotion hot threshold is workload and system configuration
dependent. So in this patchset, a method to adjust the hot threshold
automatically is implemented. The basic idea is to control the number of
the candidate promotion pages to match the promotion rate limit.
We used the pmbench memory accessing benchmark tested the patchset on a
2-socket server system with DRAM and PMEM installed. The test results are
as follows,
pmbench score promote rate
(accesses/s) MB/s
------------- ------------
base 146887704.1 725.6
hot selection 165695601.2 544.0
rate limit 162814569.8 165.2
auto adjustment 170495294.0 136.9
From the results above,
With hot page selection patch [1/3], the pmbench score increases about
12.8%, and promote rate (overhead) decreases about 25.0%, compared with
base kernel.
With rate limit patch [2/3], pmbench score decreases about 1.7%, and
promote rate decreases about 69.6%, compared with hot page selection
patch.
With threshold auto adjustment patch [3/3], pmbench score increases about
4.7%, and promote rate decrease about 17.1%, compared with rate limit
patch.
Baolin helped to test the patchset with MySQL on a machine which contains
1 DRAM node (30G) and 1 PMEM node (126G).
sysbench /usr/share/sysbench/oltp_read_write.lua \
......
--tables=200 \
--table-size=1000000 \
--report-interval=10 \
--threads=16 \
--time=120
The tps can be improved about 5%.
This patch (of 3):
To optimize page placement in a memory tiering system with NUMA balancing,
the hot pages in the slow memory node need to be identified. Essentially,
the original NUMA balancing implementation selects the mostly recently
accessed (MRU) pages to promote. But this isn't a perfect algorithm to
identify the hot pages. Because the pages with quite low access frequency
may be accessed eventually given the NUMA balancing page table scanning
period could be quite long (e.g. 60 seconds). The most frequently
accessed (MFU) algorithm is better.
So, in this patch we implemented a better hot page selection algorithm.
Which is based on NUMA balancing page table scanning and hint page fault
as follows,
- When the page tables of the processes are scanned to change PTE/PMD
to be PROT_NONE, the current time is recorded in struct page as scan
time.
- When the page is accessed, hint page fault will occur. The scan
time is gotten from the struct page. And The hint page fault
latency is defined as
hint page fault time - scan time
The shorter the hint page fault latency of a page is, the higher the
probability of their access frequency to be higher. So the hint page
fault latency is a better estimation of the page hot/cold.
It's hard to find some extra space in struct page to hold the scan time.
Fortunately, we can reuse some bits used by the original NUMA balancing.
NUMA balancing uses some bits in struct page to store the page accessing
CPU and PID (referring to page_cpupid_xchg_last()). Which is used by the
multi-stage node selection algorithm to avoid to migrate pages shared
accessed by the NUMA nodes back and forth. But for pages in the slow
memory node, even if they are shared accessed by multiple NUMA nodes, as
long as the pages are hot, they need to be promoted to the fast memory
node. So the accessing CPU and PID information are unnecessary for the
slow memory pages. We can reuse these bits in struct page to record the
scan time. For the fast memory pages, these bits are used as before.
For the hot threshold, the default value is 1 second, which works well in
our performance test. All pages with hint page fault latency < hot
threshold will be considered hot.
It's hard for users to determine the hot threshold. So we don't provide a
kernel ABI to set it, just provide a debugfs interface for advanced users
to experiment. We will continue to work on a hot threshold automatic
adjustment mechanism.
The downside of the above method is that the response time to the workload
hot spot changing may be much longer. For example,
- A previous cold memory area becomes hot
- The hint page fault will be triggered. But the hint page fault
latency isn't shorter than the hot threshold. So the pages will
not be promoted.
- When the memory area is scanned again, maybe after a scan period,
the hint page fault latency measured will be shorter than the hot
threshold and the pages will be promoted.
To mitigate this, if there are enough free space in the fast memory node,
the hot threshold will not be used, all pages will be promoted upon the
hint page fault for fast response.
Thanks Zhong Jiang reported and tested the fix for a bug when disabling
memory tiering mode dynamically.
Link: https://lkml.kernel.org/r/20220713083954.34196-1-ying.huang@intel.com
Link: https://lkml.kernel.org/r/20220713083954.34196-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Tested-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Wei Xu <weixugc@google.com>
Cc: osalvador <osalvador@suse.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Zhong Jiang <zhongjiang-ali@linux.alibaba.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-07-13 08:39:51 +00:00
|
|
|
/*
|
|
|
|
* For memory tiering mode, cpupid of slow memory page is used
|
|
|
|
* to record page access time. So use default value.
|
|
|
|
*/
|
|
|
|
if ((sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING) &&
|
|
|
|
!node_is_toptier(page_nid))
|
|
|
|
last_cpupid = (-1 & LAST_CPUPID_MASK);
|
|
|
|
else
|
|
|
|
last_cpupid = page_cpupid_last(page);
|
2016-12-14 23:06:58 +00:00
|
|
|
target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
|
2016-07-26 22:25:20 +00:00
|
|
|
&flags);
|
2019-03-05 23:42:58 +00:00
|
|
|
if (target_nid == NUMA_NO_NODE) {
|
2012-11-02 11:33:45 +00:00
|
|
|
put_page(page);
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
goto out_map;
|
2012-11-02 11:33:45 +00:00
|
|
|
}
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2022-11-08 17:46:50 +00:00
|
|
|
writable = false;
|
2012-11-02 11:33:45 +00:00
|
|
|
|
|
|
|
/* Migrate to the requested node */
|
2021-04-30 05:57:07 +00:00
|
|
|
if (migrate_misplaced_page(page, vma, target_nid)) {
|
mm: numa: Sanitize task_numa_fault() callsites
There are three callers of task_numa_fault():
- do_huge_pmd_numa_page():
Accounts against the current node, not the node where the
page resides, unless we migrated, in which case it accounts
against the node we migrated to.
- do_numa_page():
Accounts against the current node, not the node where the
page resides, unless we migrated, in which case it accounts
against the node we migrated to.
- do_pmd_numa_page():
Accounts not at all when the page isn't migrated, otherwise
accounts against the node we migrated towards.
This seems wrong to me; all three sites should have the same
sementaics, furthermore we should accounts against where the page
really is, we already know where the task is.
So modify all three sites to always account; we did after all receive
the fault; and always account to where the page is after migration,
regardless of success.
They all still differ on when they clear the PTE/PMD; ideally that
would get sorted too.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-8-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2013-10-07 10:28:45 +00:00
|
|
|
page_nid = target_nid;
|
2013-10-07 10:29:24 +00:00
|
|
|
flags |= TNF_MIGRATED;
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
} else {
|
2015-03-25 22:55:42 +00:00
|
|
|
flags |= TNF_MIGRATE_FAIL;
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
|
|
|
|
spin_lock(vmf->ptl);
|
|
|
|
if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
goto out_map;
|
|
|
|
}
|
2012-11-02 11:33:45 +00:00
|
|
|
|
|
|
|
out:
|
2019-03-05 23:42:58 +00:00
|
|
|
if (page_nid != NUMA_NO_NODE)
|
2013-10-07 10:29:24 +00:00
|
|
|
task_numa_fault(last_cpupid, page_nid, 1, flags);
|
2012-10-25 12:16:31 +00:00
|
|
|
return 0;
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
out_map:
|
|
|
|
/*
|
|
|
|
* Make it present again, depending on how arch implements
|
|
|
|
* non-accessible ptes, some can allow access by kernel mode.
|
|
|
|
*/
|
|
|
|
old_pte = ptep_modify_prot_start(vma, vmf->address, vmf->pte);
|
|
|
|
pte = pte_modify(old_pte, vma->vm_page_prot);
|
|
|
|
pte = pte_mkyoung(pte);
|
2022-11-08 17:46:50 +00:00
|
|
|
if (writable)
|
NUMA balancing: reduce TLB flush via delaying mapping on hint page fault
With NUMA balancing, in hint page fault handler, the faulting page will be
migrated to the accessing node if necessary. During the migration, TLB
will be shot down on all CPUs that the process has run on recently.
Because in the hint page fault handler, the PTE will be made accessible
before the migration is tried. The overhead of TLB shooting down can be
high, so it's better to be avoided if possible. In fact, if we delay
mapping the page until migration, that can be avoided. This is what this
patch doing.
For the multiple threads applications, it's possible that a page is
accessed by multiple threads almost at the same time. In the original
implementation, because the first thread will install the accessible PTE
before migrating the page, the other threads may access the page directly
before the page is made inaccessible again during migration. While with
the patch, the second thread will go through the page fault handler too.
And because of the PageLRU() checking in the following code path,
migrate_misplaced_page()
numamigrate_isolate_page()
isolate_lru_page()
the migrate_misplaced_page() will return 0, and the PTE will be made
accessible in the second thread.
This will introduce a little more overhead. But we think the possibility
for a page to be accessed by the multiple threads at the same time is low,
and the overhead difference isn't too large. If this becomes a problem in
some workloads, we need to consider how to reduce the overhead.
To test the patch, we run a test case as follows on a 2-socket Intel
server (1 NUMA node per socket) with 128GB DRAM (64GB per socket).
1. Run a memory eater on NUMA node 1 to use 40GB memory before running
pmbench.
2. Run pmbench (normal accessing pattern) with 8 processes, and 8
threads per process, so there are 64 threads in total. The
working-set size of each process is 8960MB, so the total working-set
size is 8 * 8960MB = 70GB. The CPU of all pmbench processes is bound
to node 1. The pmbench processes will access some DRAM on node 0.
3. After the pmbench processes run for 10 seconds, kill the memory
eater. Now, some pages will be migrated from node 0 to node 1 via
NUMA balancing.
Test results show that, with the patch, the pmbench throughput (page
accesses/s) increases 5.5%. The number of the TLB shootdowns interrupts
reduces 98% (from ~4.7e7 to ~9.7e5) with about 9.2e6 pages (35.8GB)
migrated. From the perf profile, it can be found that the CPU cycles
spent by try_to_unmap() and its callees reduces from 6.02% to 0.47%. That
is, the CPU cycles spent by TLB shooting down decreases greatly.
Link: https://lkml.kernel.org/r/20210408132236.1175607-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Matthew Wilcox" <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Arjun Roy <arjunroy@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-04-30 05:57:41 +00:00
|
|
|
pte = pte_mkwrite(pte);
|
|
|
|
ptep_modify_prot_commit(vma, vmf->address, vmf->pte, old_pte, pte);
|
|
|
|
update_mmu_cache(vma, vmf->address, vmf->pte);
|
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
|
|
|
goto out;
|
2012-10-25 12:16:31 +00:00
|
|
|
}
|
|
|
|
|
2018-08-24 00:01:36 +00:00
|
|
|
static inline vm_fault_t create_huge_pmd(struct vm_fault *vmf)
|
2015-09-08 21:58:48 +00:00
|
|
|
{
|
2017-02-22 23:40:06 +00:00
|
|
|
if (vma_is_anonymous(vmf->vma))
|
2016-12-14 23:06:58 +00:00
|
|
|
return do_huge_pmd_anonymous_page(vmf);
|
mm,fs,dax: change ->pmd_fault to ->huge_fault
Patch series "1G transparent hugepage support for device dax", v2.
The following series implements support for 1G trasparent hugepage on
x86 for device dax. The bulk of the code was written by Mathew Wilcox a
while back supporting transparent 1G hugepage for fs DAX. I have
forward ported the relevant bits to 4.10-rc. The current submission has
only the necessary code to support device DAX.
Comments from Dan Williams: So the motivation and intended user of this
functionality mirrors the motivation and users of 1GB page support in
hugetlbfs. Given expected capacities of persistent memory devices an
in-memory database may want to reduce tlb pressure beyond what they can
already achieve with 2MB mappings of a device-dax file. We have
customer feedback to that effect as Willy mentioned in his previous
version of these patches [1].
[1]: https://lkml.org/lkml/2016/1/31/52
Comments from Nilesh @ Oracle:
There are applications which have a process model; and if you assume
10,000 processes attempting to mmap all the 6TB memory available on a
server; we are looking at the following:
processes : 10,000
memory : 6TB
pte @ 4k page size: 8 bytes / 4K of memory * #processes = 6TB / 4k * 8 * 10000 = 1.5GB * 80000 = 120,000GB
pmd @ 2M page size: 120,000 / 512 = ~240GB
pud @ 1G page size: 240GB / 512 = ~480MB
As you can see with 2M pages, this system will use up an exorbitant
amount of DRAM to hold the page tables; but the 1G pages finally brings
it down to a reasonable level. Memory sizes will keep increasing; so
this number will keep increasing.
An argument can be made to convert the applications from process model
to thread model, but in the real world that may not be always practical.
Hopefully this helps explain the use case where this is valuable.
This patch (of 3):
In preparation for adding the ability to handle PUD pages, convert
vm_operations_struct.pmd_fault to vm_operations_struct.huge_fault. The
vm_fault structure is extended to include a union of the different page
table pointers that may be needed, and three flag bits are reserved to
indicate which type of pointer is in the union.
[ross.zwisler@linux.intel.com: remove unused function ext4_dax_huge_fault()]
Link: http://lkml.kernel.org/r/1485813172-7284-1-git-send-email-ross.zwisler@linux.intel.com
[dave.jiang@intel.com: clear PMD or PUD size flags when in fall through path]
Link: http://lkml.kernel.org/r/148589842696.5820.16078080610311444794.stgit@djiang5-desk3.ch.intel.com
Link: http://lkml.kernel.org/r/148545058784.17912.6353162518188733642.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jan Kara <jack@suse.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Nilesh Choudhury <nilesh.choudhury@oracle.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-02-24 22:56:59 +00:00
|
|
|
if (vmf->vma->vm_ops->huge_fault)
|
2017-02-24 22:57:08 +00:00
|
|
|
return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
|
2015-09-08 21:58:48 +00:00
|
|
|
return VM_FAULT_FALLBACK;
|
|
|
|
}
|
|
|
|
|
2017-12-14 23:32:52 +00:00
|
|
|
/* `inline' is required to avoid gcc 4.1.2 build error */
|
mm: memory: add orig_pmd to struct vm_fault
Pach series "mm: thp: use generic THP migration for NUMA hinting fault", v3.
When the THP NUMA fault support was added THP migration was not supported
yet. So the ad hoc THP migration was implemented in NUMA fault handling.
Since v4.14 THP migration has been supported so it doesn't make too much
sense to still keep another THP migration implementation rather than using
the generic migration code. It is definitely a maintenance burden to keep
two THP migration implementation for different code paths and it is more
error prone. Using the generic THP migration implementation allows us
remove the duplicate code and some hacks needed by the old ad hoc
implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both
THP and NUMA balancing. The most of them support THP migration except for
S390. Zi Yan tried to add THP migration support for S390 before but it
was not accepted due to the design of S390 PMD. For the discussion,
please see: https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip
huge PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't
figure out if they have been removed or not since I just found FOLL_NUMA
code in the current gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page. The test
VM has 80 vcpus and 64G memory. The test would create 2 processes to
consume 128G memory together which would incur memory pressure to cause
THP splits. And it also creates 80 processes to hog cpu, and the memory
consumer processes are bound to different nodes periodically in order to
increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page is
as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted the
raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split is
less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory pressure.
No obvious regression is spotted. The below is the test result *w/o*
memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
This patch (of 7):
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge
page fault could be removed, just like its PTE counterpart does.
Link: https://lkml.kernel.org/r/20210518200801.7413-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20210518200801.7413-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:51:35 +00:00
|
|
|
static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf)
|
2015-09-08 21:58:48 +00:00
|
|
|
{
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
const bool unshare = vmf->flags & FAULT_FLAG_UNSHARE;
|
2022-11-16 10:26:46 +00:00
|
|
|
vm_fault_t ret;
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
|
2020-04-07 03:05:29 +00:00
|
|
|
if (vma_is_anonymous(vmf->vma)) {
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
if (likely(!unshare) &&
|
|
|
|
userfaultfd_huge_pmd_wp(vmf->vma, vmf->orig_pmd))
|
2020-04-07 03:05:29 +00:00
|
|
|
return handle_userfault(vmf, VM_UFFD_WP);
|
mm: memory: add orig_pmd to struct vm_fault
Pach series "mm: thp: use generic THP migration for NUMA hinting fault", v3.
When the THP NUMA fault support was added THP migration was not supported
yet. So the ad hoc THP migration was implemented in NUMA fault handling.
Since v4.14 THP migration has been supported so it doesn't make too much
sense to still keep another THP migration implementation rather than using
the generic migration code. It is definitely a maintenance burden to keep
two THP migration implementation for different code paths and it is more
error prone. Using the generic THP migration implementation allows us
remove the duplicate code and some hacks needed by the old ad hoc
implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both
THP and NUMA balancing. The most of them support THP migration except for
S390. Zi Yan tried to add THP migration support for S390 before but it
was not accepted due to the design of S390 PMD. For the discussion,
please see: https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip
huge PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't
figure out if they have been removed or not since I just found FOLL_NUMA
code in the current gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page. The test
VM has 80 vcpus and 64G memory. The test would create 2 processes to
consume 128G memory together which would incur memory pressure to cause
THP splits. And it also creates 80 processes to hog cpu, and the memory
consumer processes are bound to different nodes periodically in order to
increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page is
as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted the
raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split is
less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory pressure.
No obvious regression is spotted. The below is the test result *w/o*
memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
This patch (of 7):
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge
page fault could be removed, just like its PTE counterpart does.
Link: https://lkml.kernel.org/r/20210518200801.7413-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20210518200801.7413-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:51:35 +00:00
|
|
|
return do_huge_pmd_wp_page(vmf);
|
2020-04-07 03:05:29 +00:00
|
|
|
}
|
2020-03-24 17:47:47 +00:00
|
|
|
|
2022-11-16 10:26:46 +00:00
|
|
|
if (vmf->vma->vm_flags & (VM_SHARED | VM_MAYSHARE)) {
|
|
|
|
if (vmf->vma->vm_ops->huge_fault) {
|
|
|
|
ret = vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
|
|
|
|
if (!(ret & VM_FAULT_FALLBACK))
|
|
|
|
return ret;
|
|
|
|
}
|
2020-03-24 17:47:47 +00:00
|
|
|
}
|
2016-07-26 22:25:40 +00:00
|
|
|
|
2020-03-24 17:47:47 +00:00
|
|
|
/* COW or write-notify handled on pte level: split pmd. */
|
2016-12-14 23:06:58 +00:00
|
|
|
__split_huge_pmd(vmf->vma, vmf->pmd, vmf->address, false, NULL);
|
2016-07-26 22:25:40 +00:00
|
|
|
|
2015-09-08 21:58:48 +00:00
|
|
|
return VM_FAULT_FALLBACK;
|
|
|
|
}
|
|
|
|
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t create_huge_pud(struct vm_fault *vmf)
|
2017-02-24 22:57:02 +00:00
|
|
|
{
|
2022-06-23 05:24:03 +00:00
|
|
|
#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && \
|
|
|
|
defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)
|
|
|
|
/* No support for anonymous transparent PUD pages yet */
|
|
|
|
if (vma_is_anonymous(vmf->vma))
|
|
|
|
return VM_FAULT_FALLBACK;
|
|
|
|
if (vmf->vma->vm_ops->huge_fault)
|
|
|
|
return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
|
|
|
|
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
|
|
|
|
return VM_FAULT_FALLBACK;
|
|
|
|
}
|
|
|
|
|
|
|
|
static vm_fault_t wp_huge_pud(struct vm_fault *vmf, pud_t orig_pud)
|
|
|
|
{
|
2020-03-24 17:47:47 +00:00
|
|
|
#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && \
|
|
|
|
defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)
|
2022-11-16 10:26:46 +00:00
|
|
|
vm_fault_t ret;
|
|
|
|
|
2017-02-24 22:57:02 +00:00
|
|
|
/* No support for anonymous transparent PUD pages yet */
|
|
|
|
if (vma_is_anonymous(vmf->vma))
|
2020-03-24 17:47:47 +00:00
|
|
|
goto split;
|
2022-11-16 10:26:46 +00:00
|
|
|
if (vmf->vma->vm_flags & (VM_SHARED | VM_MAYSHARE)) {
|
|
|
|
if (vmf->vma->vm_ops->huge_fault) {
|
|
|
|
ret = vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
|
|
|
|
if (!(ret & VM_FAULT_FALLBACK))
|
|
|
|
return ret;
|
|
|
|
}
|
2020-03-24 17:47:47 +00:00
|
|
|
}
|
|
|
|
split:
|
|
|
|
/* COW or write-notify not handled on PUD level: split pud.*/
|
|
|
|
__split_huge_pud(vmf->vma, vmf->pud, vmf->address);
|
2022-06-23 05:24:03 +00:00
|
|
|
#endif /* CONFIG_TRANSPARENT_HUGEPAGE && CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
|
2017-02-24 22:57:02 +00:00
|
|
|
return VM_FAULT_FALLBACK;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* These routines also need to handle stuff like marking pages dirty
|
|
|
|
* and/or accessed for architectures that don't do it in hardware (most
|
|
|
|
* RISC architectures). The early dirtying is also good on the i386.
|
|
|
|
*
|
|
|
|
* There is also a hook called "update_mmu_cache()" that architectures
|
|
|
|
* with external mmu caches can use to update those (ie the Sparc or
|
|
|
|
* PowerPC hashed page tables that act as extended TLBs).
|
|
|
|
*
|
2020-06-09 04:33:54 +00:00
|
|
|
* We enter with non-exclusive mmap_lock (to exclude vma changes, but allow
|
2016-07-26 22:25:23 +00:00
|
|
|
* concurrent faults).
|
2014-08-06 23:07:24 +00:00
|
|
|
*
|
2020-06-09 04:33:54 +00:00
|
|
|
* The mmap_lock may have been released depending on flags and our return value.
|
2021-03-19 01:39:45 +00:00
|
|
|
* See filemap_fault() and __folio_lock_or_retry().
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t handle_pte_fault(struct vm_fault *vmf)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
pte_t entry;
|
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
if (unlikely(pmd_none(*vmf->pmd))) {
|
2016-07-26 22:25:23 +00:00
|
|
|
/*
|
|
|
|
* Leave __pte_alloc() until later: because vm_ops->fault may
|
|
|
|
* want to allocate huge page, and if we expose page table
|
|
|
|
* for an instant, it will be difficult to retract from
|
|
|
|
* concurrent faults and from rmap lookups.
|
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->pte = NULL;
|
2022-05-13 03:22:52 +00:00
|
|
|
vmf->flags &= ~FAULT_FLAG_ORIG_PTE_VALID;
|
2016-07-26 22:25:23 +00:00
|
|
|
} else {
|
2020-12-19 12:19:23 +00:00
|
|
|
/*
|
|
|
|
* If a huge pmd materialized under us just retry later. Use
|
|
|
|
* pmd_trans_unstable() via pmd_devmap_trans_unstable() instead
|
|
|
|
* of pmd_trans_huge() to ensure the pmd didn't become
|
|
|
|
* pmd_trans_huge under us and then back to pmd_none, as a
|
|
|
|
* result of MADV_DONTNEED running immediately after a huge pmd
|
|
|
|
* fault in a different thread of this mm, in turn leading to a
|
|
|
|
* misleading pmd_trans_huge() retval. All we have to ensure is
|
|
|
|
* that it is a regular pmd that we can walk with
|
|
|
|
* pte_offset_map() and we can do that through an atomic read
|
|
|
|
* in C, which is what pmd_trans_unstable() provides.
|
|
|
|
*/
|
2017-06-02 21:46:34 +00:00
|
|
|
if (pmd_devmap_trans_unstable(vmf->pmd))
|
2016-07-26 22:25:23 +00:00
|
|
|
return 0;
|
|
|
|
/*
|
|
|
|
* A regular pmd is established and it can't morph into a huge
|
|
|
|
* pmd from under us anymore at this point because we hold the
|
2020-06-09 04:33:54 +00:00
|
|
|
* mmap_lock read mode and khugepaged takes it in write mode.
|
2016-07-26 22:25:23 +00:00
|
|
|
* So now it's safe to run pte_offset_map().
|
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
|
2016-12-14 23:07:16 +00:00
|
|
|
vmf->orig_pte = *vmf->pte;
|
2022-05-13 03:22:52 +00:00
|
|
|
vmf->flags |= FAULT_FLAG_ORIG_PTE_VALID;
|
2016-07-26 22:25:23 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* some architectures can have larger ptes than wordsize,
|
|
|
|
* e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
|
locking/atomics, mm: Convert ACCESS_ONCE() to READ_ONCE()/WRITE_ONCE()
For several reasons, it is desirable to use {READ,WRITE}_ONCE() in
preference to ACCESS_ONCE(), and new code is expected to use one of the
former. So far, there's been no reason to change most existing uses of
ACCESS_ONCE(), as these aren't currently harmful.
However, for some features it is necessary to instrument reads and
writes separately, which is not possible with ACCESS_ONCE(). This
distinction is critical to correct operation.
It's possible to transform the bulk of kernel code using the Coccinelle
script below. However, this doesn't handle comments, leaving references
to ACCESS_ONCE() instances which have been removed. As a preparatory
step, this patch converts the mm code and comments to use
{READ,WRITE}_ONCE() consistently.
----
virtual patch
@ depends on patch @
expression E1, E2;
@@
- ACCESS_ONCE(E1) = E2
+ WRITE_ONCE(E1, E2)
@ depends on patch @
expression E;
@@
- ACCESS_ONCE(E)
+ READ_ONCE(E)
----
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: davem@davemloft.net
Cc: linux-arch@vger.kernel.org
Cc: mpe@ellerman.id.au
Cc: shuah@kernel.org
Cc: snitzer@redhat.com
Cc: thor.thayer@linux.intel.com
Cc: tj@kernel.org
Cc: viro@zeniv.linux.org.uk
Link: http://lkml.kernel.org/r/1508792849-3115-15-git-send-email-paulmck@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-10-23 21:07:25 +00:00
|
|
|
* CONFIG_32BIT=y, so READ_ONCE cannot guarantee atomic
|
|
|
|
* accesses. The code below just needs a consistent view
|
|
|
|
* for the ifs and we later double check anyway with the
|
2016-07-26 22:25:23 +00:00
|
|
|
* ptl lock held. So here a barrier will do.
|
|
|
|
*/
|
|
|
|
barrier();
|
2016-12-14 23:07:16 +00:00
|
|
|
if (pte_none(vmf->orig_pte)) {
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap(vmf->pte);
|
|
|
|
vmf->pte = NULL;
|
[PATCH] mm: page fault handlers tidyup
Impose a little more consistency on the page fault handlers do_wp_page,
do_swap_page, do_anonymous_page, do_no_page, do_file_page: why not pass their
arguments in the same order, called the same names?
break_cow is all very well, but what it did was inlined elsewhere: easier to
compare if it's brought back into do_wp_page.
do_file_page's fallback to do_no_page dates from a time when we were testing
pte_file by using it wherever possible: currently it's peculiar to nonlinear
vmas, so just check that. BUG_ON if not? Better not, it's probably page
table corruption, so just show the pte: hmm, there's a pte_ERROR macro, let's
use that for do_wp_page's invalid pfn too.
Hah! Someone in the ppc64 world noticed pte_ERROR was unused so removed it:
restored (and say "pud" not "pmd" in its pud_ERROR).
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 01:15:59 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
mm/uffd: UFFD_FEATURE_WP_UNPOPULATED
Patch series "mm/uffd: Add feature bit UFFD_FEATURE_WP_UNPOPULATED", v4.
The new feature bit makes anonymous memory acts the same as file memory on
userfaultfd-wp in that it'll also wr-protect none ptes.
It can be useful in two cases:
(1) Uffd-wp app that needs to wr-protect none ptes like QEMU snapshot,
so pre-fault can be replaced by enabling this flag and speed up
protections
(2) It helps to implement async uffd-wp mode that Muhammad is working on [1]
It's debatable whether this is the most ideal solution because with the
new feature bit set, wr-protect none pte needs to pre-populate the
pgtables to the last level (PAGE_SIZE). But it seems fine so far to
service either purpose above, so we can leave optimizations for later.
The series brings pte markers to anonymous memory too. There's some
change in the common mm code path in the 1st patch, great to have some eye
looking at it, but hopefully they're still relatively straightforward.
This patch (of 2):
This is a new feature that controls how uffd-wp handles none ptes. When
it's set, the kernel will handle anonymous memory the same way as file
memory, by allowing the user to wr-protect unpopulated ptes.
File memories handles none ptes consistently by allowing wr-protecting of
none ptes because of the unawareness of page cache being exist or not.
For anonymous it was not as persistent because we used to assume that we
don't need protections on none ptes or known zero pages.
One use case of such a feature bit was VM live snapshot, where if without
wr-protecting empty ptes the snapshot can contain random rubbish in the
holes of the anonymous memory, which can cause misbehave of the guest when
the guest OS assumes the pages should be all zeros.
QEMU worked it around by pre-populate the section with reads to fill in
zero page entries before starting the whole snapshot process [1].
Recently there's another need raised on using userfaultfd wr-protect for
detecting dirty pages (to replace soft-dirty in some cases) [2]. In that
case if without being able to wr-protect none ptes by default, the dirty
info can get lost, since we cannot treat every none pte to be dirty (the
current design is identify a page dirty based on uffd-wp bit being
cleared).
In general, we want to be able to wr-protect empty ptes too even for
anonymous.
This patch implements UFFD_FEATURE_WP_UNPOPULATED so that it'll make
uffd-wp handling on none ptes being consistent no matter what the memory
type is underneath. It doesn't have any impact on file memories so far
because we already have pte markers taking care of that. So it only
affects anonymous.
The feature bit is by default off, so the old behavior will be maintained.
Sometimes it may be wanted because the wr-protect of none ptes will
contain overheads not only during UFFDIO_WRITEPROTECT (by applying pte
markers to anonymous), but also on creating the pgtables to store the pte
markers. So there's potentially less chance of using thp on the first
fault for a none pmd or larger than a pmd.
The major implementation part is teaching the whole kernel to understand
pte markers even for anonymously mapped ranges, meanwhile allowing the
UFFDIO_WRITEPROTECT ioctl to apply pte markers for anonymous too when the
new feature bit is set.
Note that even if the patch subject starts with mm/uffd, there're a few
small refactors to major mm path of handling anonymous page faults. But
they should be straightforward.
With WP_UNPOPUATED, application like QEMU can avoid pre-read faults all
the memory before wr-protect during taking a live snapshot. Quotting from
Muhammad's test result here [3] based on a simple program [4]:
(1) With huge page disabled
echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 1111453 (pre-fault 1101011)
Test MADVISE: 278276 (pre-fault 266378)
Test WP-UNPOPULATE: 11712
(2) With Huge page enabled
echo always > /sys/kernel/mm/transparent_hugepage/enabled
./uffd_wp_perf
Test DEFAULT: 4
Test PRE-READ: 22521 (pre-fault 22348)
Test MADVISE: 4909 (pre-fault 4743)
Test WP-UNPOPULATE: 14448
There'll be a great perf boost for no-thp case, while for thp enabled with
extreme case of all-thp-zero WP_UNPOPULATED can be slower than MADVISE,
but that's low possibility in reality, also the overhead was not reduced
but postponed until a follow up write on any huge zero thp, so potentially
it is faster by making the follow up writes slower.
[1] https://lore.kernel.org/all/20210401092226.102804-4-andrey.gruzdev@virtuozzo.com/
[2] https://lore.kernel.org/all/Y+v2HJ8+3i%2FKzDBu@x1n/
[3] https://lore.kernel.org/all/d0eb0a13-16dc-1ac1-653a-78b7273781e3@collabora.com/
[4] https://github.com/xzpeter/clibs/blob/master/uffd-test/uffd-wp-perf.c
[peterx@redhat.com: comment changes, oneliner fix to khugepaged]
Link: https://lkml.kernel.org/r/ZB2/8jPhD3fpx5U8@x1n
Link: https://lkml.kernel.org/r/20230309223711.823547-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230309223711.823547-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Paul Gofman <pgofman@codeweavers.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-09 22:37:10 +00:00
|
|
|
if (!vmf->pte)
|
|
|
|
return do_pte_missing(vmf);
|
2016-07-26 22:25:23 +00:00
|
|
|
|
2016-12-14 23:07:16 +00:00
|
|
|
if (!pte_present(vmf->orig_pte))
|
|
|
|
return do_swap_page(vmf);
|
2016-07-26 22:25:23 +00:00
|
|
|
|
2016-12-14 23:07:16 +00:00
|
|
|
if (pte_protnone(vmf->orig_pte) && vma_is_accessible(vmf->vma))
|
|
|
|
return do_numa_page(vmf);
|
2012-10-25 12:16:31 +00:00
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
|
|
|
|
spin_lock(vmf->ptl);
|
2016-12-14 23:07:16 +00:00
|
|
|
entry = vmf->orig_pte;
|
2020-05-27 02:25:18 +00:00
|
|
|
if (unlikely(!pte_same(*vmf->pte, entry))) {
|
|
|
|
update_mmu_tlb(vmf->vma, vmf->address, vmf->pte);
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
goto unlock;
|
2020-05-27 02:25:18 +00:00
|
|
|
}
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
if (vmf->flags & (FAULT_FLAG_WRITE|FAULT_FLAG_UNSHARE)) {
|
Revert "mm: replace p??_write with pte_access_permitted in fault + gup paths"
This reverts commits 5c9d2d5c269c, c7da82b894e9, and e7fe7b5cae90.
We'll probably need to revisit this, but basically we should not
complicate the get_user_pages_fast() case, and checking the actual page
table protection key bits will require more care anyway, since the
protection keys depend on the exact state of the VM in question.
Particularly when doing a "remote" page lookup (ie in somebody elses VM,
not your own), you need to be much more careful than this was. Dave
Hansen says:
"So, the underlying bug here is that we now a get_user_pages_remote()
and then go ahead and do the p*_access_permitted() checks against the
current PKRU. This was introduced recently with the addition of the
new p??_access_permitted() calls.
We have checks in the VMA path for the "remote" gups and we avoid
consulting PKRU for them. This got missed in the pkeys selftests
because I did a ptrace read, but not a *write*. I also didn't
explicitly test it against something where a COW needed to be done"
It's also not entirely clear that it makes sense to check the protection
key bits at this level at all. But one possible eventual solution is to
make the get_user_pages_fast() case just abort if it sees protection key
bits set, which makes us fall back to the regular get_user_pages() case,
which then has a vma and can do the check there if we want to.
We'll see.
Somewhat related to this all: what we _do_ want to do some day is to
check the PAGE_USER bit - it should obviously always be set for user
pages, but it would be a good check to have back. Because we have no
generic way to test for it, we lost it as part of moving over from the
architecture-specific x86 GUP implementation to the generic one in
commit e585513b76f7 ("x86/mm/gup: Switch GUP to the generic
get_user_page_fast() implementation").
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-12-16 02:53:22 +00:00
|
|
|
if (!pte_write(entry))
|
2016-12-14 23:07:16 +00:00
|
|
|
return do_wp_page(vmf);
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
else if (likely(vmf->flags & FAULT_FLAG_WRITE))
|
|
|
|
entry = pte_mkdirty(entry);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
entry = pte_mkyoung(entry);
|
2016-12-14 23:06:58 +00:00
|
|
|
if (ptep_set_access_flags(vmf->vma, vmf->address, vmf->pte, entry,
|
|
|
|
vmf->flags & FAULT_FLAG_WRITE)) {
|
|
|
|
update_mmu_cache(vmf->vma, vmf->address, vmf->pte);
|
2005-10-30 01:16:48 +00:00
|
|
|
} else {
|
2020-08-15 04:30:41 +00:00
|
|
|
/* Skip spurious TLB flush for retried page fault */
|
|
|
|
if (vmf->flags & FAULT_FLAG_TRIED)
|
|
|
|
goto unlock;
|
2005-10-30 01:16:48 +00:00
|
|
|
/*
|
|
|
|
* This is needed only for protection faults but the arch code
|
|
|
|
* is not yet telling us if this is a protection fault or not.
|
|
|
|
* This still avoids useless tlb flushes for .text page faults
|
|
|
|
* with threads.
|
|
|
|
*/
|
2016-12-14 23:06:58 +00:00
|
|
|
if (vmf->flags & FAULT_FLAG_WRITE)
|
2023-03-06 16:15:48 +00:00
|
|
|
flush_tlb_fix_spurious_fault(vmf->vma, vmf->address,
|
|
|
|
vmf->pte);
|
2005-10-30 01:16:48 +00:00
|
|
|
}
|
[PATCH] mm: page fault handler locking
On the page fault path, the patch before last pushed acquiring the
page_table_lock down to the head of handle_pte_fault (though it's also taken
and dropped earlier when a new page table has to be allocated).
Now delete that line, read "entry = *pte" without it, and go off to this or
that page fault handler on the basis of this unlocked peek. Usually the
handler can proceed without the lock, relying on the subsequent locked
pte_same or pte_none test to back out when necessary; though do_wp_page needs
the lock immediately, and do_file_page doesn't check (if there's a race,
install_page just zaps the entry and reinstalls it).
But on those architectures (notably i386 with PAE) whose pte is too big to be
read atomically, if SMP or preemption is enabled, do_swap_page and
do_file_page might cause irretrievable damage if passed a Frankenstein entry
stitched together from unrelated parts. In those configs, "pte_unmap_same"
has to take page_table_lock, validate orig_pte still the same, and drop
page_table_lock before unmapping, before proceeding.
Use pte_offset_map_lock and pte_unmap_unlock throughout the handlers; but lock
avoidance leaves more lone maps and unmaps than elsewhere.
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 01:16:26 +00:00
|
|
|
unlock:
|
2016-12-14 23:06:58 +00:00
|
|
|
pte_unmap_unlock(vmf->pte, vmf->ptl);
|
2007-07-19 08:47:05 +00:00
|
|
|
return 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* By the time we get here, we already hold the mm semaphore
|
2014-08-06 23:07:24 +00:00
|
|
|
*
|
2020-06-09 04:33:54 +00:00
|
|
|
* The mmap_lock may have been released depending on flags and our
|
2021-03-19 01:39:45 +00:00
|
|
|
* return value. See filemap_fault() and __folio_lock_or_retry().
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,
|
|
|
|
unsigned long address, unsigned int flags)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2016-12-14 23:06:58 +00:00
|
|
|
struct vm_fault vmf = {
|
2016-07-26 22:25:20 +00:00
|
|
|
.vma = vma,
|
2016-12-14 23:07:01 +00:00
|
|
|
.address = address & PAGE_MASK,
|
userfaultfd: provide unmasked address on page-fault
Userfaultfd is supposed to provide the full address (i.e., unmasked) of
the faulting access back to userspace. However, that is not the case for
quite some time.
Even running "userfaultfd_demo" from the userfaultfd man page provides the
wrong output (and contradicts the man page). Notice that
"UFFD_EVENT_PAGEFAULT event" shows the masked address (7fc5e30b3000) and
not the first read address (0x7fc5e30b300f).
Address returned by mmap() = 0x7fc5e30b3000
fault_handler_thread():
poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fc5e30b3000
(uffdio_copy.copy returned 4096)
Read address 0x7fc5e30b300f in main(): A
Read address 0x7fc5e30b340f in main(): A
Read address 0x7fc5e30b380f in main(): A
Read address 0x7fc5e30b3c0f in main(): A
The exact address is useful for various reasons and specifically for
prefetching decisions. If it is known that the memory is populated by
certain objects whose size is not page-aligned, then based on the faulting
address, the uffd-monitor can decide whether to prefetch and prefault the
adjacent page.
This bug has been for quite some time in the kernel: since commit
1a29d85eb0f1 ("mm: use vmf->address instead of of vmf->virtual_address")
vmf->virtual_address"), which dates back to 2016. A concern has been
raised that existing userspace application might rely on the old/wrong
behavior in which the address is masked. Therefore, it was suggested to
provide the masked address unless the user explicitly asks for the exact
address.
Add a new userfaultfd feature UFFD_FEATURE_EXACT_ADDRESS to direct
userfaultfd to provide the exact address. Add a new "real_address" field
to vmf to hold the unmasked address. Provide the address to userspace
accordingly.
Initialize real_address in various code-paths to be consistent with
address, even when it is not used, to be on the safe side.
[namit@vmware.com: initialize real_address on all code paths, per Jan]
Link: https://lkml.kernel.org/r/20220226022655.350562-1-namit@vmware.com
[akpm@linux-foundation.org: fix typo in comment, per Jan]
Link: https://lkml.kernel.org/r/20220218041003.3508-1-namit@vmware.com
Signed-off-by: Nadav Amit <namit@vmware.com>
Acked-by: Peter Xu <peterx@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-03-22 21:45:32 +00:00
|
|
|
.real_address = address,
|
2016-07-26 22:25:20 +00:00
|
|
|
.flags = flags,
|
2016-12-14 23:07:04 +00:00
|
|
|
.pgoff = linear_page_index(vma, address),
|
2016-12-14 23:07:07 +00:00
|
|
|
.gfp_mask = __get_fault_gfp_mask(vma),
|
2016-07-26 22:25:20 +00:00
|
|
|
};
|
2016-07-26 22:25:18 +00:00
|
|
|
struct mm_struct *mm = vma->vm_mm;
|
2022-06-16 17:48:38 +00:00
|
|
|
unsigned long vm_flags = vma->vm_flags;
|
2005-04-16 22:20:36 +00:00
|
|
|
pgd_t *pgd;
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d_t *p4d;
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
pgd = pgd_offset(mm, address);
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d = p4d_alloc(mm, pgd, address);
|
|
|
|
if (!p4d)
|
|
|
|
return VM_FAULT_OOM;
|
2017-02-24 22:57:02 +00:00
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
vmf.pud = pud_alloc(mm, p4d, address);
|
2017-02-24 22:57:02 +00:00
|
|
|
if (!vmf.pud)
|
2005-10-30 01:16:23 +00:00
|
|
|
return VM_FAULT_OOM;
|
2019-12-01 01:51:32 +00:00
|
|
|
retry_pud:
|
2022-06-16 17:48:38 +00:00
|
|
|
if (pud_none(*vmf.pud) &&
|
2022-07-06 23:59:25 +00:00
|
|
|
hugepage_vma_check(vma, vm_flags, false, true, true)) {
|
2017-02-24 22:57:02 +00:00
|
|
|
ret = create_huge_pud(&vmf);
|
|
|
|
if (!(ret & VM_FAULT_FALLBACK))
|
|
|
|
return ret;
|
|
|
|
} else {
|
|
|
|
pud_t orig_pud = *vmf.pud;
|
|
|
|
|
|
|
|
barrier();
|
|
|
|
if (pud_trans_huge(orig_pud) || pud_devmap(orig_pud)) {
|
|
|
|
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
/*
|
|
|
|
* TODO once we support anonymous PUDs: NUMA case and
|
|
|
|
* FAULT_FLAG_UNSHARE handling.
|
|
|
|
*/
|
|
|
|
if ((flags & FAULT_FLAG_WRITE) && !pud_write(orig_pud)) {
|
2017-02-24 22:57:02 +00:00
|
|
|
ret = wp_huge_pud(&vmf, orig_pud);
|
|
|
|
if (!(ret & VM_FAULT_FALLBACK))
|
|
|
|
return ret;
|
|
|
|
} else {
|
|
|
|
huge_pud_set_accessed(&vmf, orig_pud);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
vmf.pmd = pmd_alloc(mm, vmf.pud, address);
|
2016-12-14 23:06:58 +00:00
|
|
|
if (!vmf.pmd)
|
2005-10-30 01:16:23 +00:00
|
|
|
return VM_FAULT_OOM;
|
2019-12-01 01:51:32 +00:00
|
|
|
|
|
|
|
/* Huge pud page fault raced with pmd_alloc? */
|
|
|
|
if (pud_trans_unstable(vmf.pud))
|
|
|
|
goto retry_pud;
|
|
|
|
|
2022-06-16 17:48:38 +00:00
|
|
|
if (pmd_none(*vmf.pmd) &&
|
2022-07-06 23:59:25 +00:00
|
|
|
hugepage_vma_check(vma, vm_flags, false, true, true)) {
|
mm,fs,dax: change ->pmd_fault to ->huge_fault
Patch series "1G transparent hugepage support for device dax", v2.
The following series implements support for 1G trasparent hugepage on
x86 for device dax. The bulk of the code was written by Mathew Wilcox a
while back supporting transparent 1G hugepage for fs DAX. I have
forward ported the relevant bits to 4.10-rc. The current submission has
only the necessary code to support device DAX.
Comments from Dan Williams: So the motivation and intended user of this
functionality mirrors the motivation and users of 1GB page support in
hugetlbfs. Given expected capacities of persistent memory devices an
in-memory database may want to reduce tlb pressure beyond what they can
already achieve with 2MB mappings of a device-dax file. We have
customer feedback to that effect as Willy mentioned in his previous
version of these patches [1].
[1]: https://lkml.org/lkml/2016/1/31/52
Comments from Nilesh @ Oracle:
There are applications which have a process model; and if you assume
10,000 processes attempting to mmap all the 6TB memory available on a
server; we are looking at the following:
processes : 10,000
memory : 6TB
pte @ 4k page size: 8 bytes / 4K of memory * #processes = 6TB / 4k * 8 * 10000 = 1.5GB * 80000 = 120,000GB
pmd @ 2M page size: 120,000 / 512 = ~240GB
pud @ 1G page size: 240GB / 512 = ~480MB
As you can see with 2M pages, this system will use up an exorbitant
amount of DRAM to hold the page tables; but the 1G pages finally brings
it down to a reasonable level. Memory sizes will keep increasing; so
this number will keep increasing.
An argument can be made to convert the applications from process model
to thread model, but in the real world that may not be always practical.
Hopefully this helps explain the use case where this is valuable.
This patch (of 3):
In preparation for adding the ability to handle PUD pages, convert
vm_operations_struct.pmd_fault to vm_operations_struct.huge_fault. The
vm_fault structure is extended to include a union of the different page
table pointers that may be needed, and three flag bits are reserved to
indicate which type of pointer is in the union.
[ross.zwisler@linux.intel.com: remove unused function ext4_dax_huge_fault()]
Link: http://lkml.kernel.org/r/1485813172-7284-1-git-send-email-ross.zwisler@linux.intel.com
[dave.jiang@intel.com: clear PMD or PUD size flags when in fall through path]
Link: http://lkml.kernel.org/r/148589842696.5820.16078080610311444794.stgit@djiang5-desk3.ch.intel.com
Link: http://lkml.kernel.org/r/148545058784.17912.6353162518188733642.stgit@djiang5-desk3.ch.intel.com
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jan Kara <jack@suse.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Nilesh Choudhury <nilesh.choudhury@oracle.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-02-24 22:56:59 +00:00
|
|
|
ret = create_huge_pmd(&vmf);
|
2013-09-12 22:14:05 +00:00
|
|
|
if (!(ret & VM_FAULT_FALLBACK))
|
|
|
|
return ret;
|
thp: transparent hugepage core
Lately I've been working to make KVM use hugepages transparently without
the usual restrictions of hugetlbfs. Some of the restrictions I'd like to
see removed:
1) hugepages have to be swappable or the guest physical memory remains
locked in RAM and can't be paged out to swap
2) if a hugepage allocation fails, regular pages should be allocated
instead and mixed in the same vma without any failure and without
userland noticing
3) if some task quits and more hugepages become available in the
buddy, guest physical memory backed by regular pages should be
relocated on hugepages automatically in regions under
madvise(MADV_HUGEPAGE) (ideally event driven by waking up the
kernel deamon if the order=HPAGE_PMD_SHIFT-PAGE_SHIFT list becomes
not null)
4) avoidance of reservation and maximization of use of hugepages whenever
possible. Reservation (needed to avoid runtime fatal faliures) may be ok for
1 machine with 1 database with 1 database cache with 1 database cache size
known at boot time. It's definitely not feasible with a virtualization
hypervisor usage like RHEV-H that runs an unknown number of virtual machines
with an unknown size of each virtual machine with an unknown amount of
pagecache that could be potentially useful in the host for guest not using
O_DIRECT (aka cache=off).
hugepages in the virtualization hypervisor (and also in the guest!) are
much more important than in a regular host not using virtualization,
becasue with NPT/EPT they decrease the tlb-miss cacheline accesses from 24
to 19 in case only the hypervisor uses transparent hugepages, and they
decrease the tlb-miss cacheline accesses from 19 to 15 in case both the
linux hypervisor and the linux guest both uses this patch (though the
guest will limit the addition speedup to anonymous regions only for
now...). Even more important is that the tlb miss handler is much slower
on a NPT/EPT guest than for a regular shadow paging or no-virtualization
scenario. So maximizing the amount of virtual memory cached by the TLB
pays off significantly more with NPT/EPT than without (even if there would
be no significant speedup in the tlb-miss runtime).
The first (and more tedious) part of this work requires allowing the VM to
handle anonymous hugepages mixed with regular pages transparently on
regular anonymous vmas. This is what this patch tries to achieve in the
least intrusive possible way. We want hugepages and hugetlb to be used in
a way so that all applications can benefit without changes (as usual we
leverage the KVM virtualization design: by improving the Linux VM at
large, KVM gets the performance boost too).
The most important design choice is: always fallback to 4k allocation if
the hugepage allocation fails! This is the _very_ opposite of some large
pagecache patches that failed with -EIO back then if a 64k (or similar)
allocation failed...
Second important decision (to reduce the impact of the feature on the
existing pagetable handling code) is that at any time we can split an
hugepage into 512 regular pages and it has to be done with an operation
that can't fail. This way the reliability of the swapping isn't decreased
(no need to allocate memory when we are short on memory to swap) and it's
trivial to plug a split_huge_page* one-liner where needed without
polluting the VM. Over time we can teach mprotect, mremap and friends to
handle pmd_trans_huge natively without calling split_huge_page*. The fact
it can't fail isn't just for swap: if split_huge_page would return -ENOMEM
(instead of the current void) we'd need to rollback the mprotect from the
middle of it (ideally including undoing the split_vma) which would be a
big change and in the very wrong direction (it'd likely be simpler not to
call split_huge_page at all and to teach mprotect and friends to handle
hugepages instead of rolling them back from the middle). In short the
very value of split_huge_page is that it can't fail.
The collapsing and madvise(MADV_HUGEPAGE) part will remain separated and
incremental and it'll just be an "harmless" addition later if this initial
part is agreed upon. It also should be noted that locking-wise replacing
regular pages with hugepages is going to be very easy if compared to what
I'm doing below in split_huge_page, as it will only happen when
page_count(page) matches page_mapcount(page) if we can take the PG_lock
and mmap_sem in write mode. collapse_huge_page will be a "best effort"
that (unlike split_huge_page) can fail at the minimal sign of trouble and
we can try again later. collapse_huge_page will be similar to how KSM
works and the madvise(MADV_HUGEPAGE) will work similar to
madvise(MADV_MERGEABLE).
The default I like is that transparent hugepages are used at page fault
time. This can be changed with
/sys/kernel/mm/transparent_hugepage/enabled. The control knob can be set
to three values "always", "madvise", "never" which mean respectively that
hugepages are always used, or only inside madvise(MADV_HUGEPAGE) regions,
or never used. /sys/kernel/mm/transparent_hugepage/defrag instead
controls if the hugepage allocation should defrag memory aggressively
"always", only inside "madvise" regions, or "never".
The pmd_trans_splitting/pmd_trans_huge locking is very solid. The
put_page (from get_user_page users that can't use mmu notifier like
O_DIRECT) that runs against a __split_huge_page_refcount instead was a
pain to serialize in a way that would result always in a coherent page
count for both tail and head. I think my locking solution with a
compound_lock taken only after the page_first is valid and is still a
PageHead should be safe but it surely needs review from SMP race point of
view. In short there is no current existing way to serialize the O_DIRECT
final put_page against split_huge_page_refcount so I had to invent a new
one (O_DIRECT loses knowledge on the mapping status by the time gup_fast
returns so...). And I didn't want to impact all gup/gup_fast users for
now, maybe if we change the gup interface substantially we can avoid this
locking, I admit I didn't think too much about it because changing the gup
unpinning interface would be invasive.
If we ignored O_DIRECT we could stick to the existing compound refcounting
code, by simply adding a get_user_pages_fast_flags(foll_flags) where KVM
(and any other mmu notifier user) would call it without FOLL_GET (and if
FOLL_GET isn't set we'd just BUG_ON if nobody registered itself in the
current task mmu notifier list yet). But O_DIRECT is fundamental for
decent performance of virtualized I/O on fast storage so we can't avoid it
to solve the race of put_page against split_huge_page_refcount to achieve
a complete hugepage feature for KVM.
Swap and oom works fine (well just like with regular pages ;). MMU
notifier is handled transparently too, with the exception of the young bit
on the pmd, that didn't have a range check but I think KVM will be fine
because the whole point of hugepages is that EPT/NPT will also use a huge
pmd when they notice gup returns pages with PageCompound set, so they
won't care of a range and there's just the pmd young bit to check in that
case.
NOTE: in some cases if the L2 cache is small, this may slowdown and waste
memory during COWs because 4M of memory are accessed in a single fault
instead of 8k (the payoff is that after COW the program can run faster).
So we might want to switch the copy_huge_page (and clear_huge_page too) to
not temporal stores. I also extensively researched ways to avoid this
cache trashing with a full prefault logic that would cow in 8k/16k/32k/64k
up to 1M (I can send those patches that fully implemented prefault) but I
concluded they're not worth it and they add an huge additional complexity
and they remove all tlb benefits until the full hugepage has been faulted
in, to save a little bit of memory and some cache during app startup, but
they still don't improve substantially the cache-trashing during startup
if the prefault happens in >4k chunks. One reason is that those 4k pte
entries copied are still mapped on a perfectly cache-colored hugepage, so
the trashing is the worst one can generate in those copies (cow of 4k page
copies aren't so well colored so they trashes less, but again this results
in software running faster after the page fault). Those prefault patches
allowed things like a pte where post-cow pages were local 4k regular anon
pages and the not-yet-cowed pte entries were pointing in the middle of
some hugepage mapped read-only. If it doesn't payoff substantially with
todays hardware it will payoff even less in the future with larger l2
caches, and the prefault logic would blot the VM a lot. If one is
emebdded transparent_hugepage can be disabled during boot with sysfs or
with the boot commandline parameter transparent_hugepage=0 (or
transparent_hugepage=2 to restrict hugepages inside madvise regions) that
will ensure not a single hugepage is allocated at boot time. It is simple
enough to just disable transparent hugepage globally and let transparent
hugepages be allocated selectively by applications in the MADV_HUGEPAGE
region (both at page fault time, and if enabled with the
collapse_huge_page too through the kernel daemon).
This patch supports only hugepages mapped in the pmd, archs that have
smaller hugepages will not fit in this patch alone. Also some archs like
power have certain tlb limits that prevents mixing different page size in
the same regions so they will not fit in this framework that requires
"graceful fallback" to basic PAGE_SIZE in case of physical memory
fragmentation. hugetlbfs remains a perfect fit for those because its
software limits happen to match the hardware limits. hugetlbfs also
remains a perfect fit for hugepage sizes like 1GByte that cannot be hoped
to be found not fragmented after a certain system uptime and that would be
very expensive to defragment with relocation, so requiring reservation.
hugetlbfs is the "reservation way", the point of transparent hugepages is
not to have any reservation at all and maximizing the use of cache and
hugepages at all times automatically.
Some performance result:
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largep
ages3
memset page fault 1566023
memset tlb miss 453854
memset second tlb miss 453321
random access tlb miss 41635
random access second tlb miss 41658
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largepages3
memset page fault 1566471
memset tlb miss 453375
memset second tlb miss 453320
random access tlb miss 41636
random access second tlb miss 41637
vmx andrea # ./largepages3
memset page fault 1566642
memset tlb miss 453417
memset second tlb miss 453313
random access tlb miss 41630
random access second tlb miss 41647
vmx andrea # ./largepages3
memset page fault 1566872
memset tlb miss 453418
memset second tlb miss 453315
random access tlb miss 41618
random access second tlb miss 41659
vmx andrea # echo 0 > /proc/sys/vm/transparent_hugepage
vmx andrea # ./largepages3
memset page fault 2182476
memset tlb miss 460305
memset second tlb miss 460179
random access tlb miss 44483
random access second tlb miss 44186
vmx andrea # ./largepages3
memset page fault 2182791
memset tlb miss 460742
memset second tlb miss 459962
random access tlb miss 43981
random access second tlb miss 43988
============
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#define SIZE (3UL*1024*1024*1024)
int main()
{
char *p = malloc(SIZE), *p2;
struct timeval before, after;
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset page fault %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
return 0;
}
============
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:46:52 +00:00
|
|
|
} else {
|
mm: memory: add orig_pmd to struct vm_fault
Pach series "mm: thp: use generic THP migration for NUMA hinting fault", v3.
When the THP NUMA fault support was added THP migration was not supported
yet. So the ad hoc THP migration was implemented in NUMA fault handling.
Since v4.14 THP migration has been supported so it doesn't make too much
sense to still keep another THP migration implementation rather than using
the generic migration code. It is definitely a maintenance burden to keep
two THP migration implementation for different code paths and it is more
error prone. Using the generic THP migration implementation allows us
remove the duplicate code and some hacks needed by the old ad hoc
implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both
THP and NUMA balancing. The most of them support THP migration except for
S390. Zi Yan tried to add THP migration support for S390 before but it
was not accepted due to the design of S390 PMD. For the discussion,
please see: https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip
huge PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't
figure out if they have been removed or not since I just found FOLL_NUMA
code in the current gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page. The test
VM has 80 vcpus and 64G memory. The test would create 2 processes to
consume 128G memory together which would incur memory pressure to cause
THP splits. And it also creates 80 processes to hog cpu, and the memory
consumer processes are bound to different nodes periodically in order to
increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page is
as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted the
raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split is
less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory pressure.
No obvious regression is spotted. The below is the test result *w/o*
memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
This patch (of 7):
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge
page fault could be removed, just like its PTE counterpart does.
Link: https://lkml.kernel.org/r/20210518200801.7413-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20210518200801.7413-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:51:35 +00:00
|
|
|
vmf.orig_pmd = *vmf.pmd;
|
2012-05-29 22:06:23 +00:00
|
|
|
|
thp: transparent hugepage core
Lately I've been working to make KVM use hugepages transparently without
the usual restrictions of hugetlbfs. Some of the restrictions I'd like to
see removed:
1) hugepages have to be swappable or the guest physical memory remains
locked in RAM and can't be paged out to swap
2) if a hugepage allocation fails, regular pages should be allocated
instead and mixed in the same vma without any failure and without
userland noticing
3) if some task quits and more hugepages become available in the
buddy, guest physical memory backed by regular pages should be
relocated on hugepages automatically in regions under
madvise(MADV_HUGEPAGE) (ideally event driven by waking up the
kernel deamon if the order=HPAGE_PMD_SHIFT-PAGE_SHIFT list becomes
not null)
4) avoidance of reservation and maximization of use of hugepages whenever
possible. Reservation (needed to avoid runtime fatal faliures) may be ok for
1 machine with 1 database with 1 database cache with 1 database cache size
known at boot time. It's definitely not feasible with a virtualization
hypervisor usage like RHEV-H that runs an unknown number of virtual machines
with an unknown size of each virtual machine with an unknown amount of
pagecache that could be potentially useful in the host for guest not using
O_DIRECT (aka cache=off).
hugepages in the virtualization hypervisor (and also in the guest!) are
much more important than in a regular host not using virtualization,
becasue with NPT/EPT they decrease the tlb-miss cacheline accesses from 24
to 19 in case only the hypervisor uses transparent hugepages, and they
decrease the tlb-miss cacheline accesses from 19 to 15 in case both the
linux hypervisor and the linux guest both uses this patch (though the
guest will limit the addition speedup to anonymous regions only for
now...). Even more important is that the tlb miss handler is much slower
on a NPT/EPT guest than for a regular shadow paging or no-virtualization
scenario. So maximizing the amount of virtual memory cached by the TLB
pays off significantly more with NPT/EPT than without (even if there would
be no significant speedup in the tlb-miss runtime).
The first (and more tedious) part of this work requires allowing the VM to
handle anonymous hugepages mixed with regular pages transparently on
regular anonymous vmas. This is what this patch tries to achieve in the
least intrusive possible way. We want hugepages and hugetlb to be used in
a way so that all applications can benefit without changes (as usual we
leverage the KVM virtualization design: by improving the Linux VM at
large, KVM gets the performance boost too).
The most important design choice is: always fallback to 4k allocation if
the hugepage allocation fails! This is the _very_ opposite of some large
pagecache patches that failed with -EIO back then if a 64k (or similar)
allocation failed...
Second important decision (to reduce the impact of the feature on the
existing pagetable handling code) is that at any time we can split an
hugepage into 512 regular pages and it has to be done with an operation
that can't fail. This way the reliability of the swapping isn't decreased
(no need to allocate memory when we are short on memory to swap) and it's
trivial to plug a split_huge_page* one-liner where needed without
polluting the VM. Over time we can teach mprotect, mremap and friends to
handle pmd_trans_huge natively without calling split_huge_page*. The fact
it can't fail isn't just for swap: if split_huge_page would return -ENOMEM
(instead of the current void) we'd need to rollback the mprotect from the
middle of it (ideally including undoing the split_vma) which would be a
big change and in the very wrong direction (it'd likely be simpler not to
call split_huge_page at all and to teach mprotect and friends to handle
hugepages instead of rolling them back from the middle). In short the
very value of split_huge_page is that it can't fail.
The collapsing and madvise(MADV_HUGEPAGE) part will remain separated and
incremental and it'll just be an "harmless" addition later if this initial
part is agreed upon. It also should be noted that locking-wise replacing
regular pages with hugepages is going to be very easy if compared to what
I'm doing below in split_huge_page, as it will only happen when
page_count(page) matches page_mapcount(page) if we can take the PG_lock
and mmap_sem in write mode. collapse_huge_page will be a "best effort"
that (unlike split_huge_page) can fail at the minimal sign of trouble and
we can try again later. collapse_huge_page will be similar to how KSM
works and the madvise(MADV_HUGEPAGE) will work similar to
madvise(MADV_MERGEABLE).
The default I like is that transparent hugepages are used at page fault
time. This can be changed with
/sys/kernel/mm/transparent_hugepage/enabled. The control knob can be set
to three values "always", "madvise", "never" which mean respectively that
hugepages are always used, or only inside madvise(MADV_HUGEPAGE) regions,
or never used. /sys/kernel/mm/transparent_hugepage/defrag instead
controls if the hugepage allocation should defrag memory aggressively
"always", only inside "madvise" regions, or "never".
The pmd_trans_splitting/pmd_trans_huge locking is very solid. The
put_page (from get_user_page users that can't use mmu notifier like
O_DIRECT) that runs against a __split_huge_page_refcount instead was a
pain to serialize in a way that would result always in a coherent page
count for both tail and head. I think my locking solution with a
compound_lock taken only after the page_first is valid and is still a
PageHead should be safe but it surely needs review from SMP race point of
view. In short there is no current existing way to serialize the O_DIRECT
final put_page against split_huge_page_refcount so I had to invent a new
one (O_DIRECT loses knowledge on the mapping status by the time gup_fast
returns so...). And I didn't want to impact all gup/gup_fast users for
now, maybe if we change the gup interface substantially we can avoid this
locking, I admit I didn't think too much about it because changing the gup
unpinning interface would be invasive.
If we ignored O_DIRECT we could stick to the existing compound refcounting
code, by simply adding a get_user_pages_fast_flags(foll_flags) where KVM
(and any other mmu notifier user) would call it without FOLL_GET (and if
FOLL_GET isn't set we'd just BUG_ON if nobody registered itself in the
current task mmu notifier list yet). But O_DIRECT is fundamental for
decent performance of virtualized I/O on fast storage so we can't avoid it
to solve the race of put_page against split_huge_page_refcount to achieve
a complete hugepage feature for KVM.
Swap and oom works fine (well just like with regular pages ;). MMU
notifier is handled transparently too, with the exception of the young bit
on the pmd, that didn't have a range check but I think KVM will be fine
because the whole point of hugepages is that EPT/NPT will also use a huge
pmd when they notice gup returns pages with PageCompound set, so they
won't care of a range and there's just the pmd young bit to check in that
case.
NOTE: in some cases if the L2 cache is small, this may slowdown and waste
memory during COWs because 4M of memory are accessed in a single fault
instead of 8k (the payoff is that after COW the program can run faster).
So we might want to switch the copy_huge_page (and clear_huge_page too) to
not temporal stores. I also extensively researched ways to avoid this
cache trashing with a full prefault logic that would cow in 8k/16k/32k/64k
up to 1M (I can send those patches that fully implemented prefault) but I
concluded they're not worth it and they add an huge additional complexity
and they remove all tlb benefits until the full hugepage has been faulted
in, to save a little bit of memory and some cache during app startup, but
they still don't improve substantially the cache-trashing during startup
if the prefault happens in >4k chunks. One reason is that those 4k pte
entries copied are still mapped on a perfectly cache-colored hugepage, so
the trashing is the worst one can generate in those copies (cow of 4k page
copies aren't so well colored so they trashes less, but again this results
in software running faster after the page fault). Those prefault patches
allowed things like a pte where post-cow pages were local 4k regular anon
pages and the not-yet-cowed pte entries were pointing in the middle of
some hugepage mapped read-only. If it doesn't payoff substantially with
todays hardware it will payoff even less in the future with larger l2
caches, and the prefault logic would blot the VM a lot. If one is
emebdded transparent_hugepage can be disabled during boot with sysfs or
with the boot commandline parameter transparent_hugepage=0 (or
transparent_hugepage=2 to restrict hugepages inside madvise regions) that
will ensure not a single hugepage is allocated at boot time. It is simple
enough to just disable transparent hugepage globally and let transparent
hugepages be allocated selectively by applications in the MADV_HUGEPAGE
region (both at page fault time, and if enabled with the
collapse_huge_page too through the kernel daemon).
This patch supports only hugepages mapped in the pmd, archs that have
smaller hugepages will not fit in this patch alone. Also some archs like
power have certain tlb limits that prevents mixing different page size in
the same regions so they will not fit in this framework that requires
"graceful fallback" to basic PAGE_SIZE in case of physical memory
fragmentation. hugetlbfs remains a perfect fit for those because its
software limits happen to match the hardware limits. hugetlbfs also
remains a perfect fit for hugepage sizes like 1GByte that cannot be hoped
to be found not fragmented after a certain system uptime and that would be
very expensive to defragment with relocation, so requiring reservation.
hugetlbfs is the "reservation way", the point of transparent hugepages is
not to have any reservation at all and maximizing the use of cache and
hugepages at all times automatically.
Some performance result:
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largep
ages3
memset page fault 1566023
memset tlb miss 453854
memset second tlb miss 453321
random access tlb miss 41635
random access second tlb miss 41658
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largepages3
memset page fault 1566471
memset tlb miss 453375
memset second tlb miss 453320
random access tlb miss 41636
random access second tlb miss 41637
vmx andrea # ./largepages3
memset page fault 1566642
memset tlb miss 453417
memset second tlb miss 453313
random access tlb miss 41630
random access second tlb miss 41647
vmx andrea # ./largepages3
memset page fault 1566872
memset tlb miss 453418
memset second tlb miss 453315
random access tlb miss 41618
random access second tlb miss 41659
vmx andrea # echo 0 > /proc/sys/vm/transparent_hugepage
vmx andrea # ./largepages3
memset page fault 2182476
memset tlb miss 460305
memset second tlb miss 460179
random access tlb miss 44483
random access second tlb miss 44186
vmx andrea # ./largepages3
memset page fault 2182791
memset tlb miss 460742
memset second tlb miss 459962
random access tlb miss 43981
random access second tlb miss 43988
============
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#define SIZE (3UL*1024*1024*1024)
int main()
{
char *p = malloc(SIZE), *p2;
struct timeval before, after;
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset page fault %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
return 0;
}
============
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:46:52 +00:00
|
|
|
barrier();
|
mm: memory: add orig_pmd to struct vm_fault
Pach series "mm: thp: use generic THP migration for NUMA hinting fault", v3.
When the THP NUMA fault support was added THP migration was not supported
yet. So the ad hoc THP migration was implemented in NUMA fault handling.
Since v4.14 THP migration has been supported so it doesn't make too much
sense to still keep another THP migration implementation rather than using
the generic migration code. It is definitely a maintenance burden to keep
two THP migration implementation for different code paths and it is more
error prone. Using the generic THP migration implementation allows us
remove the duplicate code and some hacks needed by the old ad hoc
implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both
THP and NUMA balancing. The most of them support THP migration except for
S390. Zi Yan tried to add THP migration support for S390 before but it
was not accepted due to the design of S390 PMD. For the discussion,
please see: https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip
huge PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't
figure out if they have been removed or not since I just found FOLL_NUMA
code in the current gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page. The test
VM has 80 vcpus and 64G memory. The test would create 2 processes to
consume 128G memory together which would incur memory pressure to cause
THP splits. And it also creates 80 processes to hog cpu, and the memory
consumer processes are bound to different nodes periodically in order to
increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page is
as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted the
raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split is
less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory pressure.
No obvious regression is spotted. The below is the test result *w/o*
memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
This patch (of 7):
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge
page fault could be removed, just like its PTE counterpart does.
Link: https://lkml.kernel.org/r/20210518200801.7413-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20210518200801.7413-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:51:35 +00:00
|
|
|
if (unlikely(is_swap_pmd(vmf.orig_pmd))) {
|
mm: thp: check pmd migration entry in common path
When THP migration is being used, memory management code needs to handle
pmd migration entries properly. This patch uses !pmd_present() or
is_swap_pmd() (depending on whether pmd_none() needs separate code or
not) to check pmd migration entries at the places where a pmd entry is
present.
Since pmd-related code uses split_huge_page(), split_huge_pmd(),
pmd_trans_huge(), pmd_trans_unstable(), or
pmd_none_or_trans_huge_or_clear_bad(), this patch:
1. adds pmd migration entry split code in split_huge_pmd(),
2. takes care of pmd migration entries whenever pmd_trans_huge() is present,
3. makes pmd_none_or_trans_huge_or_clear_bad() pmd migration entry aware.
Since split_huge_page() uses split_huge_pmd() and pmd_trans_unstable()
is equivalent to pmd_none_or_trans_huge_or_clear_bad(), we do not change
them.
Until this commit, a pmd entry should be:
1. pointing to a pte page,
2. is_swap_pmd(),
3. pmd_trans_huge(),
4. pmd_devmap(), or
5. pmd_none().
Signed-off-by: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Nellans <dnellans@nvidia.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:11:01 +00:00
|
|
|
VM_BUG_ON(thp_migration_supported() &&
|
mm: memory: add orig_pmd to struct vm_fault
Pach series "mm: thp: use generic THP migration for NUMA hinting fault", v3.
When the THP NUMA fault support was added THP migration was not supported
yet. So the ad hoc THP migration was implemented in NUMA fault handling.
Since v4.14 THP migration has been supported so it doesn't make too much
sense to still keep another THP migration implementation rather than using
the generic migration code. It is definitely a maintenance burden to keep
two THP migration implementation for different code paths and it is more
error prone. Using the generic THP migration implementation allows us
remove the duplicate code and some hacks needed by the old ad hoc
implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both
THP and NUMA balancing. The most of them support THP migration except for
S390. Zi Yan tried to add THP migration support for S390 before but it
was not accepted due to the design of S390 PMD. For the discussion,
please see: https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip
huge PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't
figure out if they have been removed or not since I just found FOLL_NUMA
code in the current gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page. The test
VM has 80 vcpus and 64G memory. The test would create 2 processes to
consume 128G memory together which would incur memory pressure to cause
THP splits. And it also creates 80 processes to hog cpu, and the memory
consumer processes are bound to different nodes periodically in order to
increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page is
as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted the
raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split is
less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory pressure.
No obvious regression is spotted. The below is the test result *w/o*
memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
This patch (of 7):
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge
page fault could be removed, just like its PTE counterpart does.
Link: https://lkml.kernel.org/r/20210518200801.7413-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20210518200801.7413-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:51:35 +00:00
|
|
|
!is_pmd_migration_entry(vmf.orig_pmd));
|
|
|
|
if (is_pmd_migration_entry(vmf.orig_pmd))
|
mm: thp: check pmd migration entry in common path
When THP migration is being used, memory management code needs to handle
pmd migration entries properly. This patch uses !pmd_present() or
is_swap_pmd() (depending on whether pmd_none() needs separate code or
not) to check pmd migration entries at the places where a pmd entry is
present.
Since pmd-related code uses split_huge_page(), split_huge_pmd(),
pmd_trans_huge(), pmd_trans_unstable(), or
pmd_none_or_trans_huge_or_clear_bad(), this patch:
1. adds pmd migration entry split code in split_huge_pmd(),
2. takes care of pmd migration entries whenever pmd_trans_huge() is present,
3. makes pmd_none_or_trans_huge_or_clear_bad() pmd migration entry aware.
Since split_huge_page() uses split_huge_pmd() and pmd_trans_unstable()
is equivalent to pmd_none_or_trans_huge_or_clear_bad(), we do not change
them.
Until this commit, a pmd entry should be:
1. pointing to a pte page,
2. is_swap_pmd(),
3. pmd_trans_huge(),
4. pmd_devmap(), or
5. pmd_none().
Signed-off-by: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Nellans <dnellans@nvidia.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:11:01 +00:00
|
|
|
pmd_migration_entry_wait(mm, vmf.pmd);
|
|
|
|
return 0;
|
|
|
|
}
|
mm: memory: add orig_pmd to struct vm_fault
Pach series "mm: thp: use generic THP migration for NUMA hinting fault", v3.
When the THP NUMA fault support was added THP migration was not supported
yet. So the ad hoc THP migration was implemented in NUMA fault handling.
Since v4.14 THP migration has been supported so it doesn't make too much
sense to still keep another THP migration implementation rather than using
the generic migration code. It is definitely a maintenance burden to keep
two THP migration implementation for different code paths and it is more
error prone. Using the generic THP migration implementation allows us
remove the duplicate code and some hacks needed by the old ad hoc
implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both
THP and NUMA balancing. The most of them support THP migration except for
S390. Zi Yan tried to add THP migration support for S390 before but it
was not accepted due to the design of S390 PMD. For the discussion,
please see: https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip
huge PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't
figure out if they have been removed or not since I just found FOLL_NUMA
code in the current gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page. The test
VM has 80 vcpus and 64G memory. The test would create 2 processes to
consume 128G memory together which would incur memory pressure to cause
THP splits. And it also creates 80 processes to hog cpu, and the memory
consumer processes are bound to different nodes periodically in order to
increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page is
as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted the
raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split is
less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory pressure.
No obvious regression is spotted. The below is the test result *w/o*
memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
This patch (of 7):
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge
page fault could be removed, just like its PTE counterpart does.
Link: https://lkml.kernel.org/r/20210518200801.7413-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20210518200801.7413-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:51:35 +00:00
|
|
|
if (pmd_trans_huge(vmf.orig_pmd) || pmd_devmap(vmf.orig_pmd)) {
|
|
|
|
if (pmd_protnone(vmf.orig_pmd) && vma_is_accessible(vma))
|
|
|
|
return do_huge_pmd_numa_page(&vmf);
|
2012-10-25 12:16:31 +00:00
|
|
|
|
mm: support GUP-triggered unsharing of anonymous pages
Whenever GUP currently ends up taking a R/O pin on an anonymous page that
might be shared -- mapped R/O and !PageAnonExclusive() -- any write fault
on the page table entry will end up replacing the mapped anonymous page
due to COW, resulting in the GUP pin no longer being consistent with the
page actually mapped into the page table.
The possible ways to deal with this situation are:
(1) Ignore and pin -- what we do right now.
(2) Fail to pin -- which would be rather surprising to callers and
could break user space.
(3) Trigger unsharing and pin the now exclusive page -- reliable R/O
pins.
We want to implement 3) because it provides the clearest semantics and
allows for checking in unpin_user_pages() and friends for possible BUGs:
when trying to unpin a page that's no longer exclusive, clearly something
went very wrong and might result in memory corruptions that might be hard
to debug. So we better have a nice way to spot such issues.
To implement 3), we need a way for GUP to trigger unsharing:
FAULT_FLAG_UNSHARE. FAULT_FLAG_UNSHARE is only applicable to R/O mapped
anonymous pages and resembles COW logic during a write fault. However, in
contrast to a write fault, GUP-triggered unsharing will, for example,
still maintain the write protection.
Let's implement FAULT_FLAG_UNSHARE by hooking into the existing write
fault handlers for all applicable anonymous page types: ordinary pages,
THP and hugetlb.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that has been
marked exclusive in the meantime by someone else, there is nothing to do.
* If FAULT_FLAG_UNSHARE finds a R/O-mapped anonymous page that's not
marked exclusive, it will try detecting if the process is the exclusive
owner. If exclusive, it can be set exclusive similar to reuse logic
during write faults via page_move_anon_rmap() and there is nothing
else to do; otherwise, we either have to copy and map a fresh,
anonymous exclusive page R/O (ordinary pages, hugetlb), or split the
THP.
This commit is heavily based on patches by Andrea.
Link: https://lkml.kernel.org/r/20220428083441.37290-16-david@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Co-developed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Don Dutile <ddutile@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Liang Zhang <zhangliang5@huawei.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-05-10 01:20:45 +00:00
|
|
|
if ((flags & (FAULT_FLAG_WRITE|FAULT_FLAG_UNSHARE)) &&
|
|
|
|
!pmd_write(vmf.orig_pmd)) {
|
mm: memory: add orig_pmd to struct vm_fault
Pach series "mm: thp: use generic THP migration for NUMA hinting fault", v3.
When the THP NUMA fault support was added THP migration was not supported
yet. So the ad hoc THP migration was implemented in NUMA fault handling.
Since v4.14 THP migration has been supported so it doesn't make too much
sense to still keep another THP migration implementation rather than using
the generic migration code. It is definitely a maintenance burden to keep
two THP migration implementation for different code paths and it is more
error prone. Using the generic THP migration implementation allows us
remove the duplicate code and some hacks needed by the old ad hoc
implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both
THP and NUMA balancing. The most of them support THP migration except for
S390. Zi Yan tried to add THP migration support for S390 before but it
was not accepted due to the design of S390 PMD. For the discussion,
please see: https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip
huge PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't
figure out if they have been removed or not since I just found FOLL_NUMA
code in the current gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page. The test
VM has 80 vcpus and 64G memory. The test would create 2 processes to
consume 128G memory together which would incur memory pressure to cause
THP splits. And it also creates 80 processes to hog cpu, and the memory
consumer processes are bound to different nodes periodically in order to
increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page is
as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted the
raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split is
less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory pressure.
No obvious regression is spotted. The below is the test result *w/o*
memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
This patch (of 7):
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge
page fault could be removed, just like its PTE counterpart does.
Link: https://lkml.kernel.org/r/20210518200801.7413-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20210518200801.7413-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:51:35 +00:00
|
|
|
ret = wp_huge_pmd(&vmf);
|
2014-02-25 23:01:42 +00:00
|
|
|
if (!(ret & VM_FAULT_FALLBACK))
|
|
|
|
return ret;
|
2012-12-12 00:01:27 +00:00
|
|
|
} else {
|
mm: memory: add orig_pmd to struct vm_fault
Pach series "mm: thp: use generic THP migration for NUMA hinting fault", v3.
When the THP NUMA fault support was added THP migration was not supported
yet. So the ad hoc THP migration was implemented in NUMA fault handling.
Since v4.14 THP migration has been supported so it doesn't make too much
sense to still keep another THP migration implementation rather than using
the generic migration code. It is definitely a maintenance burden to keep
two THP migration implementation for different code paths and it is more
error prone. Using the generic THP migration implementation allows us
remove the duplicate code and some hacks needed by the old ad hoc
implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both
THP and NUMA balancing. The most of them support THP migration except for
S390. Zi Yan tried to add THP migration support for S390 before but it
was not accepted due to the design of S390 PMD. For the discussion,
please see: https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip
huge PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't
figure out if they have been removed or not since I just found FOLL_NUMA
code in the current gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page. The test
VM has 80 vcpus and 64G memory. The test would create 2 processes to
consume 128G memory together which would incur memory pressure to cause
THP splits. And it also creates 80 processes to hog cpu, and the memory
consumer processes are bound to different nodes periodically in order to
increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page is
as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted the
raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split is
less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory pressure.
No obvious regression is spotted. The below is the test result *w/o*
memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
This patch (of 7):
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge
page fault could be removed, just like its PTE counterpart does.
Link: https://lkml.kernel.org/r/20210518200801.7413-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20210518200801.7413-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-01 01:51:35 +00:00
|
|
|
huge_pmd_set_accessed(&vmf);
|
2014-02-25 23:01:42 +00:00
|
|
|
return 0;
|
2012-05-29 22:06:23 +00:00
|
|
|
}
|
thp: transparent hugepage core
Lately I've been working to make KVM use hugepages transparently without
the usual restrictions of hugetlbfs. Some of the restrictions I'd like to
see removed:
1) hugepages have to be swappable or the guest physical memory remains
locked in RAM and can't be paged out to swap
2) if a hugepage allocation fails, regular pages should be allocated
instead and mixed in the same vma without any failure and without
userland noticing
3) if some task quits and more hugepages become available in the
buddy, guest physical memory backed by regular pages should be
relocated on hugepages automatically in regions under
madvise(MADV_HUGEPAGE) (ideally event driven by waking up the
kernel deamon if the order=HPAGE_PMD_SHIFT-PAGE_SHIFT list becomes
not null)
4) avoidance of reservation and maximization of use of hugepages whenever
possible. Reservation (needed to avoid runtime fatal faliures) may be ok for
1 machine with 1 database with 1 database cache with 1 database cache size
known at boot time. It's definitely not feasible with a virtualization
hypervisor usage like RHEV-H that runs an unknown number of virtual machines
with an unknown size of each virtual machine with an unknown amount of
pagecache that could be potentially useful in the host for guest not using
O_DIRECT (aka cache=off).
hugepages in the virtualization hypervisor (and also in the guest!) are
much more important than in a regular host not using virtualization,
becasue with NPT/EPT they decrease the tlb-miss cacheline accesses from 24
to 19 in case only the hypervisor uses transparent hugepages, and they
decrease the tlb-miss cacheline accesses from 19 to 15 in case both the
linux hypervisor and the linux guest both uses this patch (though the
guest will limit the addition speedup to anonymous regions only for
now...). Even more important is that the tlb miss handler is much slower
on a NPT/EPT guest than for a regular shadow paging or no-virtualization
scenario. So maximizing the amount of virtual memory cached by the TLB
pays off significantly more with NPT/EPT than without (even if there would
be no significant speedup in the tlb-miss runtime).
The first (and more tedious) part of this work requires allowing the VM to
handle anonymous hugepages mixed with regular pages transparently on
regular anonymous vmas. This is what this patch tries to achieve in the
least intrusive possible way. We want hugepages and hugetlb to be used in
a way so that all applications can benefit without changes (as usual we
leverage the KVM virtualization design: by improving the Linux VM at
large, KVM gets the performance boost too).
The most important design choice is: always fallback to 4k allocation if
the hugepage allocation fails! This is the _very_ opposite of some large
pagecache patches that failed with -EIO back then if a 64k (or similar)
allocation failed...
Second important decision (to reduce the impact of the feature on the
existing pagetable handling code) is that at any time we can split an
hugepage into 512 regular pages and it has to be done with an operation
that can't fail. This way the reliability of the swapping isn't decreased
(no need to allocate memory when we are short on memory to swap) and it's
trivial to plug a split_huge_page* one-liner where needed without
polluting the VM. Over time we can teach mprotect, mremap and friends to
handle pmd_trans_huge natively without calling split_huge_page*. The fact
it can't fail isn't just for swap: if split_huge_page would return -ENOMEM
(instead of the current void) we'd need to rollback the mprotect from the
middle of it (ideally including undoing the split_vma) which would be a
big change and in the very wrong direction (it'd likely be simpler not to
call split_huge_page at all and to teach mprotect and friends to handle
hugepages instead of rolling them back from the middle). In short the
very value of split_huge_page is that it can't fail.
The collapsing and madvise(MADV_HUGEPAGE) part will remain separated and
incremental and it'll just be an "harmless" addition later if this initial
part is agreed upon. It also should be noted that locking-wise replacing
regular pages with hugepages is going to be very easy if compared to what
I'm doing below in split_huge_page, as it will only happen when
page_count(page) matches page_mapcount(page) if we can take the PG_lock
and mmap_sem in write mode. collapse_huge_page will be a "best effort"
that (unlike split_huge_page) can fail at the minimal sign of trouble and
we can try again later. collapse_huge_page will be similar to how KSM
works and the madvise(MADV_HUGEPAGE) will work similar to
madvise(MADV_MERGEABLE).
The default I like is that transparent hugepages are used at page fault
time. This can be changed with
/sys/kernel/mm/transparent_hugepage/enabled. The control knob can be set
to three values "always", "madvise", "never" which mean respectively that
hugepages are always used, or only inside madvise(MADV_HUGEPAGE) regions,
or never used. /sys/kernel/mm/transparent_hugepage/defrag instead
controls if the hugepage allocation should defrag memory aggressively
"always", only inside "madvise" regions, or "never".
The pmd_trans_splitting/pmd_trans_huge locking is very solid. The
put_page (from get_user_page users that can't use mmu notifier like
O_DIRECT) that runs against a __split_huge_page_refcount instead was a
pain to serialize in a way that would result always in a coherent page
count for both tail and head. I think my locking solution with a
compound_lock taken only after the page_first is valid and is still a
PageHead should be safe but it surely needs review from SMP race point of
view. In short there is no current existing way to serialize the O_DIRECT
final put_page against split_huge_page_refcount so I had to invent a new
one (O_DIRECT loses knowledge on the mapping status by the time gup_fast
returns so...). And I didn't want to impact all gup/gup_fast users for
now, maybe if we change the gup interface substantially we can avoid this
locking, I admit I didn't think too much about it because changing the gup
unpinning interface would be invasive.
If we ignored O_DIRECT we could stick to the existing compound refcounting
code, by simply adding a get_user_pages_fast_flags(foll_flags) where KVM
(and any other mmu notifier user) would call it without FOLL_GET (and if
FOLL_GET isn't set we'd just BUG_ON if nobody registered itself in the
current task mmu notifier list yet). But O_DIRECT is fundamental for
decent performance of virtualized I/O on fast storage so we can't avoid it
to solve the race of put_page against split_huge_page_refcount to achieve
a complete hugepage feature for KVM.
Swap and oom works fine (well just like with regular pages ;). MMU
notifier is handled transparently too, with the exception of the young bit
on the pmd, that didn't have a range check but I think KVM will be fine
because the whole point of hugepages is that EPT/NPT will also use a huge
pmd when they notice gup returns pages with PageCompound set, so they
won't care of a range and there's just the pmd young bit to check in that
case.
NOTE: in some cases if the L2 cache is small, this may slowdown and waste
memory during COWs because 4M of memory are accessed in a single fault
instead of 8k (the payoff is that after COW the program can run faster).
So we might want to switch the copy_huge_page (and clear_huge_page too) to
not temporal stores. I also extensively researched ways to avoid this
cache trashing with a full prefault logic that would cow in 8k/16k/32k/64k
up to 1M (I can send those patches that fully implemented prefault) but I
concluded they're not worth it and they add an huge additional complexity
and they remove all tlb benefits until the full hugepage has been faulted
in, to save a little bit of memory and some cache during app startup, but
they still don't improve substantially the cache-trashing during startup
if the prefault happens in >4k chunks. One reason is that those 4k pte
entries copied are still mapped on a perfectly cache-colored hugepage, so
the trashing is the worst one can generate in those copies (cow of 4k page
copies aren't so well colored so they trashes less, but again this results
in software running faster after the page fault). Those prefault patches
allowed things like a pte where post-cow pages were local 4k regular anon
pages and the not-yet-cowed pte entries were pointing in the middle of
some hugepage mapped read-only. If it doesn't payoff substantially with
todays hardware it will payoff even less in the future with larger l2
caches, and the prefault logic would blot the VM a lot. If one is
emebdded transparent_hugepage can be disabled during boot with sysfs or
with the boot commandline parameter transparent_hugepage=0 (or
transparent_hugepage=2 to restrict hugepages inside madvise regions) that
will ensure not a single hugepage is allocated at boot time. It is simple
enough to just disable transparent hugepage globally and let transparent
hugepages be allocated selectively by applications in the MADV_HUGEPAGE
region (both at page fault time, and if enabled with the
collapse_huge_page too through the kernel daemon).
This patch supports only hugepages mapped in the pmd, archs that have
smaller hugepages will not fit in this patch alone. Also some archs like
power have certain tlb limits that prevents mixing different page size in
the same regions so they will not fit in this framework that requires
"graceful fallback" to basic PAGE_SIZE in case of physical memory
fragmentation. hugetlbfs remains a perfect fit for those because its
software limits happen to match the hardware limits. hugetlbfs also
remains a perfect fit for hugepage sizes like 1GByte that cannot be hoped
to be found not fragmented after a certain system uptime and that would be
very expensive to defragment with relocation, so requiring reservation.
hugetlbfs is the "reservation way", the point of transparent hugepages is
not to have any reservation at all and maximizing the use of cache and
hugepages at all times automatically.
Some performance result:
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largep
ages3
memset page fault 1566023
memset tlb miss 453854
memset second tlb miss 453321
random access tlb miss 41635
random access second tlb miss 41658
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largepages3
memset page fault 1566471
memset tlb miss 453375
memset second tlb miss 453320
random access tlb miss 41636
random access second tlb miss 41637
vmx andrea # ./largepages3
memset page fault 1566642
memset tlb miss 453417
memset second tlb miss 453313
random access tlb miss 41630
random access second tlb miss 41647
vmx andrea # ./largepages3
memset page fault 1566872
memset tlb miss 453418
memset second tlb miss 453315
random access tlb miss 41618
random access second tlb miss 41659
vmx andrea # echo 0 > /proc/sys/vm/transparent_hugepage
vmx andrea # ./largepages3
memset page fault 2182476
memset tlb miss 460305
memset second tlb miss 460179
random access tlb miss 44483
random access second tlb miss 44186
vmx andrea # ./largepages3
memset page fault 2182791
memset tlb miss 460742
memset second tlb miss 459962
random access tlb miss 43981
random access second tlb miss 43988
============
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#define SIZE (3UL*1024*1024*1024)
int main()
{
char *p = malloc(SIZE), *p2;
struct timeval before, after;
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset page fault %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
return 0;
}
============
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-01-13 23:46:52 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-12-14 23:06:58 +00:00
|
|
|
return handle_pte_fault(&vmf);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
/**
|
2021-05-07 01:06:47 +00:00
|
|
|
* mm_account_fault - Do page fault accounting
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
*
|
|
|
|
* @regs: the pt_regs struct pointer. When set to NULL, will skip accounting
|
|
|
|
* of perf event counters, but we'll still do the per-task accounting to
|
|
|
|
* the task who triggered this page fault.
|
|
|
|
* @address: the faulted address.
|
|
|
|
* @flags: the fault flags.
|
|
|
|
* @ret: the fault retcode.
|
|
|
|
*
|
2021-05-07 01:06:47 +00:00
|
|
|
* This will take care of most of the page fault accounting. Meanwhile, it
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
* will also include the PERF_COUNT_SW_PAGE_FAULTS_[MAJ|MIN] perf counter
|
2021-05-07 01:06:47 +00:00
|
|
|
* updates. However, note that the handling of PERF_COUNT_SW_PAGE_FAULTS should
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
* still be in per-arch page fault handlers at the entry of page fault.
|
|
|
|
*/
|
2023-04-19 17:58:36 +00:00
|
|
|
static inline void mm_account_fault(struct mm_struct *mm, struct pt_regs *regs,
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
unsigned long address, unsigned int flags,
|
|
|
|
vm_fault_t ret)
|
|
|
|
{
|
|
|
|
bool major;
|
|
|
|
|
2023-04-19 17:58:36 +00:00
|
|
|
/* Incomplete faults will be accounted upon completion. */
|
|
|
|
if (ret & VM_FAULT_RETRY)
|
|
|
|
return;
|
|
|
|
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
/*
|
2023-04-19 17:58:36 +00:00
|
|
|
* To preserve the behavior of older kernels, PGFAULT counters record
|
|
|
|
* both successful and failed faults, as opposed to perf counters,
|
|
|
|
* which ignore failed cases.
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
*/
|
2023-04-19 17:58:36 +00:00
|
|
|
count_vm_event(PGFAULT);
|
|
|
|
count_memcg_event_mm(mm, PGFAULT);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do not account for unsuccessful faults (e.g. when the address wasn't
|
|
|
|
* valid). That includes arch_vma_access_permitted() failing before
|
|
|
|
* reaching here. So this is not a "this many hardware page faults"
|
|
|
|
* counter. We should use the hw profiling for that.
|
|
|
|
*/
|
|
|
|
if (ret & VM_FAULT_ERROR)
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We define the fault as a major fault when the final successful fault
|
|
|
|
* is VM_FAULT_MAJOR, or if it retried (which implies that we couldn't
|
|
|
|
* handle it immediately previously).
|
|
|
|
*/
|
|
|
|
major = (ret & VM_FAULT_MAJOR) || (flags & FAULT_FLAG_TRIED);
|
|
|
|
|
2020-08-12 01:38:57 +00:00
|
|
|
if (major)
|
|
|
|
current->maj_flt++;
|
|
|
|
else
|
|
|
|
current->min_flt++;
|
|
|
|
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
/*
|
2020-08-12 01:38:57 +00:00
|
|
|
* If the fault is done for GUP, regs will be NULL. We only do the
|
|
|
|
* accounting for the per thread fault counters who triggered the
|
|
|
|
* fault, and we skip the perf event updates.
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
*/
|
|
|
|
if (!regs)
|
|
|
|
return;
|
|
|
|
|
2020-08-12 01:38:57 +00:00
|
|
|
if (major)
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address);
|
2020-08-12 01:38:57 +00:00
|
|
|
else
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address);
|
|
|
|
}
|
|
|
|
|
mm: multi-gen LRU: groundwork
Evictable pages are divided into multiple generations for each lruvec.
The youngest generation number is stored in lrugen->max_seq for both
anon and file types as they are aged on an equal footing. The oldest
generation numbers are stored in lrugen->min_seq[] separately for anon
and file types as clean file pages can be evicted regardless of swap
constraints. These three variables are monotonically increasing.
Generation numbers are truncated into order_base_2(MAX_NR_GENS+1) bits
in order to fit into the gen counter in folio->flags. Each truncated
generation number is an index to lrugen->lists[]. The sliding window
technique is used to track at least MIN_NR_GENS and at most
MAX_NR_GENS generations. The gen counter stores a value within [1,
MAX_NR_GENS] while a page is on one of lrugen->lists[]. Otherwise it
stores 0.
There are two conceptually independent procedures: "the aging", which
produces young generations, and "the eviction", which consumes old
generations. They form a closed-loop system, i.e., "the page reclaim".
Both procedures can be invoked from userspace for the purposes of working
set estimation and proactive reclaim. These techniques are commonly used
to optimize job scheduling (bin packing) in data centers [1][2].
To avoid confusion, the terms "hot" and "cold" will be applied to the
multi-gen LRU, as a new convention; the terms "active" and "inactive" will
be applied to the active/inactive LRU, as usual.
The protection of hot pages and the selection of cold pages are based
on page access channels and patterns. There are two access channels:
one through page tables and the other through file descriptors. The
protection of the former channel is by design stronger because:
1. The uncertainty in determining the access patterns of the former
channel is higher due to the approximation of the accessed bit.
2. The cost of evicting the former channel is higher due to the TLB
flushes required and the likelihood of encountering the dirty bit.
3. The penalty of underprotecting the former channel is higher because
applications usually do not prepare themselves for major page
faults like they do for blocked I/O. E.g., GUI applications
commonly use dedicated I/O threads to avoid blocking rendering
threads.
There are also two access patterns: one with temporal locality and the
other without. For the reasons listed above, the former channel is
assumed to follow the former pattern unless VM_SEQ_READ or VM_RAND_READ is
present; the latter channel is assumed to follow the latter pattern unless
outlying refaults have been observed [3][4].
The next patch will address the "outlying refaults". Three macros, i.e.,
LRU_REFS_WIDTH, LRU_REFS_PGOFF and LRU_REFS_MASK, used later are added in
this patch to make the entire patchset less diffy.
A page is added to the youngest generation on faulting. The aging needs
to check the accessed bit at least twice before handing this page over to
the eviction. The first check takes care of the accessed bit set on the
initial fault; the second check makes sure this page has not been used
since then. This protocol, AKA second chance, requires a minimum of two
generations, hence MIN_NR_GENS.
[1] https://dl.acm.org/doi/10.1145/3297858.3304053
[2] https://dl.acm.org/doi/10.1145/3503222.3507731
[3] https://lwn.net/Articles/495543/
[4] https://lwn.net/Articles/815342/
Link: https://lkml.kernel.org/r/20220918080010.2920238-6-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-18 08:00:02 +00:00
|
|
|
#ifdef CONFIG_LRU_GEN
|
|
|
|
static void lru_gen_enter_fault(struct vm_area_struct *vma)
|
|
|
|
{
|
2022-12-30 21:52:51 +00:00
|
|
|
/* the LRU algorithm only applies to accesses with recency */
|
|
|
|
current->in_lru_fault = vma_has_recency(vma);
|
mm: multi-gen LRU: groundwork
Evictable pages are divided into multiple generations for each lruvec.
The youngest generation number is stored in lrugen->max_seq for both
anon and file types as they are aged on an equal footing. The oldest
generation numbers are stored in lrugen->min_seq[] separately for anon
and file types as clean file pages can be evicted regardless of swap
constraints. These three variables are monotonically increasing.
Generation numbers are truncated into order_base_2(MAX_NR_GENS+1) bits
in order to fit into the gen counter in folio->flags. Each truncated
generation number is an index to lrugen->lists[]. The sliding window
technique is used to track at least MIN_NR_GENS and at most
MAX_NR_GENS generations. The gen counter stores a value within [1,
MAX_NR_GENS] while a page is on one of lrugen->lists[]. Otherwise it
stores 0.
There are two conceptually independent procedures: "the aging", which
produces young generations, and "the eviction", which consumes old
generations. They form a closed-loop system, i.e., "the page reclaim".
Both procedures can be invoked from userspace for the purposes of working
set estimation and proactive reclaim. These techniques are commonly used
to optimize job scheduling (bin packing) in data centers [1][2].
To avoid confusion, the terms "hot" and "cold" will be applied to the
multi-gen LRU, as a new convention; the terms "active" and "inactive" will
be applied to the active/inactive LRU, as usual.
The protection of hot pages and the selection of cold pages are based
on page access channels and patterns. There are two access channels:
one through page tables and the other through file descriptors. The
protection of the former channel is by design stronger because:
1. The uncertainty in determining the access patterns of the former
channel is higher due to the approximation of the accessed bit.
2. The cost of evicting the former channel is higher due to the TLB
flushes required and the likelihood of encountering the dirty bit.
3. The penalty of underprotecting the former channel is higher because
applications usually do not prepare themselves for major page
faults like they do for blocked I/O. E.g., GUI applications
commonly use dedicated I/O threads to avoid blocking rendering
threads.
There are also two access patterns: one with temporal locality and the
other without. For the reasons listed above, the former channel is
assumed to follow the former pattern unless VM_SEQ_READ or VM_RAND_READ is
present; the latter channel is assumed to follow the latter pattern unless
outlying refaults have been observed [3][4].
The next patch will address the "outlying refaults". Three macros, i.e.,
LRU_REFS_WIDTH, LRU_REFS_PGOFF and LRU_REFS_MASK, used later are added in
this patch to make the entire patchset less diffy.
A page is added to the youngest generation on faulting. The aging needs
to check the accessed bit at least twice before handing this page over to
the eviction. The first check takes care of the accessed bit set on the
initial fault; the second check makes sure this page has not been used
since then. This protocol, AKA second chance, requires a minimum of two
generations, hence MIN_NR_GENS.
[1] https://dl.acm.org/doi/10.1145/3297858.3304053
[2] https://dl.acm.org/doi/10.1145/3503222.3507731
[3] https://lwn.net/Articles/495543/
[4] https://lwn.net/Articles/815342/
Link: https://lkml.kernel.org/r/20220918080010.2920238-6-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-18 08:00:02 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void lru_gen_exit_fault(void)
|
|
|
|
{
|
|
|
|
current->in_lru_fault = false;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static void lru_gen_enter_fault(struct vm_area_struct *vma)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
static void lru_gen_exit_fault(void)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_LRU_GEN */
|
|
|
|
|
2022-11-16 10:26:43 +00:00
|
|
|
static vm_fault_t sanitize_fault_flags(struct vm_area_struct *vma,
|
|
|
|
unsigned int *flags)
|
|
|
|
{
|
|
|
|
if (unlikely(*flags & FAULT_FLAG_UNSHARE)) {
|
|
|
|
if (WARN_ON_ONCE(*flags & FAULT_FLAG_WRITE))
|
|
|
|
return VM_FAULT_SIGSEGV;
|
|
|
|
/*
|
|
|
|
* FAULT_FLAG_UNSHARE only applies to COW mappings. Let's
|
|
|
|
* just treat it like an ordinary read-fault otherwise.
|
|
|
|
*/
|
|
|
|
if (!is_cow_mapping(vma->vm_flags))
|
|
|
|
*flags &= ~FAULT_FLAG_UNSHARE;
|
2022-11-16 10:26:44 +00:00
|
|
|
} else if (*flags & FAULT_FLAG_WRITE) {
|
|
|
|
/* Write faults on read-only mappings are impossible ... */
|
|
|
|
if (WARN_ON_ONCE(!(vma->vm_flags & VM_MAYWRITE)))
|
|
|
|
return VM_FAULT_SIGSEGV;
|
|
|
|
/* ... and FOLL_FORCE only applies to COW mappings. */
|
|
|
|
if (WARN_ON_ONCE(!(vma->vm_flags & VM_WRITE) &&
|
|
|
|
!is_cow_mapping(vma->vm_flags)))
|
|
|
|
return VM_FAULT_SIGSEGV;
|
2022-11-16 10:26:43 +00:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-08-06 23:07:24 +00:00
|
|
|
/*
|
|
|
|
* By the time we get here, we already hold the mm semaphore
|
|
|
|
*
|
2020-06-09 04:33:54 +00:00
|
|
|
* The mmap_lock may have been released depending on flags and our
|
2021-03-19 01:39:45 +00:00
|
|
|
* return value. See filemap_fault() and __folio_lock_or_retry().
|
2014-08-06 23:07:24 +00:00
|
|
|
*/
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
unsigned int flags, struct pt_regs *regs)
|
2013-09-12 22:13:42 +00:00
|
|
|
{
|
2023-04-19 17:58:36 +00:00
|
|
|
/* If the fault handler drops the mmap_lock, vma may be freed */
|
|
|
|
struct mm_struct *mm = vma->vm_mm;
|
2018-08-24 00:01:36 +00:00
|
|
|
vm_fault_t ret;
|
2013-09-12 22:13:42 +00:00
|
|
|
|
|
|
|
__set_current_state(TASK_RUNNING);
|
|
|
|
|
2022-11-16 10:26:43 +00:00
|
|
|
ret = sanitize_fault_flags(vma, &flags);
|
|
|
|
if (ret)
|
2023-04-19 17:58:36 +00:00
|
|
|
goto out;
|
2022-11-16 10:26:43 +00:00
|
|
|
|
2017-09-08 23:13:12 +00:00
|
|
|
if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
|
|
|
|
flags & FAULT_FLAG_INSTRUCTION,
|
2023-04-19 17:58:36 +00:00
|
|
|
flags & FAULT_FLAG_REMOTE)) {
|
|
|
|
ret = VM_FAULT_SIGSEGV;
|
|
|
|
goto out;
|
|
|
|
}
|
2017-09-08 23:13:12 +00:00
|
|
|
|
2013-09-12 22:13:42 +00:00
|
|
|
/*
|
|
|
|
* Enable the memcg OOM handling for faults triggered in user
|
|
|
|
* space. Kernel faults are handled more gracefully.
|
|
|
|
*/
|
|
|
|
if (flags & FAULT_FLAG_USER)
|
2018-08-17 22:47:11 +00:00
|
|
|
mem_cgroup_enter_user_fault();
|
2013-09-12 22:13:42 +00:00
|
|
|
|
mm: multi-gen LRU: groundwork
Evictable pages are divided into multiple generations for each lruvec.
The youngest generation number is stored in lrugen->max_seq for both
anon and file types as they are aged on an equal footing. The oldest
generation numbers are stored in lrugen->min_seq[] separately for anon
and file types as clean file pages can be evicted regardless of swap
constraints. These three variables are monotonically increasing.
Generation numbers are truncated into order_base_2(MAX_NR_GENS+1) bits
in order to fit into the gen counter in folio->flags. Each truncated
generation number is an index to lrugen->lists[]. The sliding window
technique is used to track at least MIN_NR_GENS and at most
MAX_NR_GENS generations. The gen counter stores a value within [1,
MAX_NR_GENS] while a page is on one of lrugen->lists[]. Otherwise it
stores 0.
There are two conceptually independent procedures: "the aging", which
produces young generations, and "the eviction", which consumes old
generations. They form a closed-loop system, i.e., "the page reclaim".
Both procedures can be invoked from userspace for the purposes of working
set estimation and proactive reclaim. These techniques are commonly used
to optimize job scheduling (bin packing) in data centers [1][2].
To avoid confusion, the terms "hot" and "cold" will be applied to the
multi-gen LRU, as a new convention; the terms "active" and "inactive" will
be applied to the active/inactive LRU, as usual.
The protection of hot pages and the selection of cold pages are based
on page access channels and patterns. There are two access channels:
one through page tables and the other through file descriptors. The
protection of the former channel is by design stronger because:
1. The uncertainty in determining the access patterns of the former
channel is higher due to the approximation of the accessed bit.
2. The cost of evicting the former channel is higher due to the TLB
flushes required and the likelihood of encountering the dirty bit.
3. The penalty of underprotecting the former channel is higher because
applications usually do not prepare themselves for major page
faults like they do for blocked I/O. E.g., GUI applications
commonly use dedicated I/O threads to avoid blocking rendering
threads.
There are also two access patterns: one with temporal locality and the
other without. For the reasons listed above, the former channel is
assumed to follow the former pattern unless VM_SEQ_READ or VM_RAND_READ is
present; the latter channel is assumed to follow the latter pattern unless
outlying refaults have been observed [3][4].
The next patch will address the "outlying refaults". Three macros, i.e.,
LRU_REFS_WIDTH, LRU_REFS_PGOFF and LRU_REFS_MASK, used later are added in
this patch to make the entire patchset less diffy.
A page is added to the youngest generation on faulting. The aging needs
to check the accessed bit at least twice before handing this page over to
the eviction. The first check takes care of the accessed bit set on the
initial fault; the second check makes sure this page has not been used
since then. This protocol, AKA second chance, requires a minimum of two
generations, hence MIN_NR_GENS.
[1] https://dl.acm.org/doi/10.1145/3297858.3304053
[2] https://dl.acm.org/doi/10.1145/3503222.3507731
[3] https://lwn.net/Articles/495543/
[4] https://lwn.net/Articles/815342/
Link: https://lkml.kernel.org/r/20220918080010.2920238-6-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-18 08:00:02 +00:00
|
|
|
lru_gen_enter_fault(vma);
|
|
|
|
|
2016-07-26 22:25:20 +00:00
|
|
|
if (unlikely(is_vm_hugetlb_page(vma)))
|
|
|
|
ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
|
|
|
|
else
|
|
|
|
ret = __handle_mm_fault(vma, address, flags);
|
2013-09-12 22:13:42 +00:00
|
|
|
|
mm: multi-gen LRU: groundwork
Evictable pages are divided into multiple generations for each lruvec.
The youngest generation number is stored in lrugen->max_seq for both
anon and file types as they are aged on an equal footing. The oldest
generation numbers are stored in lrugen->min_seq[] separately for anon
and file types as clean file pages can be evicted regardless of swap
constraints. These three variables are monotonically increasing.
Generation numbers are truncated into order_base_2(MAX_NR_GENS+1) bits
in order to fit into the gen counter in folio->flags. Each truncated
generation number is an index to lrugen->lists[]. The sliding window
technique is used to track at least MIN_NR_GENS and at most
MAX_NR_GENS generations. The gen counter stores a value within [1,
MAX_NR_GENS] while a page is on one of lrugen->lists[]. Otherwise it
stores 0.
There are two conceptually independent procedures: "the aging", which
produces young generations, and "the eviction", which consumes old
generations. They form a closed-loop system, i.e., "the page reclaim".
Both procedures can be invoked from userspace for the purposes of working
set estimation and proactive reclaim. These techniques are commonly used
to optimize job scheduling (bin packing) in data centers [1][2].
To avoid confusion, the terms "hot" and "cold" will be applied to the
multi-gen LRU, as a new convention; the terms "active" and "inactive" will
be applied to the active/inactive LRU, as usual.
The protection of hot pages and the selection of cold pages are based
on page access channels and patterns. There are two access channels:
one through page tables and the other through file descriptors. The
protection of the former channel is by design stronger because:
1. The uncertainty in determining the access patterns of the former
channel is higher due to the approximation of the accessed bit.
2. The cost of evicting the former channel is higher due to the TLB
flushes required and the likelihood of encountering the dirty bit.
3. The penalty of underprotecting the former channel is higher because
applications usually do not prepare themselves for major page
faults like they do for blocked I/O. E.g., GUI applications
commonly use dedicated I/O threads to avoid blocking rendering
threads.
There are also two access patterns: one with temporal locality and the
other without. For the reasons listed above, the former channel is
assumed to follow the former pattern unless VM_SEQ_READ or VM_RAND_READ is
present; the latter channel is assumed to follow the latter pattern unless
outlying refaults have been observed [3][4].
The next patch will address the "outlying refaults". Three macros, i.e.,
LRU_REFS_WIDTH, LRU_REFS_PGOFF and LRU_REFS_MASK, used later are added in
this patch to make the entire patchset less diffy.
A page is added to the youngest generation on faulting. The aging needs
to check the accessed bit at least twice before handing this page over to
the eviction. The first check takes care of the accessed bit set on the
initial fault; the second check makes sure this page has not been used
since then. This protocol, AKA second chance, requires a minimum of two
generations, hence MIN_NR_GENS.
[1] https://dl.acm.org/doi/10.1145/3297858.3304053
[2] https://dl.acm.org/doi/10.1145/3503222.3507731
[3] https://lwn.net/Articles/495543/
[4] https://lwn.net/Articles/815342/
Link: https://lkml.kernel.org/r/20220918080010.2920238-6-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-09-18 08:00:02 +00:00
|
|
|
lru_gen_exit_fault();
|
|
|
|
|
2013-10-16 20:46:59 +00:00
|
|
|
if (flags & FAULT_FLAG_USER) {
|
2018-08-17 22:47:11 +00:00
|
|
|
mem_cgroup_exit_user_fault();
|
2017-02-24 22:59:01 +00:00
|
|
|
/*
|
|
|
|
* The task may have entered a memcg OOM situation but
|
|
|
|
* if the allocation error was handled gracefully (no
|
|
|
|
* VM_FAULT_OOM), there is no need to kill anything.
|
|
|
|
* Just clean up the OOM state peacefully.
|
|
|
|
*/
|
|
|
|
if (task_in_memcg_oom(current) && !(ret & VM_FAULT_OOM))
|
|
|
|
mem_cgroup_oom_synchronize(false);
|
2013-10-16 20:46:59 +00:00
|
|
|
}
|
2023-04-19 17:58:36 +00:00
|
|
|
out:
|
|
|
|
mm_account_fault(mm, regs, address, flags, ret);
|
mm: do page fault accounting in handle_mm_fault
Patch series "mm: Page fault accounting cleanups", v5.
This is v5 of the pf accounting cleanup series. It originates from Gerald
Schaefer's report on an issue a week ago regarding to incorrect page fault
accountings for retried page fault after commit 4064b9827063 ("mm: allow
VM_FAULT_RETRY for multiple times"):
https://lore.kernel.org/lkml/20200610174811.44b94525@thinkpad/
What this series did:
- Correct page fault accounting: we do accounting for a page fault
(no matter whether it's from #PF handling, or gup, or anything else)
only with the one that completed the fault. For example, page fault
retries should not be counted in page fault counters. Same to the
perf events.
- Unify definition of PERF_COUNT_SW_PAGE_FAULTS: currently this perf
event is used in an adhoc way across different archs.
Case (1): for many archs it's done at the entry of a page fault
handler, so that it will also cover e.g. errornous faults.
Case (2): for some other archs, it is only accounted when the page
fault is resolved successfully.
Case (3): there're still quite some archs that have not enabled
this perf event.
Since this series will touch merely all the archs, we unify this
perf event to always follow case (1), which is the one that makes most
sense. And since we moved the accounting into handle_mm_fault, the
other two MAJ/MIN perf events are well taken care of naturally.
- Unify definition of "major faults": the definition of "major
fault" is slightly changed when used in accounting (not
VM_FAULT_MAJOR). More information in patch 1.
- Always account the page fault onto the one that triggered the page
fault. This does not matter much for #PF handlings, but mostly for
gup. More information on this in patch 25.
Patchset layout:
Patch 1: Introduced the accounting in handle_mm_fault(), not enabled.
Patch 2-23: Enable the new accounting for arch #PF handlers one by one.
Patch 24: Enable the new accounting for the rest outliers (gup, iommu, etc.)
Patch 25: Cleanup GUP task_struct pointer since it's not needed any more
This patch (of 25):
This is a preparation patch to move page fault accountings into the
general code in handle_mm_fault(). This includes both the per task
flt_maj/flt_min counters, and the major/minor page fault perf events. To
do this, the pt_regs pointer is passed into handle_mm_fault().
PERF_COUNT_SW_PAGE_FAULTS should still be kept in per-arch page fault
handlers.
So far, all the pt_regs pointer that passed into handle_mm_fault() is
NULL, which means this patch should have no intented functional change.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Cain <bcain@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Greentime Hu <green.hu@gmail.com>
Cc: Guo Ren <guoren@kernel.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Helge Deller <deller@gmx.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James E.J. Bottomley <James.Bottomley@HansenPartnership.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Ley Foon Tan <ley.foon.tan@intel.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Rich Felker <dalias@libc.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Stafford Horne <shorne@gmail.com>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Link: http://lkml.kernel.org/r/20200707225021.200906-1-peterx@redhat.com
Link: http://lkml.kernel.org/r/20200707225021.200906-2-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-08-12 01:37:44 +00:00
|
|
|
|
2013-09-12 22:13:42 +00:00
|
|
|
return ret;
|
|
|
|
}
|
2014-12-13 00:55:27 +00:00
|
|
|
EXPORT_SYMBOL_GPL(handle_mm_fault);
|
2013-09-12 22:13:42 +00:00
|
|
|
|
mm: introduce new 'lock_mm_and_find_vma()' page fault helper
commit c2508ec5a58db67093f4fb8bf89a9a7c53a109e9 upstream.
.. and make x86 use it.
This basically extracts the existing x86 "find and expand faulting vma"
code, but extends it to also take the mmap lock for writing in case we
actually do need to expand the vma.
We've historically short-circuited that case, and have some rather ugly
special logic to serialize the stack segment expansion (since we only
hold the mmap lock for reading) that doesn't match the normal VM
locking.
That slight violation of locking worked well, right up until it didn't:
the maple tree code really does want proper locking even for simple
extension of an existing vma.
So extract the code for "look up the vma of the fault" from x86, fix it
up to do the necessary write locking, and make it available as a helper
function for other architectures that can use the common helper.
Note: I say "common helper", but it really only handles the normal
stack-grows-down case. Which is all architectures except for PA-RISC
and IA64. So some rare architectures can't use the helper, but if they
care they'll just need to open-code this logic.
It's also worth pointing out that this code really would like to have an
optimistic "mmap_upgrade_trylock()" to make it quicker to go from a
read-lock (for the common case) to taking the write lock (for having to
extend the vma) in the normal single-threaded situation where there is
no other locking activity.
But that _is_ all the very uncommon special case, so while it would be
nice to have such an operation, it probably doesn't matter in reality.
I did put in the skeleton code for such a possible future expansion,
even if it only acts as pseudo-documentation for what we're doing.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2023-06-15 22:17:36 +00:00
|
|
|
#ifdef CONFIG_LOCK_MM_AND_FIND_VMA
|
|
|
|
#include <linux/extable.h>
|
|
|
|
|
|
|
|
static inline bool get_mmap_lock_carefully(struct mm_struct *mm, struct pt_regs *regs)
|
|
|
|
{
|
|
|
|
/* Even if this succeeds, make it clear we *might* have slept */
|
|
|
|
if (likely(mmap_read_trylock(mm))) {
|
|
|
|
might_sleep();
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (regs && !user_mode(regs)) {
|
|
|
|
unsigned long ip = instruction_pointer(regs);
|
|
|
|
if (!search_exception_tables(ip))
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2023-06-15 23:17:48 +00:00
|
|
|
return !mmap_read_lock_killable(mm);
|
mm: introduce new 'lock_mm_and_find_vma()' page fault helper
commit c2508ec5a58db67093f4fb8bf89a9a7c53a109e9 upstream.
.. and make x86 use it.
This basically extracts the existing x86 "find and expand faulting vma"
code, but extends it to also take the mmap lock for writing in case we
actually do need to expand the vma.
We've historically short-circuited that case, and have some rather ugly
special logic to serialize the stack segment expansion (since we only
hold the mmap lock for reading) that doesn't match the normal VM
locking.
That slight violation of locking worked well, right up until it didn't:
the maple tree code really does want proper locking even for simple
extension of an existing vma.
So extract the code for "look up the vma of the fault" from x86, fix it
up to do the necessary write locking, and make it available as a helper
function for other architectures that can use the common helper.
Note: I say "common helper", but it really only handles the normal
stack-grows-down case. Which is all architectures except for PA-RISC
and IA64. So some rare architectures can't use the helper, but if they
care they'll just need to open-code this logic.
It's also worth pointing out that this code really would like to have an
optimistic "mmap_upgrade_trylock()" to make it quicker to go from a
read-lock (for the common case) to taking the write lock (for having to
extend the vma) in the normal single-threaded situation where there is
no other locking activity.
But that _is_ all the very uncommon special case, so while it would be
nice to have such an operation, it probably doesn't matter in reality.
I did put in the skeleton code for such a possible future expansion,
even if it only acts as pseudo-documentation for what we're doing.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2023-06-15 22:17:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline bool mmap_upgrade_trylock(struct mm_struct *mm)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* We don't have this operation yet.
|
|
|
|
*
|
|
|
|
* It should be easy enough to do: it's basically a
|
|
|
|
* atomic_long_try_cmpxchg_acquire()
|
|
|
|
* from RWSEM_READER_BIAS -> RWSEM_WRITER_LOCKED, but
|
|
|
|
* it also needs the proper lockdep magic etc.
|
|
|
|
*/
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline bool upgrade_mmap_lock_carefully(struct mm_struct *mm, struct pt_regs *regs)
|
|
|
|
{
|
|
|
|
mmap_read_unlock(mm);
|
|
|
|
if (regs && !user_mode(regs)) {
|
|
|
|
unsigned long ip = instruction_pointer(regs);
|
|
|
|
if (!search_exception_tables(ip))
|
|
|
|
return false;
|
|
|
|
}
|
2023-06-15 23:17:48 +00:00
|
|
|
return !mmap_write_lock_killable(mm);
|
mm: introduce new 'lock_mm_and_find_vma()' page fault helper
commit c2508ec5a58db67093f4fb8bf89a9a7c53a109e9 upstream.
.. and make x86 use it.
This basically extracts the existing x86 "find and expand faulting vma"
code, but extends it to also take the mmap lock for writing in case we
actually do need to expand the vma.
We've historically short-circuited that case, and have some rather ugly
special logic to serialize the stack segment expansion (since we only
hold the mmap lock for reading) that doesn't match the normal VM
locking.
That slight violation of locking worked well, right up until it didn't:
the maple tree code really does want proper locking even for simple
extension of an existing vma.
So extract the code for "look up the vma of the fault" from x86, fix it
up to do the necessary write locking, and make it available as a helper
function for other architectures that can use the common helper.
Note: I say "common helper", but it really only handles the normal
stack-grows-down case. Which is all architectures except for PA-RISC
and IA64. So some rare architectures can't use the helper, but if they
care they'll just need to open-code this logic.
It's also worth pointing out that this code really would like to have an
optimistic "mmap_upgrade_trylock()" to make it quicker to go from a
read-lock (for the common case) to taking the write lock (for having to
extend the vma) in the normal single-threaded situation where there is
no other locking activity.
But that _is_ all the very uncommon special case, so while it would be
nice to have such an operation, it probably doesn't matter in reality.
I did put in the skeleton code for such a possible future expansion,
even if it only acts as pseudo-documentation for what we're doing.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2023-06-15 22:17:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Helper for page fault handling.
|
|
|
|
*
|
|
|
|
* This is kind of equivalend to "mmap_read_lock()" followed
|
|
|
|
* by "find_extend_vma()", except it's a lot more careful about
|
|
|
|
* the locking (and will drop the lock on failure).
|
|
|
|
*
|
|
|
|
* For example, if we have a kernel bug that causes a page
|
|
|
|
* fault, we don't want to just use mmap_read_lock() to get
|
|
|
|
* the mm lock, because that would deadlock if the bug were
|
|
|
|
* to happen while we're holding the mm lock for writing.
|
|
|
|
*
|
|
|
|
* So this checks the exception tables on kernel faults in
|
|
|
|
* order to only do this all for instructions that are actually
|
|
|
|
* expected to fault.
|
|
|
|
*
|
|
|
|
* We can also actually take the mm lock for writing if we
|
|
|
|
* need to extend the vma, which helps the VM layer a lot.
|
|
|
|
*/
|
|
|
|
struct vm_area_struct *lock_mm_and_find_vma(struct mm_struct *mm,
|
|
|
|
unsigned long addr, struct pt_regs *regs)
|
|
|
|
{
|
|
|
|
struct vm_area_struct *vma;
|
|
|
|
|
|
|
|
if (!get_mmap_lock_carefully(mm, regs))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
vma = find_vma(mm, addr);
|
|
|
|
if (likely(vma && (vma->vm_start <= addr)))
|
|
|
|
return vma;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Well, dang. We might still be successful, but only
|
|
|
|
* if we can extend a vma to do so.
|
|
|
|
*/
|
|
|
|
if (!vma || !(vma->vm_flags & VM_GROWSDOWN)) {
|
|
|
|
mmap_read_unlock(mm);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We can try to upgrade the mmap lock atomically,
|
|
|
|
* in which case we can continue to use the vma
|
|
|
|
* we already looked up.
|
|
|
|
*
|
|
|
|
* Otherwise we'll have to drop the mmap lock and
|
|
|
|
* re-take it, and also look up the vma again,
|
|
|
|
* re-checking it.
|
|
|
|
*/
|
|
|
|
if (!mmap_upgrade_trylock(mm)) {
|
|
|
|
if (!upgrade_mmap_lock_carefully(mm, regs))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
vma = find_vma(mm, addr);
|
|
|
|
if (!vma)
|
|
|
|
goto fail;
|
|
|
|
if (vma->vm_start <= addr)
|
|
|
|
goto success;
|
|
|
|
if (!(vma->vm_flags & VM_GROWSDOWN))
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
2023-06-24 20:45:51 +00:00
|
|
|
if (expand_stack_locked(vma, addr))
|
mm: introduce new 'lock_mm_and_find_vma()' page fault helper
commit c2508ec5a58db67093f4fb8bf89a9a7c53a109e9 upstream.
.. and make x86 use it.
This basically extracts the existing x86 "find and expand faulting vma"
code, but extends it to also take the mmap lock for writing in case we
actually do need to expand the vma.
We've historically short-circuited that case, and have some rather ugly
special logic to serialize the stack segment expansion (since we only
hold the mmap lock for reading) that doesn't match the normal VM
locking.
That slight violation of locking worked well, right up until it didn't:
the maple tree code really does want proper locking even for simple
extension of an existing vma.
So extract the code for "look up the vma of the fault" from x86, fix it
up to do the necessary write locking, and make it available as a helper
function for other architectures that can use the common helper.
Note: I say "common helper", but it really only handles the normal
stack-grows-down case. Which is all architectures except for PA-RISC
and IA64. So some rare architectures can't use the helper, but if they
care they'll just need to open-code this logic.
It's also worth pointing out that this code really would like to have an
optimistic "mmap_upgrade_trylock()" to make it quicker to go from a
read-lock (for the common case) to taking the write lock (for having to
extend the vma) in the normal single-threaded situation where there is
no other locking activity.
But that _is_ all the very uncommon special case, so while it would be
nice to have such an operation, it probably doesn't matter in reality.
I did put in the skeleton code for such a possible future expansion,
even if it only acts as pseudo-documentation for what we're doing.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2023-06-15 22:17:36 +00:00
|
|
|
goto fail;
|
|
|
|
|
|
|
|
success:
|
|
|
|
mmap_write_downgrade(mm);
|
|
|
|
return vma;
|
|
|
|
|
|
|
|
fail:
|
|
|
|
mmap_write_unlock(mm);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2023-02-27 17:36:22 +00:00
|
|
|
#ifdef CONFIG_PER_VMA_LOCK
|
|
|
|
/*
|
|
|
|
* Lookup and lock a VMA under RCU protection. Returned VMA is guaranteed to be
|
|
|
|
* stable and not isolated. If the VMA is not found or is being modified the
|
|
|
|
* function returns NULL.
|
|
|
|
*/
|
|
|
|
struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
|
|
|
|
unsigned long address)
|
|
|
|
{
|
|
|
|
MA_STATE(mas, &mm->mm_mt, address, address);
|
|
|
|
struct vm_area_struct *vma;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
retry:
|
|
|
|
vma = mas_walk(&mas);
|
|
|
|
if (!vma)
|
|
|
|
goto inval;
|
|
|
|
|
|
|
|
/* Only anonymous vmas are supported for now */
|
|
|
|
if (!vma_is_anonymous(vma))
|
|
|
|
goto inval;
|
|
|
|
|
|
|
|
if (!vma_start_read(vma))
|
|
|
|
goto inval;
|
|
|
|
|
2023-07-26 21:41:03 +00:00
|
|
|
/*
|
|
|
|
* find_mergeable_anon_vma uses adjacent vmas which are not locked.
|
|
|
|
* This check must happen after vma_start_read(); otherwise, a
|
|
|
|
* concurrent mremap() with MREMAP_DONTUNMAP could dissociate the VMA
|
|
|
|
* from its anon_vma.
|
|
|
|
*/
|
|
|
|
if (unlikely(!vma->anon_vma))
|
|
|
|
goto inval_end_read;
|
|
|
|
|
2023-02-27 17:36:26 +00:00
|
|
|
/*
|
|
|
|
* Due to the possibility of userfault handler dropping mmap_lock, avoid
|
|
|
|
* it for now and fall back to page fault handling under mmap_lock.
|
|
|
|
*/
|
2023-07-26 21:41:03 +00:00
|
|
|
if (userfaultfd_armed(vma))
|
|
|
|
goto inval_end_read;
|
2023-02-27 17:36:26 +00:00
|
|
|
|
2023-02-27 17:36:22 +00:00
|
|
|
/* Check since vm_start/vm_end might change before we lock the VMA */
|
2023-07-26 21:41:03 +00:00
|
|
|
if (unlikely(address < vma->vm_start || address >= vma->vm_end))
|
|
|
|
goto inval_end_read;
|
2023-02-27 17:36:22 +00:00
|
|
|
|
|
|
|
/* Check if the VMA got isolated after we found it */
|
|
|
|
if (vma->detached) {
|
|
|
|
vma_end_read(vma);
|
2023-02-27 17:36:27 +00:00
|
|
|
count_vm_vma_lock_event(VMA_LOCK_MISS);
|
2023-02-27 17:36:22 +00:00
|
|
|
/* The area was replaced with another one */
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
|
|
|
return vma;
|
2023-07-26 21:41:03 +00:00
|
|
|
|
|
|
|
inval_end_read:
|
|
|
|
vma_end_read(vma);
|
2023-02-27 17:36:22 +00:00
|
|
|
inval:
|
|
|
|
rcu_read_unlock();
|
2023-02-27 17:36:27 +00:00
|
|
|
count_vm_vma_lock_event(VMA_LOCK_ABORT);
|
2023-02-27 17:36:22 +00:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_PER_VMA_LOCK */
|
|
|
|
|
2017-03-09 14:24:08 +00:00
|
|
|
#ifndef __PAGETABLE_P4D_FOLDED
|
|
|
|
/*
|
|
|
|
* Allocate p4d page table.
|
|
|
|
* We've already handled the fast-path in-line.
|
|
|
|
*/
|
|
|
|
int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
|
|
|
|
{
|
|
|
|
p4d_t *new = p4d_alloc_one(mm, address);
|
|
|
|
if (!new)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
spin_lock(&mm->page_table_lock);
|
2021-11-05 20:38:41 +00:00
|
|
|
if (pgd_present(*pgd)) { /* Another has populated it */
|
2017-03-09 14:24:08 +00:00
|
|
|
p4d_free(mm, new);
|
2021-11-05 20:38:41 +00:00
|
|
|
} else {
|
|
|
|
smp_wmb(); /* See comment in pmd_install() */
|
2017-03-09 14:24:08 +00:00
|
|
|
pgd_populate(mm, pgd, new);
|
2021-11-05 20:38:41 +00:00
|
|
|
}
|
2017-03-09 14:24:08 +00:00
|
|
|
spin_unlock(&mm->page_table_lock);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#endif /* __PAGETABLE_P4D_FOLDED */
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
#ifndef __PAGETABLE_PUD_FOLDED
|
|
|
|
/*
|
|
|
|
* Allocate page upper directory.
|
[PATCH] mm: init_mm without ptlock
First step in pushing down the page_table_lock. init_mm.page_table_lock has
been used throughout the architectures (usually for ioremap): not to serialize
kernel address space allocation (that's usually vmlist_lock), but because
pud_alloc,pmd_alloc,pte_alloc_kernel expect caller holds it.
Reverse that: don't lock or unlock init_mm.page_table_lock in any of the
architectures; instead rely on pud_alloc,pmd_alloc,pte_alloc_kernel to take
and drop it when allocating a new one, to check lest a racing task already
did. Similarly no page_table_lock in vmalloc's map_vm_area.
Some temporary ugliness in __pud_alloc and __pmd_alloc: since they also handle
user mms, which are converted only by a later patch, for now they have to lock
differently according to whether or not it's init_mm.
If sources get muddled, there's a danger that an arch source taking
init_mm.page_table_lock will be mixed with common source also taking it (or
neither take it). So break the rules and make another change, which should
break the build for such a mismatch: remove the redundant mm arg from
pte_alloc_kernel (ppc64 scrapped its distinct ioremap_mm in 2.6.13).
Exceptions: arm26 used pte_alloc_kernel on user mm, now pte_alloc_map; ia64
used pte_alloc_map on init_mm, now pte_alloc_kernel; parisc had bad args to
pmd_alloc and pte_alloc_kernel in unused USE_HPPA_IOREMAP code; ppc64
map_io_page forgot to unlock on failure; ppc mmu_mapin_ram and ppc64 im_free
took page_table_lock for no good reason.
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 01:16:21 +00:00
|
|
|
* We've already handled the fast-path in-line.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2017-03-09 14:24:07 +00:00
|
|
|
int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2005-10-30 01:16:23 +00:00
|
|
|
pud_t *new = pud_alloc_one(mm, address);
|
|
|
|
if (!new)
|
2005-10-30 01:16:22 +00:00
|
|
|
return -ENOMEM;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
[PATCH] mm: init_mm without ptlock
First step in pushing down the page_table_lock. init_mm.page_table_lock has
been used throughout the architectures (usually for ioremap): not to serialize
kernel address space allocation (that's usually vmlist_lock), but because
pud_alloc,pmd_alloc,pte_alloc_kernel expect caller holds it.
Reverse that: don't lock or unlock init_mm.page_table_lock in any of the
architectures; instead rely on pud_alloc,pmd_alloc,pte_alloc_kernel to take
and drop it when allocating a new one, to check lest a racing task already
did. Similarly no page_table_lock in vmalloc's map_vm_area.
Some temporary ugliness in __pud_alloc and __pmd_alloc: since they also handle
user mms, which are converted only by a later patch, for now they have to lock
differently according to whether or not it's init_mm.
If sources get muddled, there's a danger that an arch source taking
init_mm.page_table_lock will be mixed with common source also taking it (or
neither take it). So break the rules and make another change, which should
break the build for such a mismatch: remove the redundant mm arg from
pte_alloc_kernel (ppc64 scrapped its distinct ioremap_mm in 2.6.13).
Exceptions: arm26 used pte_alloc_kernel on user mm, now pte_alloc_map; ia64
used pte_alloc_map on init_mm, now pte_alloc_kernel; parisc had bad args to
pmd_alloc and pte_alloc_kernel in unused USE_HPPA_IOREMAP code; ppc64
map_io_page forgot to unlock on failure; ppc mmu_mapin_ram and ppc64 im_free
took page_table_lock for no good reason.
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 01:16:21 +00:00
|
|
|
spin_lock(&mm->page_table_lock);
|
2017-11-16 01:35:33 +00:00
|
|
|
if (!p4d_present(*p4d)) {
|
|
|
|
mm_inc_nr_puds(mm);
|
2021-11-05 20:38:41 +00:00
|
|
|
smp_wmb(); /* See comment in pmd_install() */
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d_populate(mm, p4d, new);
|
2017-11-16 01:35:33 +00:00
|
|
|
} else /* Another has populated it */
|
2008-02-05 06:29:14 +00:00
|
|
|
pud_free(mm, new);
|
2005-10-30 01:16:23 +00:00
|
|
|
spin_unlock(&mm->page_table_lock);
|
2005-10-30 01:16:22 +00:00
|
|
|
return 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
#endif /* __PAGETABLE_PUD_FOLDED */
|
|
|
|
|
|
|
|
#ifndef __PAGETABLE_PMD_FOLDED
|
|
|
|
/*
|
|
|
|
* Allocate page middle directory.
|
[PATCH] mm: init_mm without ptlock
First step in pushing down the page_table_lock. init_mm.page_table_lock has
been used throughout the architectures (usually for ioremap): not to serialize
kernel address space allocation (that's usually vmlist_lock), but because
pud_alloc,pmd_alloc,pte_alloc_kernel expect caller holds it.
Reverse that: don't lock or unlock init_mm.page_table_lock in any of the
architectures; instead rely on pud_alloc,pmd_alloc,pte_alloc_kernel to take
and drop it when allocating a new one, to check lest a racing task already
did. Similarly no page_table_lock in vmalloc's map_vm_area.
Some temporary ugliness in __pud_alloc and __pmd_alloc: since they also handle
user mms, which are converted only by a later patch, for now they have to lock
differently according to whether or not it's init_mm.
If sources get muddled, there's a danger that an arch source taking
init_mm.page_table_lock will be mixed with common source also taking it (or
neither take it). So break the rules and make another change, which should
break the build for such a mismatch: remove the redundant mm arg from
pte_alloc_kernel (ppc64 scrapped its distinct ioremap_mm in 2.6.13).
Exceptions: arm26 used pte_alloc_kernel on user mm, now pte_alloc_map; ia64
used pte_alloc_map on init_mm, now pte_alloc_kernel; parisc had bad args to
pmd_alloc and pte_alloc_kernel in unused USE_HPPA_IOREMAP code; ppc64
map_io_page forgot to unlock on failure; ppc mmu_mapin_ram and ppc64 im_free
took page_table_lock for no good reason.
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 01:16:21 +00:00
|
|
|
* We've already handled the fast-path in-line.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2005-10-30 01:16:22 +00:00
|
|
|
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2017-02-24 22:57:02 +00:00
|
|
|
spinlock_t *ptl;
|
2005-10-30 01:16:23 +00:00
|
|
|
pmd_t *new = pmd_alloc_one(mm, address);
|
|
|
|
if (!new)
|
2005-10-30 01:16:22 +00:00
|
|
|
return -ENOMEM;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2017-02-24 22:57:02 +00:00
|
|
|
ptl = pud_lock(mm, pud);
|
mm: account pmd page tables to the process
Dave noticed that unprivileged process can allocate significant amount of
memory -- >500 MiB on x86_64 -- and stay unnoticed by oom-killer and
memory cgroup. The trick is to allocate a lot of PMD page tables. Linux
kernel doesn't account PMD tables to the process, only PTE.
The use-cases below use few tricks to allocate a lot of PMD page tables
while keeping VmRSS and VmPTE low. oom_score for the process will be 0.
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/prctl.h>
#define PUD_SIZE (1UL << 30)
#define PMD_SIZE (1UL << 21)
#define NR_PUD 130000
int main(void)
{
char *addr = NULL;
unsigned long i;
prctl(PR_SET_THP_DISABLE);
for (i = 0; i < NR_PUD ; i++) {
addr = mmap(addr + PUD_SIZE, PUD_SIZE, PROT_WRITE|PROT_READ,
MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
if (addr == MAP_FAILED) {
perror("mmap");
break;
}
*addr = 'x';
munmap(addr, PMD_SIZE);
mmap(addr, PMD_SIZE, PROT_WRITE|PROT_READ,
MAP_ANONYMOUS|MAP_PRIVATE|MAP_FIXED, -1, 0);
if (addr == MAP_FAILED)
perror("re-mmap"), exit(1);
}
printf("PID %d consumed %lu KiB in PMD page tables\n",
getpid(), i * 4096 >> 10);
return pause();
}
The patch addresses the issue by account PMD tables to the process the
same way we account PTE.
The main place where PMD tables is accounted is __pmd_alloc() and
free_pmd_range(). But there're few corner cases:
- HugeTLB can share PMD page tables. The patch handles by accounting
the table to all processes who share it.
- x86 PAE pre-allocates few PMD tables on fork.
- Architectures with FIRST_USER_ADDRESS > 0. We need to adjust sanity
check on exit(2).
Accounting only happens on configuration where PMD page table's level is
present (PMD is not folded). As with nr_ptes we use per-mm counter. The
counter value is used to calculate baseline for badness score by
oom-killer.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: David Rientjes <rientjes@google.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-11 23:26:50 +00:00
|
|
|
if (!pud_present(*pud)) {
|
|
|
|
mm_inc_nr_pmds(mm);
|
2021-11-05 20:38:41 +00:00
|
|
|
smp_wmb(); /* See comment in pmd_install() */
|
2005-10-30 01:16:22 +00:00
|
|
|
pud_populate(mm, pud, new);
|
2021-11-05 20:38:41 +00:00
|
|
|
} else { /* Another has populated it */
|
2008-02-05 06:29:14 +00:00
|
|
|
pmd_free(mm, new);
|
2021-11-05 20:38:41 +00:00
|
|
|
}
|
2017-02-24 22:57:02 +00:00
|
|
|
spin_unlock(ptl);
|
2005-10-30 01:16:22 +00:00
|
|
|
return 0;
|
[PATCH] Workaround for gcc 2.96 (undefined references)
LD .tmp_vmlinux1
mm/built-in.o(.text+0x100d6): In function `copy_page_range':
: undefined reference to `__pud_alloc'
mm/built-in.o(.text+0x1010b): In function `copy_page_range':
: undefined reference to `__pmd_alloc'
mm/built-in.o(.text+0x11ef4): In function `__handle_mm_fault':
: undefined reference to `__pud_alloc'
fs/built-in.o(.text+0xc930): In function `install_arg_page':
: undefined reference to `__pud_alloc'
make: *** [.tmp_vmlinux1] Error 1
Those missing references in mm/memory.c arise from this code in
include/linux/mm.h, combined with the fact that __PGTABLE_PMD_FOLDED and
__PGTABLE_PUD_FOLDED are both set and __ARCH_HAS_4LEVEL_HACK is not:
/*
* The following ifdef needed to get the 4level-fixup.h header to work.
* Remove it when 4level-fixup.h has been removed.
*/
#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
{
return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
NULL: pud_offset(pgd, address);
}
static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
{
return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
NULL: pmd_offset(pud, address);
}
#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
With my configuration the pgd_none and pud_none routines are inlines
returning a constant 0. Apparently the old compiler avoids generating
calls to __pud_alloc and __pmd_alloc but still lists them as undefined
references in the module's symbol table.
I don't know which change caused this problem. I think it was added
somewhere between 2.6.14 and 2.6.15-rc1, because I remember building
several 2.6.14-rc kernels without difficulty. However I can't point to an
individual culprit.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-28 21:43:44 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
#endif /* __PAGETABLE_PMD_FOLDED */
|
|
|
|
|
2022-04-29 06:16:10 +00:00
|
|
|
/**
|
|
|
|
* follow_pte - look up PTE at a user virtual address
|
|
|
|
* @mm: the mm_struct of the target address space
|
|
|
|
* @address: user virtual address
|
|
|
|
* @ptepp: location to store found PTE
|
|
|
|
* @ptlp: location to store the lock for the PTE
|
|
|
|
*
|
|
|
|
* On a successful return, the pointer to the PTE is stored in @ptepp;
|
|
|
|
* the corresponding lock is taken and its location is stored in @ptlp.
|
|
|
|
* The contents of the PTE are only stable until @ptlp is released;
|
|
|
|
* any further use, if any, must be protected against invalidation
|
|
|
|
* with MMU notifiers.
|
|
|
|
*
|
|
|
|
* Only IO mappings and raw PFN mappings are allowed. The mmap semaphore
|
|
|
|
* should be taken for read.
|
|
|
|
*
|
|
|
|
* KVM uses this function. While it is arguably less bad than ``follow_pfn``,
|
|
|
|
* it is not a good general-purpose API.
|
|
|
|
*
|
|
|
|
* Return: zero on success, -ve otherwise.
|
|
|
|
*/
|
|
|
|
int follow_pte(struct mm_struct *mm, unsigned long address,
|
|
|
|
pte_t **ptepp, spinlock_t **ptlp)
|
2009-06-16 22:32:33 +00:00
|
|
|
{
|
|
|
|
pgd_t *pgd;
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d_t *p4d;
|
2009-06-16 22:32:33 +00:00
|
|
|
pud_t *pud;
|
|
|
|
pmd_t *pmd;
|
|
|
|
pte_t *ptep;
|
|
|
|
|
|
|
|
pgd = pgd_offset(mm, address);
|
|
|
|
if (pgd_none(*pgd) || unlikely(pgd_bad(*pgd)))
|
|
|
|
goto out;
|
|
|
|
|
2017-03-09 14:24:07 +00:00
|
|
|
p4d = p4d_offset(pgd, address);
|
|
|
|
if (p4d_none(*p4d) || unlikely(p4d_bad(*p4d)))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
pud = pud_offset(p4d, address);
|
2009-06-16 22:32:33 +00:00
|
|
|
if (pud_none(*pud) || unlikely(pud_bad(*pud)))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
pmd = pmd_offset(pud, address);
|
2011-01-13 23:46:54 +00:00
|
|
|
VM_BUG_ON(pmd_trans_huge(*pmd));
|
2009-06-16 22:32:33 +00:00
|
|
|
|
2017-01-11 00:57:21 +00:00
|
|
|
if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
|
2009-06-16 22:32:33 +00:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
ptep = pte_offset_map_lock(mm, pmd, address, ptlp);
|
|
|
|
if (!pte_present(*ptep))
|
|
|
|
goto unlock;
|
|
|
|
*ptepp = ptep;
|
|
|
|
return 0;
|
|
|
|
unlock:
|
|
|
|
pte_unmap_unlock(ptep, *ptlp);
|
|
|
|
out:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2021-02-05 10:07:11 +00:00
|
|
|
EXPORT_SYMBOL_GPL(follow_pte);
|
|
|
|
|
2009-06-16 22:32:35 +00:00
|
|
|
/**
|
|
|
|
* follow_pfn - look up PFN at a user virtual address
|
|
|
|
* @vma: memory mapping
|
|
|
|
* @address: user virtual address
|
|
|
|
* @pfn: location to store found PFN
|
|
|
|
*
|
|
|
|
* Only IO mappings and raw PFN mappings are allowed.
|
|
|
|
*
|
2021-02-05 10:07:11 +00:00
|
|
|
* This function does not allow the caller to read the permissions
|
|
|
|
* of the PTE. Do not use it.
|
|
|
|
*
|
2019-03-05 23:48:42 +00:00
|
|
|
* Return: zero and the pfn at @pfn on success, -ve otherwise.
|
2009-06-16 22:32:35 +00:00
|
|
|
*/
|
|
|
|
int follow_pfn(struct vm_area_struct *vma, unsigned long address,
|
|
|
|
unsigned long *pfn)
|
|
|
|
{
|
|
|
|
int ret = -EINVAL;
|
|
|
|
spinlock_t *ptl;
|
|
|
|
pte_t *ptep;
|
|
|
|
|
|
|
|
if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
|
|
|
|
return ret;
|
|
|
|
|
2021-02-05 10:07:11 +00:00
|
|
|
ret = follow_pte(vma->vm_mm, address, &ptep, &ptl);
|
2009-06-16 22:32:35 +00:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
*pfn = pte_pfn(*ptep);
|
|
|
|
pte_unmap_unlock(ptep, ptl);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(follow_pfn);
|
|
|
|
|
2008-07-24 04:27:05 +00:00
|
|
|
#ifdef CONFIG_HAVE_IOREMAP_PROT
|
2008-12-19 21:47:27 +00:00
|
|
|
int follow_phys(struct vm_area_struct *vma,
|
|
|
|
unsigned long address, unsigned int flags,
|
|
|
|
unsigned long *prot, resource_size_t *phys)
|
2008-07-24 04:27:05 +00:00
|
|
|
{
|
2009-06-16 22:32:34 +00:00
|
|
|
int ret = -EINVAL;
|
2008-07-24 04:27:05 +00:00
|
|
|
pte_t *ptep, pte;
|
|
|
|
spinlock_t *ptl;
|
|
|
|
|
2008-12-19 21:47:27 +00:00
|
|
|
if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
|
|
|
|
goto out;
|
2008-07-24 04:27:05 +00:00
|
|
|
|
2021-02-05 10:07:11 +00:00
|
|
|
if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
|
2008-12-19 21:47:27 +00:00
|
|
|
goto out;
|
2008-07-24 04:27:05 +00:00
|
|
|
pte = *ptep;
|
2009-06-16 22:32:34 +00:00
|
|
|
|
Revert "mm: replace p??_write with pte_access_permitted in fault + gup paths"
This reverts commits 5c9d2d5c269c, c7da82b894e9, and e7fe7b5cae90.
We'll probably need to revisit this, but basically we should not
complicate the get_user_pages_fast() case, and checking the actual page
table protection key bits will require more care anyway, since the
protection keys depend on the exact state of the VM in question.
Particularly when doing a "remote" page lookup (ie in somebody elses VM,
not your own), you need to be much more careful than this was. Dave
Hansen says:
"So, the underlying bug here is that we now a get_user_pages_remote()
and then go ahead and do the p*_access_permitted() checks against the
current PKRU. This was introduced recently with the addition of the
new p??_access_permitted() calls.
We have checks in the VMA path for the "remote" gups and we avoid
consulting PKRU for them. This got missed in the pkeys selftests
because I did a ptrace read, but not a *write*. I also didn't
explicitly test it against something where a COW needed to be done"
It's also not entirely clear that it makes sense to check the protection
key bits at this level at all. But one possible eventual solution is to
make the get_user_pages_fast() case just abort if it sees protection key
bits set, which makes us fall back to the regular get_user_pages() case,
which then has a vma and can do the check there if we want to.
We'll see.
Somewhat related to this all: what we _do_ want to do some day is to
check the PAGE_USER bit - it should obviously always be set for user
pages, but it would be a good check to have back. Because we have no
generic way to test for it, we lost it as part of moving over from the
architecture-specific x86 GUP implementation to the generic one in
commit e585513b76f7 ("x86/mm/gup: Switch GUP to the generic
get_user_page_fast() implementation").
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-12-16 02:53:22 +00:00
|
|
|
if ((flags & FOLL_WRITE) && !pte_write(pte))
|
2008-07-24 04:27:05 +00:00
|
|
|
goto unlock;
|
|
|
|
|
|
|
|
*prot = pgprot_val(pte_pgprot(pte));
|
2009-06-16 22:32:34 +00:00
|
|
|
*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
|
2008-07-24 04:27:05 +00:00
|
|
|
|
2009-06-16 22:32:34 +00:00
|
|
|
ret = 0;
|
2008-07-24 04:27:05 +00:00
|
|
|
unlock:
|
|
|
|
pte_unmap_unlock(ptep, ptl);
|
|
|
|
out:
|
2008-12-19 21:47:27 +00:00
|
|
|
return ret;
|
2008-07-24 04:27:05 +00:00
|
|
|
}
|
|
|
|
|
2020-11-27 16:41:21 +00:00
|
|
|
/**
|
|
|
|
* generic_access_phys - generic implementation for iomem mmap access
|
|
|
|
* @vma: the vma to access
|
2021-05-07 01:06:47 +00:00
|
|
|
* @addr: userspace address, not relative offset within @vma
|
2020-11-27 16:41:21 +00:00
|
|
|
* @buf: buffer to read/write
|
|
|
|
* @len: length of transfer
|
|
|
|
* @write: set to FOLL_WRITE when writing, otherwise reading
|
|
|
|
*
|
|
|
|
* This is a generic implementation for &vm_operations_struct.access for an
|
|
|
|
* iomem mapping. This callback is used by access_process_vm() when the @vma is
|
|
|
|
* not page based.
|
|
|
|
*/
|
2008-07-24 04:27:05 +00:00
|
|
|
int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
|
|
|
|
void *buf, int len, int write)
|
|
|
|
{
|
|
|
|
resource_size_t phys_addr;
|
|
|
|
unsigned long prot = 0;
|
2009-01-06 22:39:43 +00:00
|
|
|
void __iomem *maddr;
|
2020-11-27 16:41:21 +00:00
|
|
|
pte_t *ptep, pte;
|
|
|
|
spinlock_t *ptl;
|
|
|
|
int offset = offset_in_page(addr);
|
|
|
|
int ret = -EINVAL;
|
|
|
|
|
|
|
|
if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
retry:
|
2021-02-23 01:45:02 +00:00
|
|
|
if (follow_pte(vma->vm_mm, addr, &ptep, &ptl))
|
2020-11-27 16:41:21 +00:00
|
|
|
return -EINVAL;
|
|
|
|
pte = *ptep;
|
|
|
|
pte_unmap_unlock(ptep, ptl);
|
2008-07-24 04:27:05 +00:00
|
|
|
|
2020-11-27 16:41:21 +00:00
|
|
|
prot = pgprot_val(pte_pgprot(pte));
|
|
|
|
phys_addr = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
|
|
|
|
|
|
|
|
if ((write & FOLL_WRITE) && !pte_write(pte))
|
2008-07-24 04:27:05 +00:00
|
|
|
return -EINVAL;
|
|
|
|
|
2015-02-12 23:00:19 +00:00
|
|
|
maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
|
2018-08-11 00:23:06 +00:00
|
|
|
if (!maddr)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2021-02-23 01:45:02 +00:00
|
|
|
if (follow_pte(vma->vm_mm, addr, &ptep, &ptl))
|
2020-11-27 16:41:21 +00:00
|
|
|
goto out_unmap;
|
|
|
|
|
|
|
|
if (!pte_same(pte, *ptep)) {
|
|
|
|
pte_unmap_unlock(ptep, ptl);
|
|
|
|
iounmap(maddr);
|
|
|
|
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
|
2008-07-24 04:27:05 +00:00
|
|
|
if (write)
|
|
|
|
memcpy_toio(maddr + offset, buf, len);
|
|
|
|
else
|
|
|
|
memcpy_fromio(buf, maddr + offset, len);
|
2020-11-27 16:41:21 +00:00
|
|
|
ret = len;
|
|
|
|
pte_unmap_unlock(ptep, ptl);
|
|
|
|
out_unmap:
|
2008-07-24 04:27:05 +00:00
|
|
|
iounmap(maddr);
|
|
|
|
|
2020-11-27 16:41:21 +00:00
|
|
|
return ret;
|
2008-07-24 04:27:05 +00:00
|
|
|
}
|
2013-08-07 11:02:52 +00:00
|
|
|
EXPORT_SYMBOL_GPL(generic_access_phys);
|
2008-07-24 04:27:05 +00:00
|
|
|
#endif
|
|
|
|
|
2006-09-27 08:50:15 +00:00
|
|
|
/*
|
2020-12-15 03:07:45 +00:00
|
|
|
* Access another process' address space as given in mm.
|
2006-09-27 08:50:15 +00:00
|
|
|
*/
|
2020-12-15 03:07:45 +00:00
|
|
|
int __access_remote_vm(struct mm_struct *mm, unsigned long addr, void *buf,
|
|
|
|
int len, unsigned int gup_flags)
|
2006-09-27 08:50:15 +00:00
|
|
|
{
|
|
|
|
struct vm_area_struct *vma;
|
|
|
|
void *old_buf = buf;
|
2016-10-13 00:20:18 +00:00
|
|
|
int write = gup_flags & FOLL_WRITE;
|
2006-09-27 08:50:15 +00:00
|
|
|
|
2020-06-09 04:33:25 +00:00
|
|
|
if (mmap_read_lock_killable(mm))
|
2019-07-12 04:00:07 +00:00
|
|
|
return 0;
|
|
|
|
|
2023-06-24 20:45:51 +00:00
|
|
|
/* We might need to expand the stack to access it */
|
|
|
|
vma = vma_lookup(mm, addr);
|
|
|
|
if (!vma) {
|
|
|
|
vma = expand_stack(mm, addr);
|
|
|
|
if (!vma)
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2007-10-19 23:27:18 +00:00
|
|
|
/* ignore errors, just check how much was successfully transferred */
|
2006-09-27 08:50:15 +00:00
|
|
|
while (len) {
|
|
|
|
int bytes, ret, offset;
|
|
|
|
void *maddr;
|
2008-07-24 04:27:05 +00:00
|
|
|
struct page *page = NULL;
|
2006-09-27 08:50:15 +00:00
|
|
|
|
2020-08-12 01:39:01 +00:00
|
|
|
ret = get_user_pages_remote(mm, addr, 1,
|
2016-12-14 23:06:52 +00:00
|
|
|
gup_flags, &page, &vma, NULL);
|
2008-07-24 04:27:05 +00:00
|
|
|
if (ret <= 0) {
|
2014-08-06 23:08:12 +00:00
|
|
|
#ifndef CONFIG_HAVE_IOREMAP_PROT
|
|
|
|
break;
|
|
|
|
#else
|
2008-07-24 04:27:05 +00:00
|
|
|
/*
|
|
|
|
* Check if this is a VM_IO | VM_PFNMAP VMA, which
|
|
|
|
* we can access using slightly different code.
|
|
|
|
*/
|
2021-06-29 02:39:50 +00:00
|
|
|
vma = vma_lookup(mm, addr);
|
|
|
|
if (!vma)
|
2008-07-24 04:27:05 +00:00
|
|
|
break;
|
|
|
|
if (vma->vm_ops && vma->vm_ops->access)
|
|
|
|
ret = vma->vm_ops->access(vma, addr, buf,
|
|
|
|
len, write);
|
|
|
|
if (ret <= 0)
|
|
|
|
break;
|
|
|
|
bytes = ret;
|
2014-08-06 23:08:12 +00:00
|
|
|
#endif
|
2006-09-27 08:50:15 +00:00
|
|
|
} else {
|
2008-07-24 04:27:05 +00:00
|
|
|
bytes = len;
|
|
|
|
offset = addr & (PAGE_SIZE-1);
|
|
|
|
if (bytes > PAGE_SIZE-offset)
|
|
|
|
bytes = PAGE_SIZE-offset;
|
|
|
|
|
|
|
|
maddr = kmap(page);
|
|
|
|
if (write) {
|
|
|
|
copy_to_user_page(vma, page, addr,
|
|
|
|
maddr + offset, buf, bytes);
|
|
|
|
set_page_dirty_lock(page);
|
|
|
|
} else {
|
|
|
|
copy_from_user_page(vma, page, addr,
|
|
|
|
buf, maddr + offset, bytes);
|
|
|
|
}
|
|
|
|
kunmap(page);
|
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
|
|
|
put_page(page);
|
2006-09-27 08:50:15 +00:00
|
|
|
}
|
|
|
|
len -= bytes;
|
|
|
|
buf += bytes;
|
|
|
|
addr += bytes;
|
|
|
|
}
|
2020-06-09 04:33:25 +00:00
|
|
|
mmap_read_unlock(mm);
|
2006-09-27 08:50:15 +00:00
|
|
|
|
|
|
|
return buf - old_buf;
|
|
|
|
}
|
2008-01-30 12:33:18 +00:00
|
|
|
|
2011-03-13 19:49:20 +00:00
|
|
|
/**
|
2011-03-26 20:27:01 +00:00
|
|
|
* access_remote_vm - access another process' address space
|
2011-03-13 19:49:20 +00:00
|
|
|
* @mm: the mm_struct of the target address space
|
|
|
|
* @addr: start address to access
|
|
|
|
* @buf: source or destination buffer
|
|
|
|
* @len: number of bytes to transfer
|
2016-10-13 00:20:19 +00:00
|
|
|
* @gup_flags: flags modifying lookup behaviour
|
2011-03-13 19:49:20 +00:00
|
|
|
*
|
|
|
|
* The caller must hold a reference on @mm.
|
2019-03-05 23:48:42 +00:00
|
|
|
*
|
|
|
|
* Return: number of bytes copied from source to destination.
|
2011-03-13 19:49:20 +00:00
|
|
|
*/
|
|
|
|
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
|
2016-10-13 00:20:19 +00:00
|
|
|
void *buf, int len, unsigned int gup_flags)
|
2011-03-13 19:49:20 +00:00
|
|
|
{
|
2020-12-15 03:07:45 +00:00
|
|
|
return __access_remote_vm(mm, addr, buf, len, gup_flags);
|
2011-03-13 19:49:20 +00:00
|
|
|
}
|
|
|
|
|
2011-03-13 19:49:19 +00:00
|
|
|
/*
|
|
|
|
* Access another process' address space.
|
|
|
|
* Source/target buffer must be kernel space,
|
|
|
|
* Do not walk the page table directly, use get_user_pages
|
|
|
|
*/
|
|
|
|
int access_process_vm(struct task_struct *tsk, unsigned long addr,
|
2016-10-13 00:20:20 +00:00
|
|
|
void *buf, int len, unsigned int gup_flags)
|
2011-03-13 19:49:19 +00:00
|
|
|
{
|
|
|
|
struct mm_struct *mm;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
mm = get_task_mm(tsk);
|
|
|
|
if (!mm)
|
|
|
|
return 0;
|
|
|
|
|
2020-12-15 03:07:45 +00:00
|
|
|
ret = __access_remote_vm(mm, addr, buf, len, gup_flags);
|
2016-10-13 00:20:18 +00:00
|
|
|
|
2011-03-13 19:49:19 +00:00
|
|
|
mmput(mm);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
2016-11-01 21:43:25 +00:00
|
|
|
EXPORT_SYMBOL_GPL(access_process_vm);
|
2011-03-13 19:49:19 +00:00
|
|
|
|
2008-01-30 12:33:18 +00:00
|
|
|
/*
|
|
|
|
* Print the name of a VMA.
|
|
|
|
*/
|
|
|
|
void print_vma_addr(char *prefix, unsigned long ip)
|
|
|
|
{
|
|
|
|
struct mm_struct *mm = current->mm;
|
|
|
|
struct vm_area_struct *vma;
|
|
|
|
|
2008-02-13 19:21:06 +00:00
|
|
|
/*
|
2017-11-16 01:38:59 +00:00
|
|
|
* we might be running from an atomic context so we cannot sleep
|
2008-02-13 19:21:06 +00:00
|
|
|
*/
|
2020-06-09 04:33:25 +00:00
|
|
|
if (!mmap_read_trylock(mm))
|
2008-02-13 19:21:06 +00:00
|
|
|
return;
|
|
|
|
|
2008-01-30 12:33:18 +00:00
|
|
|
vma = find_vma(mm, ip);
|
|
|
|
if (vma && vma->vm_file) {
|
|
|
|
struct file *f = vma->vm_file;
|
2017-11-16 01:38:59 +00:00
|
|
|
char *buf = (char *)__get_free_page(GFP_NOWAIT);
|
2008-01-30 12:33:18 +00:00
|
|
|
if (buf) {
|
2012-12-18 00:01:23 +00:00
|
|
|
char *p;
|
2008-01-30 12:33:18 +00:00
|
|
|
|
2015-06-19 08:29:13 +00:00
|
|
|
p = file_path(f, buf, PAGE_SIZE);
|
2008-01-30 12:33:18 +00:00
|
|
|
if (IS_ERR(p))
|
|
|
|
p = "?";
|
2012-12-18 00:01:23 +00:00
|
|
|
printk("%s%s[%lx+%lx]", prefix, kbasename(p),
|
2008-01-30 12:33:18 +00:00
|
|
|
vma->vm_start,
|
|
|
|
vma->vm_end - vma->vm_start);
|
|
|
|
free_page((unsigned long)buf);
|
|
|
|
}
|
|
|
|
}
|
2020-06-09 04:33:25 +00:00
|
|
|
mmap_read_unlock(mm);
|
2008-01-30 12:33:18 +00:00
|
|
|
}
|
2008-09-10 11:37:17 +00:00
|
|
|
|
2013-05-26 14:32:23 +00:00
|
|
|
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
|
sched/preempt, mm/fault: Trigger might_sleep() in might_fault() with disabled pagefaults
Commit 662bbcb2747c ("mm, sched: Allow uaccess in atomic with
pagefault_disable()") removed might_sleep() checks for all user access
code (that uses might_fault()).
The reason was to disable wrong "sleep in atomic" warnings in the
following scenario:
pagefault_disable()
rc = copy_to_user(...)
pagefault_enable()
Which is valid, as pagefault_disable() increments the preempt counter
and therefore disables the pagefault handler. copy_to_user() will not
sleep and return an error code if a page is not available.
However, as all might_sleep() checks are removed,
CONFIG_DEBUG_ATOMIC_SLEEP would no longer detect the following scenario:
spin_lock(&lock);
rc = copy_to_user(...)
spin_unlock(&lock)
If the kernel is compiled with preemption turned on, preempt_disable()
will make in_atomic() detect disabled preemption. The fault handler would
correctly never sleep on user access.
However, with preemption turned off, preempt_disable() is usually a NOP
(with !CONFIG_PREEMPT_COUNT), therefore in_atomic() will not be able to
detect disabled preemption nor disabled pagefaults. The fault handler
could sleep.
We really want to enable CONFIG_DEBUG_ATOMIC_SLEEP checks for user access
functions again, otherwise we can end up with horrible deadlocks.
Root of all evil is that pagefault_disable() acts almost as
preempt_disable(), depending on preemption being turned on/off.
As we now have pagefault_disabled(), we can use it to distinguish
whether user acces functions might sleep.
Convert might_fault() into a makro that calls __might_fault(), to
allow proper file + line messages in case of a might_sleep() warning.
Reviewed-and-tested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David.Laight@ACULAB.COM
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: airlied@linux.ie
Cc: akpm@linux-foundation.org
Cc: benh@kernel.crashing.org
Cc: bigeasy@linutronix.de
Cc: borntraeger@de.ibm.com
Cc: daniel.vetter@intel.com
Cc: heiko.carstens@de.ibm.com
Cc: herbert@gondor.apana.org.au
Cc: hocko@suse.cz
Cc: hughd@google.com
Cc: mst@redhat.com
Cc: paulus@samba.org
Cc: ralf@linux-mips.org
Cc: schwidefsky@de.ibm.com
Cc: yang.shi@windriver.com
Link: http://lkml.kernel.org/r/1431359540-32227-3-git-send-email-dahi@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-05-11 15:52:07 +00:00
|
|
|
void __might_fault(const char *file, int line)
|
2008-09-10 11:37:17 +00:00
|
|
|
{
|
sched/preempt, mm/fault: Trigger might_sleep() in might_fault() with disabled pagefaults
Commit 662bbcb2747c ("mm, sched: Allow uaccess in atomic with
pagefault_disable()") removed might_sleep() checks for all user access
code (that uses might_fault()).
The reason was to disable wrong "sleep in atomic" warnings in the
following scenario:
pagefault_disable()
rc = copy_to_user(...)
pagefault_enable()
Which is valid, as pagefault_disable() increments the preempt counter
and therefore disables the pagefault handler. copy_to_user() will not
sleep and return an error code if a page is not available.
However, as all might_sleep() checks are removed,
CONFIG_DEBUG_ATOMIC_SLEEP would no longer detect the following scenario:
spin_lock(&lock);
rc = copy_to_user(...)
spin_unlock(&lock)
If the kernel is compiled with preemption turned on, preempt_disable()
will make in_atomic() detect disabled preemption. The fault handler would
correctly never sleep on user access.
However, with preemption turned off, preempt_disable() is usually a NOP
(with !CONFIG_PREEMPT_COUNT), therefore in_atomic() will not be able to
detect disabled preemption nor disabled pagefaults. The fault handler
could sleep.
We really want to enable CONFIG_DEBUG_ATOMIC_SLEEP checks for user access
functions again, otherwise we can end up with horrible deadlocks.
Root of all evil is that pagefault_disable() acts almost as
preempt_disable(), depending on preemption being turned on/off.
As we now have pagefault_disabled(), we can use it to distinguish
whether user acces functions might sleep.
Convert might_fault() into a makro that calls __might_fault(), to
allow proper file + line messages in case of a might_sleep() warning.
Reviewed-and-tested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David.Laight@ACULAB.COM
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: airlied@linux.ie
Cc: akpm@linux-foundation.org
Cc: benh@kernel.crashing.org
Cc: bigeasy@linutronix.de
Cc: borntraeger@de.ibm.com
Cc: daniel.vetter@intel.com
Cc: heiko.carstens@de.ibm.com
Cc: herbert@gondor.apana.org.au
Cc: hocko@suse.cz
Cc: hughd@google.com
Cc: mst@redhat.com
Cc: paulus@samba.org
Cc: ralf@linux-mips.org
Cc: schwidefsky@de.ibm.com
Cc: yang.shi@windriver.com
Link: http://lkml.kernel.org/r/1431359540-32227-3-git-send-email-dahi@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-05-11 15:52:07 +00:00
|
|
|
if (pagefault_disabled())
|
2013-05-26 14:32:23 +00:00
|
|
|
return;
|
2021-09-23 16:54:38 +00:00
|
|
|
__might_sleep(file, line);
|
sched/preempt, mm/fault: Trigger might_sleep() in might_fault() with disabled pagefaults
Commit 662bbcb2747c ("mm, sched: Allow uaccess in atomic with
pagefault_disable()") removed might_sleep() checks for all user access
code (that uses might_fault()).
The reason was to disable wrong "sleep in atomic" warnings in the
following scenario:
pagefault_disable()
rc = copy_to_user(...)
pagefault_enable()
Which is valid, as pagefault_disable() increments the preempt counter
and therefore disables the pagefault handler. copy_to_user() will not
sleep and return an error code if a page is not available.
However, as all might_sleep() checks are removed,
CONFIG_DEBUG_ATOMIC_SLEEP would no longer detect the following scenario:
spin_lock(&lock);
rc = copy_to_user(...)
spin_unlock(&lock)
If the kernel is compiled with preemption turned on, preempt_disable()
will make in_atomic() detect disabled preemption. The fault handler would
correctly never sleep on user access.
However, with preemption turned off, preempt_disable() is usually a NOP
(with !CONFIG_PREEMPT_COUNT), therefore in_atomic() will not be able to
detect disabled preemption nor disabled pagefaults. The fault handler
could sleep.
We really want to enable CONFIG_DEBUG_ATOMIC_SLEEP checks for user access
functions again, otherwise we can end up with horrible deadlocks.
Root of all evil is that pagefault_disable() acts almost as
preempt_disable(), depending on preemption being turned on/off.
As we now have pagefault_disabled(), we can use it to distinguish
whether user acces functions might sleep.
Convert might_fault() into a makro that calls __might_fault(), to
allow proper file + line messages in case of a might_sleep() warning.
Reviewed-and-tested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David.Laight@ACULAB.COM
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: airlied@linux.ie
Cc: akpm@linux-foundation.org
Cc: benh@kernel.crashing.org
Cc: bigeasy@linutronix.de
Cc: borntraeger@de.ibm.com
Cc: daniel.vetter@intel.com
Cc: heiko.carstens@de.ibm.com
Cc: herbert@gondor.apana.org.au
Cc: hocko@suse.cz
Cc: hughd@google.com
Cc: mst@redhat.com
Cc: paulus@samba.org
Cc: ralf@linux-mips.org
Cc: schwidefsky@de.ibm.com
Cc: yang.shi@windriver.com
Link: http://lkml.kernel.org/r/1431359540-32227-3-git-send-email-dahi@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-05-11 15:52:07 +00:00
|
|
|
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
|
2013-05-26 14:32:23 +00:00
|
|
|
if (current->mm)
|
2020-06-09 04:33:47 +00:00
|
|
|
might_lock_read(¤t->mm->mmap_lock);
|
sched/preempt, mm/fault: Trigger might_sleep() in might_fault() with disabled pagefaults
Commit 662bbcb2747c ("mm, sched: Allow uaccess in atomic with
pagefault_disable()") removed might_sleep() checks for all user access
code (that uses might_fault()).
The reason was to disable wrong "sleep in atomic" warnings in the
following scenario:
pagefault_disable()
rc = copy_to_user(...)
pagefault_enable()
Which is valid, as pagefault_disable() increments the preempt counter
and therefore disables the pagefault handler. copy_to_user() will not
sleep and return an error code if a page is not available.
However, as all might_sleep() checks are removed,
CONFIG_DEBUG_ATOMIC_SLEEP would no longer detect the following scenario:
spin_lock(&lock);
rc = copy_to_user(...)
spin_unlock(&lock)
If the kernel is compiled with preemption turned on, preempt_disable()
will make in_atomic() detect disabled preemption. The fault handler would
correctly never sleep on user access.
However, with preemption turned off, preempt_disable() is usually a NOP
(with !CONFIG_PREEMPT_COUNT), therefore in_atomic() will not be able to
detect disabled preemption nor disabled pagefaults. The fault handler
could sleep.
We really want to enable CONFIG_DEBUG_ATOMIC_SLEEP checks for user access
functions again, otherwise we can end up with horrible deadlocks.
Root of all evil is that pagefault_disable() acts almost as
preempt_disable(), depending on preemption being turned on/off.
As we now have pagefault_disabled(), we can use it to distinguish
whether user acces functions might sleep.
Convert might_fault() into a makro that calls __might_fault(), to
allow proper file + line messages in case of a might_sleep() warning.
Reviewed-and-tested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David.Laight@ACULAB.COM
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: airlied@linux.ie
Cc: akpm@linux-foundation.org
Cc: benh@kernel.crashing.org
Cc: bigeasy@linutronix.de
Cc: borntraeger@de.ibm.com
Cc: daniel.vetter@intel.com
Cc: heiko.carstens@de.ibm.com
Cc: herbert@gondor.apana.org.au
Cc: hocko@suse.cz
Cc: hughd@google.com
Cc: mst@redhat.com
Cc: paulus@samba.org
Cc: ralf@linux-mips.org
Cc: schwidefsky@de.ibm.com
Cc: yang.shi@windriver.com
Link: http://lkml.kernel.org/r/1431359540-32227-3-git-send-email-dahi@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-05-11 15:52:07 +00:00
|
|
|
#endif
|
2008-09-10 11:37:17 +00:00
|
|
|
}
|
sched/preempt, mm/fault: Trigger might_sleep() in might_fault() with disabled pagefaults
Commit 662bbcb2747c ("mm, sched: Allow uaccess in atomic with
pagefault_disable()") removed might_sleep() checks for all user access
code (that uses might_fault()).
The reason was to disable wrong "sleep in atomic" warnings in the
following scenario:
pagefault_disable()
rc = copy_to_user(...)
pagefault_enable()
Which is valid, as pagefault_disable() increments the preempt counter
and therefore disables the pagefault handler. copy_to_user() will not
sleep and return an error code if a page is not available.
However, as all might_sleep() checks are removed,
CONFIG_DEBUG_ATOMIC_SLEEP would no longer detect the following scenario:
spin_lock(&lock);
rc = copy_to_user(...)
spin_unlock(&lock)
If the kernel is compiled with preemption turned on, preempt_disable()
will make in_atomic() detect disabled preemption. The fault handler would
correctly never sleep on user access.
However, with preemption turned off, preempt_disable() is usually a NOP
(with !CONFIG_PREEMPT_COUNT), therefore in_atomic() will not be able to
detect disabled preemption nor disabled pagefaults. The fault handler
could sleep.
We really want to enable CONFIG_DEBUG_ATOMIC_SLEEP checks for user access
functions again, otherwise we can end up with horrible deadlocks.
Root of all evil is that pagefault_disable() acts almost as
preempt_disable(), depending on preemption being turned on/off.
As we now have pagefault_disabled(), we can use it to distinguish
whether user acces functions might sleep.
Convert might_fault() into a makro that calls __might_fault(), to
allow proper file + line messages in case of a might_sleep() warning.
Reviewed-and-tested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David.Laight@ACULAB.COM
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: airlied@linux.ie
Cc: akpm@linux-foundation.org
Cc: benh@kernel.crashing.org
Cc: bigeasy@linutronix.de
Cc: borntraeger@de.ibm.com
Cc: daniel.vetter@intel.com
Cc: heiko.carstens@de.ibm.com
Cc: herbert@gondor.apana.org.au
Cc: hocko@suse.cz
Cc: hughd@google.com
Cc: mst@redhat.com
Cc: paulus@samba.org
Cc: ralf@linux-mips.org
Cc: schwidefsky@de.ibm.com
Cc: yang.shi@windriver.com
Link: http://lkml.kernel.org/r/1431359540-32227-3-git-send-email-dahi@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-05-11 15:52:07 +00:00
|
|
|
EXPORT_SYMBOL(__might_fault);
|
2008-09-10 11:37:17 +00:00
|
|
|
#endif
|
2011-01-13 23:46:47 +00:00
|
|
|
|
|
|
|
#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
/*
|
|
|
|
* Process all subpages of the specified huge page with the specified
|
|
|
|
* operation. The target subpage will be processed last to keep its
|
|
|
|
* cache lines hot.
|
|
|
|
*/
|
2023-04-13 13:13:49 +00:00
|
|
|
static inline int process_huge_page(
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
unsigned long addr_hint, unsigned int pages_per_huge_page,
|
2023-04-13 13:13:49 +00:00
|
|
|
int (*process_subpage)(unsigned long addr, int idx, void *arg),
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
void *arg)
|
2011-01-13 23:46:47 +00:00
|
|
|
{
|
2023-04-13 13:13:49 +00:00
|
|
|
int i, n, base, l, ret;
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
unsigned long addr = addr_hint &
|
|
|
|
~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
|
2011-01-13 23:46:47 +00:00
|
|
|
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
/* Process target subpage last to keep its cache lines hot */
|
2011-01-13 23:46:47 +00:00
|
|
|
might_sleep();
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
n = (addr_hint - addr) / PAGE_SIZE;
|
|
|
|
if (2 * n <= pages_per_huge_page) {
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
/* If target subpage in first half of huge page */
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
base = 0;
|
|
|
|
l = n;
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
/* Process subpages at the end of huge page */
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
for (i = pages_per_huge_page - 1; i >= 2 * n; i--) {
|
|
|
|
cond_resched();
|
2023-04-13 13:13:49 +00:00
|
|
|
ret = process_subpage(addr + i * PAGE_SIZE, i, arg);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
}
|
|
|
|
} else {
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
/* If target subpage in second half of huge page */
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
base = pages_per_huge_page - 2 * (pages_per_huge_page - n);
|
|
|
|
l = pages_per_huge_page - n;
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
/* Process subpages at the begin of huge page */
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
for (i = 0; i < base; i++) {
|
|
|
|
cond_resched();
|
2023-04-13 13:13:49 +00:00
|
|
|
ret = process_subpage(addr + i * PAGE_SIZE, i, arg);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
/*
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
* Process remaining subpages in left-right-left-right pattern
|
|
|
|
* towards the target subpage
|
mm: hugetlb: clear target sub-page last when clearing huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
clearing huge page on x86_64 platform, the cache footprint is 2M. But
on a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M
LLC (last level cache). That is, in average, there are 2.5M LLC for
each core and 1.25M LLC for each thread.
If the cache pressure is heavy when clearing the huge page, and we clear
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing clearing the
end of the huge page. And it is possible for the application to access
the begin of the huge page after clearing the huge page.
To help the above situation, in this patch, when we clear a huge page,
the order to clear sub-pages is changed. In quite some situation, we
can get the address that the application will access after we clear the
huge page, for example, in a page fault handler. Instead of clearing
the huge page from begin to end, we will clear the sub-pages farthest
from the the sub-page to access firstly, and clear the sub-page to
access last. This will make the sub-page to access most cache-hot and
sub-pages around it more cache-hot too. If we cannot know the address
the application will access, the begin of the huge page is assumed to be
the the address the application will access.
With this patch, the throughput increases ~28.3% in vm-scalability
anon-w-seq test case with 72 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case creates 72 processes, each
process mmap a big anonymous memory area and writes to it from the begin
to the end. For each process, other processes could be seen as other
workload which generates heavy cache pressure. At the same time, the
cache miss rate reduced from ~33.4% to ~31.7%, the IPC (instruction per
cycle) increased from 0.56 to 0.74, and the time spent in user space is
reduced ~7.9%
Christopher Lameter suggests to clear bytes inside a sub-page from end
to begin too. But tests show no visible performance difference in the
tests. May because the size of page is small compared with the cache
size.
Thanks Andi Kleen to propose to use address to access to determine the
order of sub-pages to clear.
The hugetlbfs access address could be improved, will do that in another
patch.
[ying.huang@intel.com: improve readability of clear_huge_page()]
Link: http://lkml.kernel.org/r/20170830051842.1397-1-ying.huang@intel.com
Link: http://lkml.kernel.org/r/20170815014618.15842-1-ying.huang@intel.com
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Nadia Yvette Chambers <nyc@holomorphy.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-06 23:25:04 +00:00
|
|
|
*/
|
|
|
|
for (i = 0; i < l; i++) {
|
|
|
|
int left_idx = base + i;
|
|
|
|
int right_idx = base + 2 * l - 1 - i;
|
|
|
|
|
|
|
|
cond_resched();
|
2023-04-13 13:13:49 +00:00
|
|
|
ret = process_subpage(addr + left_idx * PAGE_SIZE, left_idx, arg);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2011-01-13 23:46:47 +00:00
|
|
|
cond_resched();
|
2023-04-13 13:13:49 +00:00
|
|
|
ret = process_subpage(addr + right_idx * PAGE_SIZE, right_idx, arg);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2011-01-13 23:46:47 +00:00
|
|
|
}
|
2023-04-13 13:13:49 +00:00
|
|
|
return 0;
|
2011-01-13 23:46:47 +00:00
|
|
|
}
|
|
|
|
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
static void clear_gigantic_page(struct page *page,
|
|
|
|
unsigned long addr,
|
|
|
|
unsigned int pages_per_huge_page)
|
|
|
|
{
|
|
|
|
int i;
|
2022-09-09 07:31:09 +00:00
|
|
|
struct page *p;
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
|
|
|
|
might_sleep();
|
2022-09-09 07:31:09 +00:00
|
|
|
for (i = 0; i < pages_per_huge_page; i++) {
|
|
|
|
p = nth_page(page, i);
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
cond_resched();
|
|
|
|
clear_user_highpage(p, addr + i * PAGE_SIZE);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2023-04-13 13:13:49 +00:00
|
|
|
static int clear_subpage(unsigned long addr, int idx, void *arg)
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
{
|
|
|
|
struct page *page = arg;
|
|
|
|
|
|
|
|
clear_user_highpage(page + idx, addr);
|
2023-04-13 13:13:49 +00:00
|
|
|
return 0;
|
mm, clear_huge_page: move order algorithm into a separate function
Patch series "mm, huge page: Copy target sub-page last when copy huge
page", v2.
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patchset.
The patchset is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patchset, we have tested it with vm-scalability run on
transparent huge page.
With this patchset, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
This patch (of 4):
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. This optimization could
be applied to copying huge page too with the same order algorithm. To
avoid code duplication and reduce maintenance overhead, in this patch,
the order algorithm is moved out of clear_huge_page() into a separate
function: process_huge_page(). So that we can use it for copying huge
page too.
This will change the direct calls to clear_user_highpage() into the
indirect calls. But with the proper inline support of the compilers,
the indirect call will be optimized to be the direct call. Our tests
show no performance change with the patch.
This patch is a code cleanup without functionality change.
Link: http://lkml.kernel.org/r/20180524005851.4079-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:46 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void clear_huge_page(struct page *page,
|
|
|
|
unsigned long addr_hint, unsigned int pages_per_huge_page)
|
|
|
|
{
|
|
|
|
unsigned long addr = addr_hint &
|
|
|
|
~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
|
|
|
|
|
|
|
|
if (unlikely(pages_per_huge_page > MAX_ORDER_NR_PAGES)) {
|
|
|
|
clear_gigantic_page(page, addr, pages_per_huge_page);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
process_huge_page(addr_hint, pages_per_huge_page, clear_subpage, page);
|
|
|
|
}
|
|
|
|
|
2023-04-13 13:13:49 +00:00
|
|
|
static int copy_user_gigantic_page(struct folio *dst, struct folio *src,
|
2023-04-10 13:39:31 +00:00
|
|
|
unsigned long addr,
|
|
|
|
struct vm_area_struct *vma,
|
|
|
|
unsigned int pages_per_huge_page)
|
2011-01-13 23:46:47 +00:00
|
|
|
{
|
|
|
|
int i;
|
2023-04-10 13:39:31 +00:00
|
|
|
struct page *dst_page;
|
|
|
|
struct page *src_page;
|
2011-01-13 23:46:47 +00:00
|
|
|
|
2022-09-09 07:31:09 +00:00
|
|
|
for (i = 0; i < pages_per_huge_page; i++) {
|
2023-04-10 13:39:31 +00:00
|
|
|
dst_page = folio_page(dst, i);
|
|
|
|
src_page = folio_page(src, i);
|
2022-09-09 07:31:09 +00:00
|
|
|
|
2011-01-13 23:46:47 +00:00
|
|
|
cond_resched();
|
2023-04-13 13:13:49 +00:00
|
|
|
if (copy_mc_user_highpage(dst_page, src_page,
|
|
|
|
addr + i*PAGE_SIZE, vma)) {
|
|
|
|
memory_failure_queue(page_to_pfn(src_page), 0);
|
|
|
|
return -EHWPOISON;
|
|
|
|
}
|
2011-01-13 23:46:47 +00:00
|
|
|
}
|
2023-04-13 13:13:49 +00:00
|
|
|
return 0;
|
2011-01-13 23:46:47 +00:00
|
|
|
}
|
|
|
|
|
mm, huge page: copy target sub-page last when copy huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patch. Because we have put the order algorithm into a separate
function, the implementation is quite simple.
The patch is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patch, we tested it with vm-scalability run on
transparent huge page.
With this patch, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
Link: http://lkml.kernel.org/r/20180524005851.4079-3-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:49 +00:00
|
|
|
struct copy_subpage_arg {
|
|
|
|
struct page *dst;
|
|
|
|
struct page *src;
|
|
|
|
struct vm_area_struct *vma;
|
|
|
|
};
|
|
|
|
|
2023-04-13 13:13:49 +00:00
|
|
|
static int copy_subpage(unsigned long addr, int idx, void *arg)
|
mm, huge page: copy target sub-page last when copy huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patch. Because we have put the order algorithm into a separate
function, the implementation is quite simple.
The patch is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patch, we tested it with vm-scalability run on
transparent huge page.
With this patch, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
Link: http://lkml.kernel.org/r/20180524005851.4079-3-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:49 +00:00
|
|
|
{
|
|
|
|
struct copy_subpage_arg *copy_arg = arg;
|
|
|
|
|
2023-04-13 13:13:49 +00:00
|
|
|
if (copy_mc_user_highpage(copy_arg->dst + idx, copy_arg->src + idx,
|
|
|
|
addr, copy_arg->vma)) {
|
|
|
|
memory_failure_queue(page_to_pfn(copy_arg->src + idx), 0);
|
|
|
|
return -EHWPOISON;
|
|
|
|
}
|
|
|
|
return 0;
|
mm, huge page: copy target sub-page last when copy huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patch. Because we have put the order algorithm into a separate
function, the implementation is quite simple.
The patch is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patch, we tested it with vm-scalability run on
transparent huge page.
With this patch, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
Link: http://lkml.kernel.org/r/20180524005851.4079-3-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:49 +00:00
|
|
|
}
|
|
|
|
|
2023-04-13 13:13:49 +00:00
|
|
|
int copy_user_large_folio(struct folio *dst, struct folio *src,
|
|
|
|
unsigned long addr_hint, struct vm_area_struct *vma)
|
2011-01-13 23:46:47 +00:00
|
|
|
{
|
2023-04-10 13:39:31 +00:00
|
|
|
unsigned int pages_per_huge_page = folio_nr_pages(dst);
|
mm, huge page: copy target sub-page last when copy huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patch. Because we have put the order algorithm into a separate
function, the implementation is quite simple.
The patch is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patch, we tested it with vm-scalability run on
transparent huge page.
With this patch, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
Link: http://lkml.kernel.org/r/20180524005851.4079-3-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:49 +00:00
|
|
|
unsigned long addr = addr_hint &
|
|
|
|
~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
|
|
|
|
struct copy_subpage_arg arg = {
|
2023-04-10 13:39:31 +00:00
|
|
|
.dst = &dst->page,
|
|
|
|
.src = &src->page,
|
mm, huge page: copy target sub-page last when copy huge page
Huge page helps to reduce TLB miss rate, but it has higher cache
footprint, sometimes this may cause some issue. For example, when
copying huge page on x86_64 platform, the cache footprint is 4M. But on
a Xeon E5 v3 2699 CPU, there are 18 cores, 36 threads, and only 45M LLC
(last level cache). That is, in average, there are 2.5M LLC for each
core and 1.25M LLC for each thread.
If the cache contention is heavy when copying the huge page, and we copy
the huge page from the begin to the end, it is possible that the begin
of huge page is evicted from the cache after we finishing copying the
end of the huge page. And it is possible for the application to access
the begin of the huge page after copying the huge page.
In c79b57e462b5d ("mm: hugetlb: clear target sub-page last when clearing
huge page"), to keep the cache lines of the target subpage hot, the
order to clear the subpages in the huge page in clear_huge_page() is
changed to clearing the subpage which is furthest from the target
subpage firstly, and the target subpage last. The similar order
changing helps huge page copying too. That is implemented in this
patch. Because we have put the order algorithm into a separate
function, the implementation is quite simple.
The patch is a generic optimization which should benefit quite some
workloads, not for a specific use case. To demonstrate the performance
benefit of the patch, we tested it with vm-scalability run on
transparent huge page.
With this patch, the throughput increases ~16.6% in vm-scalability
anon-cow-seq test case with 36 processes on a 2 socket Xeon E5 v3 2699
system (36 cores, 72 threads). The test case set
/sys/kernel/mm/transparent_hugepage/enabled to be always, mmap() a big
anonymous memory area and populate it, then forked 36 child processes,
each writes to the anonymous memory area from the begin to the end, so
cause copy on write. For each child process, other child processes
could be seen as other workloads which generate heavy cache pressure.
At the same time, the IPC (instruction per cycle) increased from 0.63 to
0.78, and the time spent in user space is reduced ~7.2%.
Link: http://lkml.kernel.org/r/20180524005851.4079-3-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shaohua Li <shli@fb.com>
Cc: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-08-17 22:45:49 +00:00
|
|
|
.vma = vma,
|
|
|
|
};
|
2011-01-13 23:46:47 +00:00
|
|
|
|
2023-04-13 13:13:49 +00:00
|
|
|
if (unlikely(pages_per_huge_page > MAX_ORDER_NR_PAGES))
|
|
|
|
return copy_user_gigantic_page(dst, src, addr, vma,
|
|
|
|
pages_per_huge_page);
|
2011-01-13 23:46:47 +00:00
|
|
|
|
2023-04-13 13:13:49 +00:00
|
|
|
return process_huge_page(addr_hint, pages_per_huge_page, copy_subpage, &arg);
|
2011-01-13 23:46:47 +00:00
|
|
|
}
|
2017-02-22 23:42:49 +00:00
|
|
|
|
2023-04-10 13:39:29 +00:00
|
|
|
long copy_folio_from_user(struct folio *dst_folio,
|
|
|
|
const void __user *usr_src,
|
|
|
|
bool allow_pagefault)
|
2017-02-22 23:42:49 +00:00
|
|
|
{
|
2023-04-10 13:39:29 +00:00
|
|
|
void *kaddr;
|
2017-02-22 23:42:49 +00:00
|
|
|
unsigned long i, rc = 0;
|
2023-04-10 13:39:29 +00:00
|
|
|
unsigned int nr_pages = folio_nr_pages(dst_folio);
|
|
|
|
unsigned long ret_val = nr_pages * PAGE_SIZE;
|
2022-09-09 07:31:09 +00:00
|
|
|
struct page *subpage;
|
2017-02-22 23:42:49 +00:00
|
|
|
|
2023-04-10 13:39:29 +00:00
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
subpage = folio_page(dst_folio, i);
|
|
|
|
kaddr = kmap_local_page(subpage);
|
2023-04-10 13:39:28 +00:00
|
|
|
if (!allow_pagefault)
|
|
|
|
pagefault_disable();
|
2023-04-10 13:39:29 +00:00
|
|
|
rc = copy_from_user(kaddr, usr_src + i * PAGE_SIZE, PAGE_SIZE);
|
2023-04-10 13:39:28 +00:00
|
|
|
if (!allow_pagefault)
|
|
|
|
pagefault_enable();
|
2023-04-10 13:39:29 +00:00
|
|
|
kunmap_local(kaddr);
|
2017-02-22 23:42:49 +00:00
|
|
|
|
|
|
|
ret_val -= (PAGE_SIZE - rc);
|
|
|
|
if (rc)
|
|
|
|
break;
|
|
|
|
|
2022-03-22 21:41:59 +00:00
|
|
|
flush_dcache_page(subpage);
|
|
|
|
|
2017-02-22 23:42:49 +00:00
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
return ret_val;
|
|
|
|
}
|
2011-01-13 23:46:47 +00:00
|
|
|
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
|
2013-11-14 22:31:51 +00:00
|
|
|
|
2013-12-20 22:28:05 +00:00
|
|
|
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
|
2014-01-21 23:49:07 +00:00
|
|
|
|
|
|
|
static struct kmem_cache *page_ptl_cachep;
|
|
|
|
|
|
|
|
void __init ptlock_cache_init(void)
|
|
|
|
{
|
|
|
|
page_ptl_cachep = kmem_cache_create("page->ptl", sizeof(spinlock_t), 0,
|
|
|
|
SLAB_PANIC, NULL);
|
|
|
|
}
|
|
|
|
|
2013-11-14 22:31:52 +00:00
|
|
|
bool ptlock_alloc(struct page *page)
|
2013-11-14 22:31:51 +00:00
|
|
|
{
|
|
|
|
spinlock_t *ptl;
|
|
|
|
|
2014-01-21 23:49:07 +00:00
|
|
|
ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
|
2013-11-14 22:31:51 +00:00
|
|
|
if (!ptl)
|
|
|
|
return false;
|
2013-11-14 22:31:52 +00:00
|
|
|
page->ptl = ptl;
|
2013-11-14 22:31:51 +00:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2013-11-14 22:31:52 +00:00
|
|
|
void ptlock_free(struct page *page)
|
2013-11-14 22:31:51 +00:00
|
|
|
{
|
2014-01-21 23:49:07 +00:00
|
|
|
kmem_cache_free(page_ptl_cachep, page->ptl);
|
2013-11-14 22:31:51 +00:00
|
|
|
}
|
|
|
|
#endif
|