by moving cond_resched_rcu() to rcupdate_wait.h, we can kill another big
sched.h dependency.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Pablo Neira Ayuso says:
====================
Netfilter/IPVS updates for net-next
1) Incorrect error check in nft_expr_inner_parse(), from Dan Carpenter.
2) Add DATA_SENT state to SCTP connection tracking helper, from
Sriram Yagnaraman.
3) Consolidate nf_confirm for ipv4 and ipv6, from Florian Westphal.
4) Add bitmask support for ipset, from Vishwanath Pai.
5) Handle icmpv6 redirects as RELATED, from Florian Westphal.
6) Add WARN_ON_ONCE() to impossible case in flowtable datapath,
from Li Qiong.
7) A large batch of IPVS updates to replace timer-based estimators by
kthreads to scale up wrt. CPUs and workload (millions of estimators).
Julian Anastasov says:
This patchset implements stats estimation in kthread context.
It replaces the code that runs on single CPU in timer context every 2
seconds and causing latency splats as shown in reports [1], [2], [3].
The solution targets setups with thousands of IPVS services,
destinations and multi-CPU boxes.
Spread the estimation on multiple (configured) CPUs and multiple
time slots (timer ticks) by using multiple chains organized under RCU
rules. When stats are not needed, it is recommended to use
run_estimation=0 as already implemented before this change.
RCU Locking:
- As stats are now RCU-locked, tot_stats, svc and dest which
hold estimator structures are now always freed from RCU
callback. This ensures RCU grace period after the
ip_vs_stop_estimator() call.
Kthread data:
- every kthread works over its own data structure and all
such structures are attached to array. For now we limit
kthreads depending on the number of CPUs.
- even while there can be a kthread structure, its task
may not be running, eg. before first service is added or
while the sysctl var is set to an empty cpulist or
when run_estimation is set to 0 to disable the estimation.
- the allocated kthread context may grow from 1 to 50
allocated structures for timer ticks which saves memory for
setups with small number of estimators
- a task and its structure may be released if all
estimators are unlinked from its chains, leaving the
slot in the array empty
- every kthread data structure allows limited number
of estimators. Kthread 0 is also used to initially
calculate the max number of estimators to allow in every
chain considering a sub-100 microsecond cond_resched
rate. This number can be from 1 to hundreds.
- kthread 0 has an additional job of optimizing the
adding of estimators: they are first added in
temp list (est_temp_list) and later kthread 0
distributes them to other kthreads. The optimization
is based on the fact that newly added estimator
should be estimated after 2 seconds, so we have the
time to offload the adding to chain from controlling
process to kthread 0.
- to add new estimators we use the last added kthread
context (est_add_ktid). The new estimators are linked to
the chains just before the estimated one, based on add_row.
This ensures their estimation will start after 2 seconds.
If estimators are added in bursts, common case if all
services and dests are initially configured, we may
spread the estimators to more chains and as result,
reducing the initial delay below 2 seconds.
Many thanks to Jiri Wiesner for his valuable comments
and for spending a lot of time reviewing and testing
the changes on different platforms with 48-256 CPUs and
1-8 NUMA nodes under different cpufreq governors.
The new IPVS estimators do not use workqueue infrastructure
because:
- The estimation can take long time when using multiple IPVS rules (eg.
millions estimator structures) and especially when box has multiple
CPUs due to the for_each_possible_cpu usage that expects packets from
any CPU. With est_nice sysctl we have more control how to prioritize the
estimation kthreads compared to other processes/kthreads that have
latency requirements (such as servers). As a benefit, we can see these
kthreads in top and decide if we will need some further control to limit
their CPU usage (max number of structure to estimate per kthread).
- with kthreads we run code that is read-mostly, no write/lock
operations to process the estimators in 2-second intervals.
- work items are one-shot: as estimators are processed every
2 seconds, they need to be re-added every time. This again
loads the timers (add_timer) if we use delayed works, as there are
no kthreads to do the timings.
[1] Report from Yunhong Jiang:
https://lore.kernel.org/netdev/D25792C1-1B89-45DE-9F10-EC350DC04ADC@gmail.com/
[2] https://marc.info/?l=linux-virtual-server&m=159679809118027&w=2
[3] Report from Dust:
https://archive.linuxvirtualserver.org/html/lvs-devel/2020-12/msg00000.html
* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next:
ipvs: run_estimation should control the kthread tasks
ipvs: add est_cpulist and est_nice sysctl vars
ipvs: use kthreads for stats estimation
ipvs: use u64_stats_t for the per-cpu counters
ipvs: use common functions for stats allocation
ipvs: add rcu protection to stats
netfilter: flowtable: add a 'default' case to flowtable datapath
netfilter: conntrack: set icmpv6 redirects as RELATED
netfilter: ipset: Add support for new bitmask parameter
netfilter: conntrack: merge ipv4+ipv6 confirm functions
netfilter: conntrack: add sctp DATA_SENT state
netfilter: nft_inner: fix IS_ERR() vs NULL check
====================
Link: https://lore.kernel.org/r/20221211101204.1751-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a new parameter to complement the existing 'netmask' option. The
main difference between netmask and bitmask is that bitmask takes any
arbitrary ip address as input, it does not have to be a valid netmask.
The name of the new parameter is 'bitmask'. This lets us mask out
arbitrary bits in the ip address, for example:
ipset create set1 hash:ip bitmask 255.128.255.0
ipset create set2 hash:ip,port family inet6 bitmask ffff::ff80
Signed-off-by: Vishwanath Pai <vpai@akamai.com>
Signed-off-by: Joshua Hunt <johunt@akamai.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The commit 510841da1f ("netfilter: ipset: enforce documented limit to
prevent allocating huge memory") was too strict and prevented to add up to
64 clashing elements to a hash:net,iface type of set. This patch fixes the
issue and now the type behaves as documented.
Fixes: 510841da1f ("netfilter: ipset: enforce documented limit to prevent allocating huge memory")
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Daniel Xu reported that the hash:net,iface type of the ipset subsystem does
not limit adding the same network with different interfaces to a set, which
can lead to huge memory usage or allocation failure.
The quick reproducer is
$ ipset create ACL.IN.ALL_PERMIT hash:net,iface hashsize 1048576 timeout 0
$ for i in $(seq 0 100); do /sbin/ipset add ACL.IN.ALL_PERMIT 0.0.0.0/0,kaf_$i timeout 0 -exist; done
The backtrace when vmalloc fails:
[Tue Oct 25 00:13:08 2022] ipset: vmalloc error: size 1073741848, exceeds total pages
<...>
[Tue Oct 25 00:13:08 2022] Call Trace:
[Tue Oct 25 00:13:08 2022] <TASK>
[Tue Oct 25 00:13:08 2022] dump_stack_lvl+0x48/0x60
[Tue Oct 25 00:13:08 2022] warn_alloc+0x155/0x180
[Tue Oct 25 00:13:08 2022] __vmalloc_node_range+0x72a/0x760
[Tue Oct 25 00:13:08 2022] ? hash_netiface4_add+0x7c0/0xb20
[Tue Oct 25 00:13:08 2022] ? __kmalloc_large_node+0x4a/0x90
[Tue Oct 25 00:13:08 2022] kvmalloc_node+0xa6/0xd0
[Tue Oct 25 00:13:08 2022] ? hash_netiface4_resize+0x99/0x710
<...>
The fix is to enforce the limit documented in the ipset(8) manpage:
> The internal restriction of the hash:net,iface set type is that the same
> network prefix cannot be stored with more than 64 different interfaces
> in a single set.
Fixes: ccf0a4b7fc ("netfilter: ipset: Add bucketsize parameter to all hash types")
Reported-by: Daniel Xu <dxu@dxuuu.xyz>
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
htable_bits() can call jhash_size(32) and trigger shift-out-of-bounds
UBSAN: shift-out-of-bounds in net/netfilter/ipset/ip_set_hash_gen.h:151:6
shift exponent 32 is too large for 32-bit type 'unsigned int'
CPU: 0 PID: 8498 Comm: syz-executor519
Not tainted 5.10.0-rc7-next-20201208-syzkaller #0
Call Trace:
__dump_stack lib/dump_stack.c:79 [inline]
dump_stack+0x107/0x163 lib/dump_stack.c:120
ubsan_epilogue+0xb/0x5a lib/ubsan.c:148
__ubsan_handle_shift_out_of_bounds.cold+0xb1/0x181 lib/ubsan.c:395
htable_bits net/netfilter/ipset/ip_set_hash_gen.h:151 [inline]
hash_mac_create.cold+0x58/0x9b net/netfilter/ipset/ip_set_hash_gen.h:1524
ip_set_create+0x610/0x1380 net/netfilter/ipset/ip_set_core.c:1115
nfnetlink_rcv_msg+0xecc/0x1180 net/netfilter/nfnetlink.c:252
netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2494
nfnetlink_rcv+0x1ac/0x420 net/netfilter/nfnetlink.c:600
netlink_unicast_kernel net/netlink/af_netlink.c:1304 [inline]
netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1330
netlink_sendmsg+0x907/0xe40 net/netlink/af_netlink.c:1919
sock_sendmsg_nosec net/socket.c:652 [inline]
sock_sendmsg+0xcf/0x120 net/socket.c:672
____sys_sendmsg+0x6e8/0x810 net/socket.c:2345
___sys_sendmsg+0xf3/0x170 net/socket.c:2399
__sys_sendmsg+0xe5/0x1b0 net/socket.c:2432
do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
entry_SYSCALL_64_after_hwframe+0x44/0xa9
This patch replaces htable_bits() by simple fls(hashsize - 1) call:
it alone returns valid nbits both for round and non-round hashsizes.
It is normal to set any nbits here because it is validated inside
following htable_size() call which returns 0 for nbits>31.
Fixes: 1feab10d7e6d("netfilter: ipset: Unified hash type generation")
Reported-by: syzbot+d66bfadebca46cf61a2b@syzkaller.appspotmail.com
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Acked-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
currently mtype_resize() can cause oops
t = ip_set_alloc(htable_size(htable_bits));
if (!t) {
ret = -ENOMEM;
goto out;
}
t->hregion = ip_set_alloc(ahash_sizeof_regions(htable_bits));
Increased htable_bits can force htable_size() to return 0.
In own turn ip_set_alloc(0) returns not 0 but ZERO_SIZE_PTR,
so follwoing access to t->hregion should trigger an OOPS.
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Acked-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
It makes possible to reproduce exactly the same set after a save/restore.
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The parameter defines the upper limit in any hash bucket at adding new entries
from userspace - if the limit would be exceeded, ipset doubles the hash size
and rehashes. It means the set may consume more memory but gives faster
evaluation at matching in the set.
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
Lastly, fix checkpatch.pl warning
WARNING: __aligned(size) is preferred over __attribute__((aligned(size)))
in net/bridge/netfilter/ebtables.c
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
When the forceadd option is enabled, the hash:* types should find and replace
the first entry in the bucket with the new one if there are no reuseable
(deleted or timed out) entries. However, the position index was just not set
to zero and remained the invalid -1 if there were no reuseable entries.
Reported-by: syzbot+6a86565c74ebe30aea18@syzkaller.appspotmail.com
Fixes: 23c42a403a ("netfilter: ipset: Introduction of new commands and protocol version 7")
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
In the case of huge hash:* types of sets, due to the single spinlock of
a set the processing of the whole set under spinlock protection could take
too long.
There were four places where the whole hash table of the set was processed
from bucket to bucket under holding the spinlock:
- During resizing a set, the original set was locked to exclude kernel side
add/del element operations (userspace add/del is excluded by the
nfnetlink mutex). The original set is actually just read during the
resize, so the spinlocking is replaced with rcu locking of regions.
However, thus there can be parallel kernel side add/del of entries.
In order not to loose those operations a backlog is added and replayed
after the successful resize.
- Garbage collection of timed out entries was also protected by the spinlock.
In order not to lock too long, region locking is introduced and a single
region is processed in one gc go. Also, the simple timer based gc running
is replaced with a workqueue based solution. The internal book-keeping
(number of elements, size of extensions) is moved to region level due to
the region locking.
- Adding elements: when the max number of the elements is reached, the gc
was called to evict the timed out entries. The new approach is that the gc
is called just for the matching region, assuming that if the region
(proportionally) seems to be full, then the whole set does. We could scan
the other regions to check every entry under rcu locking, but for huge
sets it'd mean a slowdown at adding elements.
- Listing the set header data: when the set was defined with timeout
support, the garbage collector was called to clean up timed out entries
to get the correct element numbers and set size values. Now the set is
scanned to check non-timed out entries, without actually calling the gc
for the whole set.
Thanks to Florian Westphal for helping me to solve the SOFTIRQ-safe ->
SOFTIRQ-unsafe lock order issues during working on the patch.
Reported-by: syzbot+4b0e9d4ff3cf117837e5@syzkaller.appspotmail.com
Reported-by: syzbot+c27b8d5010f45c666ed1@syzkaller.appspotmail.com
Reported-by: syzbot+68a806795ac89df3aa1c@syzkaller.appspotmail.com
Fixes: 23c42a403a ("netfilter: ipset: Introduction of new commands and protocol version 7")
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
The inline function-specifier should not be used for static functions
defined in .c files since it bloats the kernel. Instead leave the
compiler to decide which functions to inline.
While a couple of the files affected (ip_set_*_gen.h) are technically
headers, they contain templates for generating the common parts of
particular set-types and so we treat them like .c files.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
Acked-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
linux/netfilter/ipset/ip_set.h included four other header files:
include/linux/netfilter/ipset/ip_set_comment.h
include/linux/netfilter/ipset/ip_set_counter.h
include/linux/netfilter/ipset/ip_set_skbinfo.h
include/linux/netfilter/ipset/ip_set_timeout.h
Of these the first three were not included anywhere else. The last,
ip_set_timeout.h, was included in a couple of other places, but defined
inline functions which call other inline functions defined in ip_set.h,
so ip_set.h had to be included before it.
Inlined all four into ip_set.h, and updated the other files that
included ip_set_timeout.h.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
Acked-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Resolve conflict between d2912cb15b ("treewide: Replace GPLv2
boilerplate/reference with SPDX - rule 500") removing the GPL disclaimer
and fe03d47456 ("Update my email address") which updates Jozsef
Kadlecsik's email.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Based on 2 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license version 2 as
published by the free software foundation
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license version 2 as
published by the free software foundation #
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-only
has been chosen to replace the boilerplate/reference in 4122 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Enrico Weigelt <info@metux.net>
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: Allison Randal <allison@lohutok.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190604081206.933168790@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It's better to use my kadlec@netfilter.org email address in
the source code. I might not be able to use
kadlec@blackhole.kfki.hu in the future.
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
If a fresh array block is allocated during resize, the current in-memory
set size should be increased by the size of the block, not replaced by it.
Before the fix, adding entries to a hash set type, leading to a table
resize, caused an inconsistent memory size to be reported. This becomes
more obvious when swapping sets with similar sizes:
# cat hash_ip_size.sh
#!/bin/sh
FAIL_RETRIES=10
tries=0
while [ ${tries} -lt ${FAIL_RETRIES} ]; do
ipset create t1 hash:ip
for i in `seq 1 4345`; do
ipset add t1 1.2.$((i / 255)).$((i % 255))
done
t1_init="$(ipset list t1|sed -n 's/Size in memory: \(.*\)/\1/p')"
ipset create t2 hash:ip
for i in `seq 1 4360`; do
ipset add t2 1.2.$((i / 255)).$((i % 255))
done
t2_init="$(ipset list t2|sed -n 's/Size in memory: \(.*\)/\1/p')"
ipset swap t1 t2
t1_swap="$(ipset list t1|sed -n 's/Size in memory: \(.*\)/\1/p')"
t2_swap="$(ipset list t2|sed -n 's/Size in memory: \(.*\)/\1/p')"
ipset destroy t1
ipset destroy t2
tries=$((tries + 1))
if [ ${t1_init} -lt 10000 ] || [ ${t2_init} -lt 10000 ]; then
echo "FAIL after ${tries} tries:"
echo "T1 size ${t1_init}, after swap ${t1_swap}"
echo "T2 size ${t2_init}, after swap ${t2_swap}"
exit 1
fi
done
echo "PASS"
# echo -n 'func hash_ip4_resize +p' > /sys/kernel/debug/dynamic_debug/control
# ./hash_ip_size.sh
[ 2035.018673] attempt to resize set t1 from 10 to 11, t 00000000fe6551fa
[ 2035.078583] set t1 resized from 10 (00000000fe6551fa) to 11 (00000000172a0163)
[ 2035.080353] Table destroy by resize 00000000fe6551fa
FAIL after 4 tries:
T1 size 9064, after swap 71128
T2 size 71128, after swap 9064
Reported-by: NOYB <JunkYardMail1@Frontier.com>
Fixes: 9e41f26a50 ("netfilter: ipset: Count non-static extension memory for userspace")
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
After the previous commit, both ipset_nest_start() and ipset_nest_end() are
just aliases for nla_nest_start() and nla_nest_end() so that there is no
need to keep them.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Acked-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that call_rcu()'s callback is not invoked until after bh-disable
regions of code have completed (in addition to explicitly marked
RCU read-side critical sections), call_rcu() can be used in place
of call_rcu_bh(). Similarly, rcu_barrier() can be used in place of
rcu_barrier_bh() and synchronize_rcu() in place of synchronize_rcu_bh().
This commit therefore makes these changes.
Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
lockdep_assert_held() is better suited to checking locking requirements,
since it won't get confused when someone else holds the lock. This is
also a step towards possibly removing spin_is_locked().
Signed-off-by: Lance Roy <ldr709@gmail.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Cc: Florian Westphal <fw@strlen.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: <netfilter-devel@vger.kernel.org>
Cc: <coreteam@netfilter.org>
Cc: <netdev@vger.kernel.org>
Acked-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Userspace `ipset` command forbids family option for hash:mac type:
ipset create test hash:mac family inet4
ipset v6.30: Unknown argument: `family'
However, this check is not done in kernel itself. When someone use
external netlink applications (pyroute2 python library for example), one
can create hash:mac with invalid family and inconsistant results from
userspace (`ipset` command cannot read set content anymore).
This patch enforce the logic in kernel, and forbids insertion of
hash:mac with a family set.
Since IP_SET_PROTO_UNDEF is defined only for hash:mac, this patch has no
impact on other hash:* sets
Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
Signed-off-by: Victorien Molle <victorien.molle@wifirst.fr>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
The matching of the counters was not taken into account, fixed.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
When sets are extremely large we can get softlockup during ipset -L.
We could fix this by adding cond_resched_rcu() at the right location
during iteration, but this only works if RCU nesting depth is 1.
At this time entire variant->list() is called under under rcu_read_lock_bh.
This used to be a read_lock_bh() but as rcu doesn't really lock anything,
it does not appear to be needed, so remove it (ipset increments set
reference count before this, so a set deletion should not be possible).
Reported-by: Li Shuang <shuali@redhat.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
In preparation for unconditionally passing the struct timer_list pointer to
all timer callbacks, switch to using the new timer_setup() and from_timer()
to pass the timer pointer explicitly. This introduces a pointer back to the
struct ip_set, which is used instead of the struct timer_list .data field.
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Cc: Florian Westphal <fw@strlen.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: simran singhal <singhalsimran0@gmail.com>
Cc: Muhammad Falak R Wani <falakreyaz@gmail.com>
Cc: netfilter-devel@vger.kernel.org
Cc: coreteam@netfilter.org
Cc: netdev@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simple testcase:
$ ipset create test hash:ip timeout 5
$ ipset add test 1.2.3.4
$ ipset add test 1.2.2.2
$ sleep 5
$ ipset l
Name: test
Type: hash:ip
Revision: 5
Header: family inet hashsize 1024 maxelem 65536 timeout 5
Size in memory: 296
References: 0
Number of entries: 2
Members:
We return "Number of entries: 2" but no members are listed. That is
because mtype_list runs "ip_set_timeout_expired" and does not list the
expired entries, but set->elements is never upated (until mtype_gc
cleans it up later).
Reviewed-by: Joshua Hunt <johunt@akamai.com>
Signed-off-by: Vishwanath Pai <vpai@akamai.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Wrong index was used and therefore when shrinking a hash bucket at
deleting an entry, valid entries could be evicted as well.
Thanks to Eric Ewanco for the thorough bugreport.
Fixes netfilter bugzilla #1119
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Use setup_timer() and instead of init_timer(), being the preferred way
of setting up a timer.
Also, quoting the mod_timer() function comment:
-> mod_timer() is a more efficient way to update the expire field of an
active timer (if the timer is inactive it will be activated).
Use setup_timer() and mod_timer() to setup and arm a timer, making the
code compact and easier to read.
