Linux-mm Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Zhan Xusheng" <zhanxusheng1024@gmail.com>, akpm@linux-foundation.org
Cc: hannes@cmpxchg.org, mhocko@suse.com, roman.gushchin@linux.dev,
	shakeel.butt@linux.dev, muchun.song@linux.dev,
	penguin-kernel@i-love.sakura.ne.jp, tj@kernel.org,
	mkoutny@suse.com, chris@chrisdown.name, cui.tao@linux.dev,
	linux-mm@kvack.org, zhanxusheng@xiaomi.com
Subject: Re: [PATCH v3] mm: memcontrol: don't miss processes whose leader has exited in OOM scans
Date: Sun, 22 Nov 2026 07:32:46 +0000	[thread overview]
Message-ID: <sashiko-outbox-227786@kernel.org> (raw)
In-Reply-To: <20261121073222.834282-1-zhanxusheng1024@gmail.com>

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Medium] mm: memcontrol: mem_cgroup_scan_skip() exposes left-behind
  exiting threads
--

commit 7a7da7139a4622f1e1e4394bcb56c233c1176259
Author: Zhan Xusheng <zhanxusheng@xiaomi.com>

mm: memcontrol: don't miss processes whose leader has exited in OOM scans

This commit replaces CSS_TASK_ITER_PROCS in mem_cgroup_scan_tasks() with
a scan over all tasks filtered by mem_cgroup_scan_skip(), so that
processes migrated after their thread group leader has exited remain
visible to memcg OOM scans.

> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index afbb638e7e1cb..3355341666db5 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -1426,6 +1426,44 @@ static void invalidate_reclaim_iterators(struct mem_cgroup *dead_memcg)
>  						dead_memcg);
>  }
>  
> +/*
> + * Let one task of each thread group in @memcg stand for the group: the
> + * first one that is not exiting, or, once the whole group is exiting, the
> + * first one that has not exited yet, so that an OOM victim stays visible.
> + * The leader is not enough: a group whose leader has exited can be moved
> + * without it.  Exiting threads left behind in @memcg do not count while the
> + * group has live threads elsewhere.
> + */
> +static bool mem_cgroup_scan_skip(struct mem_cgroup *memcg,
> +				 struct task_struct *task)
> +{
> +	struct task_struct *t, *dying = NULL;
> +	bool live_elsewhere = false;
> +
> +	guard(rcu)();
> +	__for_each_thread(task->signal, t) {
> +		bool here;
> +
> +		if (READ_ONCE(t->exit_state))
> +			continue;
> +		here = mem_cgroup_from_task(t) == memcg;
> +		if (!(READ_ONCE(t->flags) & PF_EXITING)) {
> +			if (here)
> +				return t != task;
> +			live_elsewhere = true;
> +		} else if (here && !dying) {
> +			dying = t;
> +		}
> +	}
> +	/*
> +	 * Once every thread has exited, the leader stands for the group until
> +	 * it is gone, as with CSS_TASK_ITER_PROCS.
> +	 */
> +	if (!dying)
> +		return live_elsewhere || !thread_group_leader(task);
> +	return live_elsewhere || dying != task;
> +}

[Severity: Medium]
Can mem_cgroup_scan_skip() expose an exiting thread left behind in the
source memcg once the migrated threads in the destination memcg begin
exiting?

When a multi-threaded process is migrated while one of its threads already
has PF_EXITING set, cgroup_migrate_add_task() skips the exiting thread and
leaves it on the source cset->tasks until do_cgroup_task_dead().  With
CSS_TASK_ITER_WITH_DEAD, css_task_iter_next() continues returning that
left-behind thread even after the last migrated thread enters do_exit().

Once the migrated threads in the destination memcg set PF_EXITING (for
example, after being OOM-killed there or calling exit_group()), they have
here == false and PF_EXITING set, so the loop above ignores them without
setting live_elsewhere.  If the left-behind thread in the source memcg
still has exit_state == 0, dying is set to it and mem_cgroup_scan_skip()
returns false; if the left-behind thread is the group leader with
exit_state != 0 (or a leader concurrently migrated out of @memcg after
css_task_iter_next() returned it), dying is NULL and the !dying fallback
also returns false without checking whether unexited threads exist
elsewhere or whether @task is still in @memcg.

An OOM scan in the source memcg then visits the migrated process through
the left-behind thread: if the process is an OOM victim in the destination
memcg, tsk_is_oom_victim() causes the source memcg's OOM scan to back off
without killing anything in the source memcg, and if a migrated thread has
not yet reached exit_mm(), find_lock_task_mm() on the left-behind thread
can still find the migrated process's mm and select it as the source
memcg's OOM victim.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261121073222.834282-1-zhanxusheng1024@gmail.com?part=1


      reply	other threads:[~2026-11-22  7:32 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-21  7:32 [PATCH v3] mm: memcontrol: don't miss processes whose leader has exited in OOM scans Zhan Xusheng
2026-11-22  7:32 ` sashiko-bot [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-227786@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=akpm@linux-foundation.org \
    --cc=chris@chrisdown.name \
    --cc=cui.tao@linux.dev \
    --cc=hannes@cmpxchg.org \
    --cc=linux-mm@kvack.org \
    --cc=mhocko@suse.com \
    --cc=mkoutny@suse.com \
    --cc=muchun.song@linux.dev \
    --cc=penguin-kernel@i-love.sakura.ne.jp \
    --cc=roman.gushchin@linux.dev \
    --cc=sashiko-reviews@lists.linux.dev \
    --cc=shakeel.butt@linux.dev \
    --cc=tj@kernel.org \
    --cc=zhanxusheng1024@gmail.com \
    --cc=zhanxusheng@xiaomi.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox