oom_unkillable_task() can be called from three different contexts i.e.
global OOM, memcg OOM and oom_score procfs interface. At the moment
oom_unkillable_task() does a task_in_mem_cgroup() check on the given
process. Since there is no reason to perform task_in_mem_cgroup()
check for global OOM and oom_score procfs interface, those contexts
provide NULL memcg and skips the task_in_mem_cgroup() check. However
for memcg OOM context, the oom_unkillable_task() is always called from
mem_cgroup_scan_tasks() and thus task_in_mem_cgroup() check becomes
redundant and effectively dead code. So, just remove the
task_in_mem_cgroup() check altogether.
Link: http://lkml.kernel.org/r/20190624212631.87212-2-shakeelb@google.com
Signed-off-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Acked-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Paul Jackson <pj@sgi.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The memory controller in cgroup v2 exposes memory.events file for each
memcg which shows the number of times events like low, high, max, oom
and oom_kill have happened for the whole tree rooted at that memcg.
Users can also poll or register notification to monitor the changes in
that file. Any event at any level of the tree rooted at memcg will
notify all the listeners along the path till root_mem_cgroup. There are
existing users which depend on this behavior.
However there are users which are only interested in the events
happening at a specific level of the memcg tree and not in the events in
the underlying tree rooted at that memcg. One such use-case is a
centralized resource monitor which can dynamically adjust the limits of
the jobs running on a system. The jobs can create their sub-hierarchy
for their own sub-tasks. The centralized monitor is only interested in
the events at the top level memcgs of the jobs as it can then act and
adjust the limits of the jobs. Using the current memory.events for such
centralized monitor is very inconvenient. The monitor will keep
receiving events which it is not interested and to find if the received
event is interesting, it has to read memory.event files of the next
level and compare it with the top level one. So, let's introduce
memory.events.local to the memcg which shows and notify for the events
at the memcg level.
Now, does memory.stat and memory.pressure need their local versions. IMHO
no due to the no internal process contraint of the cgroup v2. The
memory.stat file of the top level memcg of a job shows the stats and
vmevents of the whole tree. The local stats or vmevents of the top level
memcg will only change if there is a process running in that memcg but v2
does not allow that. Similarly for memory.pressure there will not be any
process in the internal nodes and thus no chance of local pressure.
Link: http://lkml.kernel.org/r/20190527174643.209172-1-shakeelb@google.com
Signed-off-by: Shakeel Butt <shakeelb@google.com>
Reviewed-by: Roman Gushchin <guro@fb.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Chris Down <chris@chrisdown.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The kernel test robot noticed a 26% will-it-scale pagefault regression
from commit 42a3003535 ("mm: memcontrol: fix recursive statistics
correctness & scalabilty"). This appears to be caused by bouncing the
additional cachelines from the new hierarchical statistics counters.
We can fix this by getting rid of the batched local counters instead.
Originally, there were *only* group-local counters, and they were fully
maintained per cpu. A reader of a stats file high up in the cgroup tree
would have to walk the entire subtree and collect each level's per-cpu
counters to get the recursive view. This was prohibitively expensive,
and so we switched to per-cpu batched updates of the local counters
during a983b5ebee ("mm: memcontrol: fix excessive complexity in
memory.stat reporting"), reducing the complexity from nr_subgroups *
nr_cpus to nr_subgroups.
With growing machines and cgroup trees, the tree walk itself became too
expensive for monitoring top-level groups, and this is when the culprit
patch added hierarchy counters on each cgroup level. When the per-cpu
batch size would be reached, both the local and the hierarchy counters
would get batch-updated from the per-cpu delta simultaneously.
This makes local and hierarchical counter reads blazingly fast, but it
unfortunately makes the write-side too cache line intense.
Since local counter reads were never a problem - we only centralized
them to accelerate the hierarchy walk - and use of the local counters
are becoming rarer due to replacement with hierarchical views (ongoing
rework in the page reclaim and workingset code), we can make those local
counters unbatched per-cpu counters again.
The scheme will then be as such:
when a memcg statistic changes, the writer will:
- update the local counter (per-cpu)
- update the batch counter (per-cpu). If the batch is full:
- spill the batch into the group's atomic_t
- spill the batch into all ancestors' atomic_ts
- empty out the batch counter (per-cpu)
when a local memcg counter is read, the reader will:
- collect the local counter from all cpus
when a hiearchy memcg counter is read, the reader will:
- read the atomic_t
We might be able to simplify this further and make the recursive
counters unbatched per-cpu counters as well (batch upward propagation,
but leave per-cpu collection to the readers), but that will require a
more in-depth analysis and testing of all the callsites. Deal with the
immediate regression for now.
Link: http://lkml.kernel.org/r/20190521151647.GB2870@cmpxchg.org
Fixes: 42a3003535 ("mm: memcontrol: fix recursive statistics correctness & scalabilty")
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reported-by: kernel test robot <rong.a.chen@intel.com>
Tested-by: kernel test robot <rong.a.chen@intel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Roman Gushchin <guro@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
memory.stat and other files already consider subtrees in their output, and
we should too in order to not present an inconsistent interface.
The current situation is fairly confusing, because people interacting with
cgroups expect hierarchical behaviour in the vein of memory.stat,
cgroup.events, and other files. For example, this causes confusion when
debugging reclaim events under low, as currently these always read "0" at
non-leaf memcg nodes, which frequently causes people to misdiagnose breach
behaviour. The same confusion applies to other counters in this file when
debugging issues.
Aggregation is done at write time instead of at read-time since these
counters aren't hot (unlike memory.stat which is per-page, so it does it
at read time), and it makes sense to bundle this with the file
notifications.
After this patch, events are propagated up the hierarchy:
[root@ktst ~]# cat /sys/fs/cgroup/system.slice/memory.events
low 0
high 0
max 0
oom 0
oom_kill 0
[root@ktst ~]# systemd-run -p MemoryMax=1 true
Running as unit: run-r251162a189fb4562b9dabfdc9b0422f5.service
[root@ktst ~]# cat /sys/fs/cgroup/system.slice/memory.events
low 0
high 0
max 7
oom 1
oom_kill 1
As this is a change in behaviour, this can be reverted to the old
behaviour by mounting with the `memory_localevents' flag set. However, we
use the new behaviour by default as there's a lack of evidence that there
are any current users of memory.events that would find this change
undesirable.
akpm: this is a behaviour change, so Cc:stable. THis is so that
forthcoming distros which use cgroup v2 are more likely to pick up the
revised behaviour.
Link: http://lkml.kernel.org/r/20190208224419.GA24772@chrisdown.name
Signed-off-by: Chris Down <chris@chrisdown.name>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Based on 3 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 as published by
the free software foundation either version 2 of the license or at
your option any later version this program is distributed in the
hope that it will be useful but without any warranty without even
the implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version [author] [kishon] [vijay] [abraham]
[i] [kishon]@[ti] [com] this program is distributed in the hope that
it will be useful but without any warranty without even the implied
warranty of merchantability or fitness for a particular purpose see
the gnu general public license for more details
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version [author] [graeme] [gregory]
[gg]@[slimlogic] [co] [uk] [author] [kishon] [vijay] [abraham] [i]
[kishon]@[ti] [com] [based] [on] [twl6030]_[usb] [c] [author] [hema]
[hk] [hemahk]@[ti] [com] this program is distributed in the hope
that it will be useful but without any warranty without even the
implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-or-later
has been chosen to replace the boilerplate/reference in 1105 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Allison Randal <allison@lohutok.net>
Reviewed-by: Richard Fontana <rfontana@redhat.com>
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190527070033.202006027@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Right now, when somebody needs to know the recursive memory statistics
and events of a cgroup subtree, they need to walk the entire subtree and
sum up the counters manually.
There are two issues with this:
1. When a cgroup gets deleted, its stats are lost. The state counters
should all be 0 at that point, of course, but the events are not.
When this happens, the event counters, which are supposed to be
monotonic, can go backwards in the parent cgroups.
2. During regular operation, we always have a certain number of lazily
freed cgroups sitting around that have been deleted, have no tasks,
but have a few cache pages remaining. These groups' statistics do not
change until we eventually hit memory pressure, but somebody
watching, say, memory.stat on an ancestor has to iterate those every
time.
This patch addresses both issues by introducing recursive counters at
each level that are propagated from the write side when stats change.
Upward propagation happens when the per-cpu caches spill over into the
local atomic counter. This is the same thing we do during charge and
uncharge, except that the latter uses atomic RMWs, which are more
expensive; stat changes happen at around the same rate. In a sparse
file test (page faults and reclaim at maximum CPU speed) with 5 cgroup
nesting levels, perf shows __mod_memcg_page state at ~1%.
Link: http://lkml.kernel.org/r/20190412151507.2769-4-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Reviewed-by: Roman Gushchin <guro@fb.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>
Patch series "mm: memcontrol: memory.stat cost & correctness".
The cgroup memory.stat file holds recursive statistics for the entire
subtree. The current implementation does this tree walk on-demand
whenever the file is read. This is giving us problems in production.
1. The cost of aggregating the statistics on-demand is high. A lot of
system service cgroups are mostly idle and their stats don't change
between reads, yet we always have to check them. There are also always
some lazily-dying cgroups sitting around that are pinned by a handful
of remaining page cache; the same applies to them.
In an application that periodically monitors memory.stat in our
fleet, we have seen the aggregation consume up to 5% CPU time.
2. When cgroups die and disappear from the cgroup tree, so do their
accumulated vm events. The result is that the event counters at
higher-level cgroups can go backwards and confuse some of our
automation, let alone people looking at the graphs over time.
To address both issues, this patch series changes the stat
implementation to spill counts upwards when the counters change.
The upward spilling is batched using the existing per-cpu cache. In a
sparse file stress test with 5 level cgroup nesting, the additional cost
of the flushing was negligible (a little under 1% of CPU at 100% CPU
utilization, compared to the 5% of reading memory.stat during regular
operation).
This patch (of 4):
memcg_page_state(), lruvec_page_state(), memcg_sum_events() are
currently returning the state of the local memcg or lruvec, not the
recursive state.
In practice there is a demand for both versions, although the callers
that want the recursive counts currently sum them up by hand.
Per default, cgroups are considered recursive entities and generally we
expect more users of the recursive counters, with the local counts being
special cases. To reflect that in the name, add a _local suffix to the
current implementations.
The following patch will re-incarnate these functions with recursive
semantics, but with an O(1) implementation.
[hannes@cmpxchg.org: fix bisection hole]
Link: http://lkml.kernel.org/r/20190417160347.GC23013@cmpxchg.org
Link: http://lkml.kernel.org/r/20190412151507.2769-2-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Reviewed-by: Roman Gushchin <guro@fb.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>
I spent literally an hour trying to work out why an earlier version of
my memory.events aggregation code doesn't work properly, only to find
out I was calling memcg->events instead of memcg->memory_events, which
is fairly confusing.
This naming seems in need of reworking, so make it harder to do the
wrong thing by using vmevents instead of events, which makes it more
clear that these are vm counters rather than memcg-specific counters.
There are also a few other inconsistent names in both the percpu and
aggregated structs, so these are all cleaned up to be more coherent and
easy to understand.
This commit contains code cleanup only: there are no logic changes.
[akpm@linux-foundation.org: fix it for preceding changes]
Link: http://lkml.kernel.org/r/20190208224319.GA23801@chrisdown.name
Signed-off-by: Chris Down <chris@chrisdown.name>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Cc: Dennis Zhou <dennis@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Since commit a983b5ebee ("mm: memcontrol: fix excessive complexity in
memory.stat reporting") memcg dirty and writeback counters are managed
as:
1) per-memcg per-cpu values in range of [-32..32]
2) per-memcg atomic counter
When a per-cpu counter cannot fit in [-32..32] it's flushed to the
atomic. Stat readers only check the atomic. Thus readers such as
balance_dirty_pages() may see a nontrivial error margin: 32 pages per
cpu.
Assuming 100 cpus:
4k x86 page_size: 13 MiB error per memcg
64k ppc page_size: 200 MiB error per memcg
Considering that dirty+writeback are used together for some decisions the
errors double.
This inaccuracy can lead to undeserved oom kills. One nasty case is
when all per-cpu counters hold positive values offsetting an atomic
negative value (i.e. per_cpu[*]=32, atomic=n_cpu*-32).
balance_dirty_pages() only consults the atomic and does not consider
throttling the next n_cpu*32 dirty pages. If the file_lru is in the
13..200 MiB range then there's absolutely no dirty throttling, which
burdens vmscan with only dirty+writeback pages thus resorting to oom
kill.
It could be argued that tiny containers are not supported, but it's more
subtle. It's the amount the space available for file lru that matters.
If a container has memory.max-200MiB of non reclaimable memory, then it
will also suffer such oom kills on a 100 cpu machine.
The following test reliably ooms without this patch. This patch avoids
oom kills.
$ cat test
mount -t cgroup2 none /dev/cgroup
cd /dev/cgroup
echo +io +memory > cgroup.subtree_control
mkdir test
cd test
echo 10M > memory.max
(echo $BASHPID > cgroup.procs && exec /memcg-writeback-stress /foo)
(echo $BASHPID > cgroup.procs && exec dd if=/dev/zero of=/foo bs=2M count=100)
$ cat memcg-writeback-stress.c
/*
* Dirty pages from all but one cpu.
* Clean pages from the non dirtying cpu.
* This is to stress per cpu counter imbalance.