Signed-off-by: Muhammad Falak R Wani <falakreyaz@gmail.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
The calculation of the full allocated memory did not take
into account the size of the base hash bucket structure at some
places.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
The set full case (with net_ratelimit()-ed pr_warn()) is already
handled, simply jump there.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Before this patch struct htype created at the first source
of ip_set_hash_gen.h and it is common for both IPv4 and IPv6
set variants.
Make struct htype per ipset family and use NLEN to make
nets array fixed size to simplify struct htype allocation.
Ported from a patch proposed by Sergey Popovich <popovich_sergei@mail.ua>.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Exit as easly as possible on error and use RCU_INIT_POINTER()
as set is not seen at creation time.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Data for hashing required to be array of u32. Make sure that
element data always multiple of u32.
Ported from a patch proposed by Sergey Popovich <popovich_sergei@mail.ua>.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Hash types define HOST_MASK before inclusion of ip_set_hash_gen.h
and the only place where NLEN needed to be calculated at runtime
is *_create() method.
Ported from a patch proposed by Sergey Popovich <popovich_sergei@mail.ua>.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Remove one leve of intendation by using continue while
iterating over elements in bucket.
Ported from a patch proposed by Sergey Popovich <popovich_sergei@mail.ua>.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Remove redundant parameters nets_length and dsize, because
they can be get from other parameters.
Ported from a patch proposed by Sergey Popovich <popovich_sergei@mail.ua>.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Non-static (i.e. comment) extension was not counted into the memory
size. A new internal counter is introduced for this. In the case of
the hash types the sizes of the arrays are counted there as well so
that we can avoid to scan the whole set when just the header data
is requested.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
It is better to list the set elements for all set types, thus the
header information is uniform. Element counts are therefore added
to the bitmap and list types.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
It would be useful for userspace to query the size of an ipset hash,
however, this data is not exposed to userspace outside of counting the
number of member entries. This patch uses the attribute
IPSET_ATTR_ELEMENTS to indicate the size in the the header that is
exported to userspace. This field is then printed by the userspace
tool for hashes.
Signed-off-by: Eric B Munson <emunson@akamai.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Josh Hunt <johunt@akamai.com>
Cc: netfilter-devel@vger.kernel.org
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
This fix adds a new reference counter (ref_netlink) for the struct ip_set.
The other reference counter (ref) can be swapped out by ip_set_swap and we
need a separate counter to keep track of references for netlink events
like dump. Using the same ref counter for dump causes a race condition
which can be demonstrated by the following script:
ipset create hash_ip1 hash:ip family inet hashsize 1024 maxelem 500000 \
counters
ipset create hash_ip2 hash:ip family inet hashsize 300000 maxelem 500000 \
counters
ipset create hash_ip3 hash:ip family inet hashsize 1024 maxelem 500000 \
counters
ipset save &
ipset swap hash_ip3 hash_ip2
ipset destroy hash_ip3 /* will crash the machine */
Swap will exchange the values of ref so destroy will see ref = 0 instead of
ref = 1. With this fix in place swap will not succeed because ipset save
still has ref_netlink on the set (ip_set_swap doesn't swap ref_netlink).
Both delete and swap will error out if ref_netlink != 0 on the set.
Note: The changes to *_head functions is because previously we would
increment ref whenever we called these functions, we don't do that
anymore.
Reviewed-by: Joshua Hunt <johunt@akamai.com>
Signed-off-by: Vishwanath Pai <vpai@akamai.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Incorrect index was used when the data blob was shrinked at expiration,
which could lead to falsely expired entries and memory leak when
the comment extension was used too.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
The data extensions in ipset lacked the proper memory alignment and
thus could lead to kernel crash on several architectures. Therefore
the structures have been reorganized and alignment attributes added
where needed. The patch was tested on armv7h by Gerhard Wiesinger and
on x86_64, sparc64 by Jozsef Kadlecsik.
Reported-by: Gerhard Wiesinger <lists@wiesinger.com>
Tested-by: Gerhard Wiesinger <lists@wiesinger.com>
Tested-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Three types of data need to be protected in the case of the hash types:
a. The hash buckets: standard rcu pointer operations are used.
b. The element blobs in the hash buckets are stored in an array and
a bitmap is used for book-keeping to tell which elements in the array
are used or free.
c. Networks per cidr values and the cidr values themselves are stored
in fix sized arrays and need no protection. The values are modified
in such an order that in the worst case an element testing is repeated
once with the same cidr value.
The ipset hash approach uses arrays instead of lists and therefore is
incompatible with rhashtable.
Performance is tested by Jesper Dangaard Brouer:
Simple drop in FORWARD
~~~~~~~~~~~~~~~~~~~~~~
Dropping via simple iptables net-mask match::
iptables -t raw -N simple || iptables -t raw -F simple
iptables -t raw -I simple -s 198.18.0.0/15 -j DROP
iptables -t raw -D PREROUTING -j simple
iptables -t raw -I PREROUTING -j simple
Drop performance in "raw": 11.3Mpps
Generator: sending 12.2Mpps (tx:12264083 pps)
Drop via original ipset in RAW table
~~~~~~~~~~~~~~~~~~~~~~~~~~~
Create a set with lots of elements::
sudo ./ipset destroy test
echo "create test hash:ip hashsize 65536" > test.set
for x in `seq 0 255`; do
for y in `seq 0 255`; do
echo "add test 198.18.$x.$y" >> test.set
done
done
sudo ./ipset restore < test.set
Dropping via ipset::
iptables -t raw -F
iptables -t raw -N net198 || iptables -t raw -F net198
iptables -t raw -I net198 -m set --match-set test src -j DROP
iptables -t raw -I PREROUTING -j net198
Drop performance in "raw" with ipset: 8Mpps
Perf report numbers ipset drop in "raw"::
+ 24.65% ksoftirqd/1 [ip_set] [k] ip_set_test
- 21.42% ksoftirqd/1 [kernel.kallsyms] [k] _raw_read_lock_bh
- _raw_read_lock_bh
+ 99.88% ip_set_test
- 19.42% ksoftirqd/1 [kernel.kallsyms] [k] _raw_read_unlock_bh
- _raw_read_unlock_bh
+ 99.72% ip_set_test
+ 4.31% ksoftirqd/1 [ip_set_hash_ip] [k] hash_ip4_kadt
+ 2.27% ksoftirqd/1 [ixgbe] [k] ixgbe_fetch_rx_buffer
+ 2.18% ksoftirqd/1 [ip_tables] [k] ipt_do_table
+ 1.81% ksoftirqd/1 [ip_set_hash_ip] [k] hash_ip4_test
+ 1.61% ksoftirqd/1 [kernel.kallsyms] [k] __netif_receive_skb_core
+ 1.44% ksoftirqd/1 [kernel.kallsyms] [k] build_skb
+ 1.42% ksoftirqd/1 [kernel.kallsyms] [k] ip_rcv
+ 1.36% ksoftirqd/1 [kernel.kallsyms] [k] __local_bh_enable_ip
+ 1.16% ksoftirqd/1 [kernel.kallsyms] [k] dev_gro_receive
+ 1.09% ksoftirqd/1 [kernel.kallsyms] [k] __rcu_read_unlock
+ 0.96% ksoftirqd/1 [ixgbe] [k] ixgbe_clean_rx_irq
+ 0.95% ksoftirqd/1 [kernel.kallsyms] [k] __netdev_alloc_frag
+ 0.88% ksoftirqd/1 [kernel.kallsyms] [k] kmem_cache_alloc
+ 0.87% ksoftirqd/1 [xt_set] [k] set_match_v3
+ 0.85% ksoftirqd/1 [kernel.kallsyms] [k] inet_gro_receive
+ 0.83% ksoftirqd/1 [kernel.kallsyms] [k] nf_iterate
+ 0.76% ksoftirqd/1 [kernel.kallsyms] [k] put_compound_page
+ 0.75% ksoftirqd/1 [kernel.kallsyms] [k] __rcu_read_lock
Drop via ipset in RAW table with RCU-locking
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
With RCU locking, the RW-lock is gone.
Drop performance in "raw" with ipset with RCU-locking: 11.3Mpps
Performance-tested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
When elements added to a hash:* type of set and resizing triggered,
parallel listing could start to list the original set (before resizing)
and "continue" with listing the new set. Fix it by references and
using the original hash table for listing. Therefore the destroying of
the original hash table may happen from the resizing or listing functions.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Kent Overstreet