* On a 100 cpu machine:
* - per memcg per cpu dirty count is 32 pages for each of 99 cpus
* - per memcg atomic is -99*32 pages
* - thus the complete dirty limit: sum of all counters 0
* - balance_dirty_pages() only sees atomic count -99*32 pages, which
* it max()s to 0.
* - So a workload can dirty -99*32 pages before balance_dirty_pages()
* cares.
*/
#define _GNU_SOURCE
#include <err.h>
#include <fcntl.h>
#include <sched.h>
#include <stdlib.h>
#include <stdio.h>
#include <sys/stat.h>
#include <sys/sysinfo.h>
#include <sys/types.h>
#include <unistd.h>
static char *buf;
static int bufSize;
static void set_affinity(int cpu)
{
cpu_set_t affinity;
CPU_ZERO(&affinity);
CPU_SET(cpu, &affinity);
if (sched_setaffinity(0, sizeof(affinity), &affinity))
err(1, "sched_setaffinity");
}
static void dirty_on(int output_fd, int cpu)
{
int i, wrote;
set_affinity(cpu);
for (i = 0; i < 32; i++) {
for (wrote = 0; wrote < bufSize; ) {
int ret = write(output_fd, buf+wrote, bufSize-wrote);
if (ret == -1)
err(1, "write");
wrote += ret;
}
}
}
int main(int argc, char **argv)
{
int cpu, flush_cpu = 1, output_fd;
const char *output;
if (argc != 2)
errx(1, "usage: output_file");
output = argv[1];
bufSize = getpagesize();
buf = malloc(getpagesize());
if (buf == NULL)
errx(1, "malloc failed");
output_fd = open(output, O_CREAT|O_RDWR);
if (output_fd == -1)
err(1, "open(%s)", output);
for (cpu = 0; cpu < get_nprocs(); cpu++) {
if (cpu != flush_cpu)
dirty_on(output_fd, cpu);
}
set_affinity(flush_cpu);
if (fsync(output_fd))
err(1, "fsync(%s)", output);
if (close(output_fd))
err(1, "close(%s)", output);
free(buf);
}
Make balance_dirty_pages() and wb_over_bg_thresh() work harder to
collect exact per memcg counters. This avoids the aforementioned oom
kills.
This does not affect the overhead of memory.stat, which still reads the
single atomic counter.
Why not use percpu_counter? memcg already handles cpus going offline, so
no need for that overhead from percpu_counter. And the percpu_counter
spinlocks are more heavyweight than is required.
It probably also makes sense to use exact dirty and writeback counters
in memcg oom reports. But that is saved for later.
Link: http://lkml.kernel.org/r/20190329174609.164344-1-gthelen@google.com
Signed-off-by: Greg Thelen <gthelen@google.com>
Reviewed-by: Roman Gushchin <guro@fb.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: <stable@vger.kernel.org> [4.16+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is the start of a series of patches similar to my earlier
DEFINE_MEMCG_MAX_OR_VAL work, but with less Macro Magic(tm).
There are a bunch of places we go from seq_file to mem_cgroup, which
currently requires manually getting the css, then getting the mem_cgroup
from the css. It's in enough places now that having mem_cgroup_from_seq
makes sense (and also makes the next patch a bit nicer).
Link: http://lkml.kernel.org/r/20190124194050.GA31341@chrisdown.name
Signed-off-by: Chris Down <chris@chrisdown.name>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The current oom report doesn't display victim's memcg context during the
global OOM situation. While this information is not strictly needed, it
can be really helpful for containerized environments to locate which
container has lost a process. Now that we have a single line for the oom
context, we can trivially add both the oom memcg (this can be either
global_oom or a specific memcg which hits its hard limits) and task_memcg
which is the victim's memcg.
Below is the single line output in the oom report after this patch.
- global oom context information:
oom-kill:constraint=<constraint>,nodemask=<nodemask>,cpuset=<cpuset>,mems_allowed=<mems_allowed>,global_oom,task_memcg=<memcg>,task=<comm>,pid=<pid>,uid=<uid>
- memcg oom context information:
oom-kill:constraint=<constraint>,nodemask=<nodemask>,cpuset=<cpuset>,mems_allowed=<mems_allowed>,oom_memcg=<memcg>,task_memcg=<memcg>,task=<comm>,pid=<pid>,uid=<uid>
[penguin-kernel@I-love.SAKURA.ne.jp: use pr_cont() in mem_cgroup_print_oom_context()]
Link: http://lkml.kernel.org/r/201812190723.wBJ7NdkN032628@www262.sakura.ne.jp
Link: http://lkml.kernel.org/r/1542799799-36184-2-git-send-email-ufo19890607@gmail.com
Signed-off-by: yuzhoujian <yuzhoujian@didichuxing.com>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: Roman Gushchin <guro@fb.com>
Cc: Yang Shi <yang.s@alibaba-inc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If CONFIG_VMAP_STACK is set, kernel stacks are allocated using
__vmalloc_node_range() with __GFP_ACCOUNT. So kernel stack pages are
charged against corresponding memory cgroups on allocation and uncharged
on releasing them.
The problem is that we do cache kernel stacks in small per-cpu caches and
do reuse them for new tasks, which can belong to different memory cgroups.
Each stack page still holds a reference to the original cgroup, so the
cgroup can't be released until the vmap area is released.
To make this happen we need more than two subsequent exits without forks
in between on the current cpu, which makes it very unlikely to happen. As
a result, I saw a significant number of dying cgroups (in theory, up to 2
* number_of_cpu + number_of_tasks), which can't be released even by
significant memory pressure.
As a cgroup structure can take a significant amount of memory (first of
all, per-cpu data like memcg statistics), it leads to a noticeable waste
of memory.
Link: http://lkml.kernel.org/r/20180827162621.30187-1-guro@fb.com
Fixes: ac496bf48d ("fork: Optimize task creation by caching two thread stacks per CPU if CONFIG_VMAP_STACK=y")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Michal Hocko <mhocko@kernel.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
For some workloads an intervention from the OOM killer can be painful.
Killing a random task can bring the workload into an inconsistent state.
Historically, there are two common solutions for this
problem:
1) enabling panic_on_oom,
2) using a userspace daemon to monitor OOMs and kill
all outstanding processes.
Both approaches have their downsides: rebooting on each OOM is an obvious
waste of capacity, and handling all in userspace is tricky and requires a
userspace agent, which will monitor all cgroups for OOMs.
In most cases an in-kernel after-OOM cleaning-up mechanism can eliminate
the necessity of enabling panic_on_oom. Also, it can simplify the cgroup
management for userspace applications.
This commit introduces a new knob for cgroup v2 memory controller:
memory.oom.group. The knob determines whether the cgroup should be
treated as an indivisible workload by the OOM killer. If set, all tasks
belonging to the cgroup or to its descendants (if the memory cgroup is not
a leaf cgroup) are killed together or not at all.
To determine which cgroup has to be killed, we do traverse the cgroup
hierarchy from the victim task's cgroup up to the OOMing cgroup (or root)
and looking for the highest-level cgroup with memory.oom.group set.
Tasks with the OOM protection (oom_score_adj set to -1000) are treated as
an exception and are never killed.
This patch doesn't change the OOM victim selection algorithm.
Link: http://lkml.kernel.org/r/20180802003201.817-4-guro@fb.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit 3812c8c8f3 ("mm: memcg: do not trap chargers with full
callstack on OOM") has changed the ENOMEM semantic of memcg charges.
Rather than invoking the oom killer from the charging context it delays
the oom killer to the page fault path (pagefault_out_of_memory). This
in turn means that many users (e.g. slab or g-u-p) will get ENOMEM when
the corresponding memcg hits the hard limit and the memcg is is OOM.
This is behavior is inconsistent with !memcg case where the oom killer
is invoked from the allocation context and the allocator keeps retrying
until it succeeds.
The difference in the behavior is user visible. mmap(MAP_POPULATE)
might result in not fully populated ranges while the mmap return code
doesn't tell that to the userspace. Random syscalls might fail with
ENOMEM etc.
The primary motivation of the different memcg oom semantic was the
deadlock avoidance. Things have changed since then, though. We have an
async oom teardown by the oom reaper now and so we do not have to rely
on the victim to tear down its memory anymore. Therefore we can return
to the original semantic as long as the memcg oom killer is not handed
over to the users space.
There is still one thing to be careful about here though. If the oom
killer is not able to make any forward progress - e.g. because there is
no eligible task to kill - then we have to bail out of the charge path
to prevent from same class of deadlocks. We have basically two options
here. Either we fail the charge with ENOMEM or force the charge and
allow overcharge. The first option has been considered more harmful
than useful because rare inconsistencies in the ENOMEM behavior is hard
to test for and error prone. Basically the same reason why the page
allocator doesn't fail allocations under such conditions. The later
might allow runaways but those should be really unlikely unless somebody
misconfigures the system. E.g. allowing to migrate tasks away from the
memcg to a different unlimited memcg with move_charge_at_immigrate
disabled.
Link: http://lkml.kernel.org/r/20180628151101.25307-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
