linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Oleg Nesterov <oleg@redhat.com>
To: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: mhocko@kernel.org, linux-mm@kvack.org, mhocko@suse.com,
	vdavydov@virtuozzo.com, rientjes@google.com
Subject: Re: [PATCH v2] mm, oom: don't set TIF_MEMDIE on a mm-less thread.
Date: Fri, 24 Jun 2016 00:58:09 +0200	[thread overview]
Message-ID: <20160623225809.GB17186@redhat.com> (raw)
In-Reply-To: <201606240124.FEI12978.OFQOSMJtOHFFLV@I-love.SAKURA.ne.jp>

On 06/24, Tetsuo Handa wrote:
>
> On CONFIG_MMU=n kernels, nothing
> will clear TIF_MEMDIE and the system can OOM livelock if TIF_MEMDIE was
> by error set to a mm-less thread group leader.

and btw this needs more cleanups imo. I mean, the fact we pass task_struct
to wake_oom_reaper() looks, well, strange. But this is off-topic right now.

> @@ -839,9 +839,19 @@ void oom_kill_process(struct oom_control *oc, struct task_struct *p,
>  	 * its children or threads, just set TIF_MEMDIE so it can die quickly
>  	 */
>  	if (task_will_free_mem(p)) {
> -		mark_oom_victim(p);
> -		wake_oom_reaper(p);
> -		put_task_struct(p);
> +		p = find_lock_task_mm(victim);
> +		if (!p) {

Well, this doesn't really matter... but imo

		victim = find_lock_task_mm(p);
		if (!victim) {

will look much more readable, and this way we won't depend on the
early "victim = p" initialization at the start.

> +			put_task_struct(victim);
> +			return;
> +		} else if (victim != p) {
> +			get_task_struct(p);
> +			put_task_struct(victim);
> +			victim = p;
> +		}

Tetsuo but this is horrible ;)

At least this needs a comment to explain _why_. Because this looks
"obviously unnecessary"; exit_oom_victim() does find_lock_task_mm()
too and "p" can exit right after we drop task_lock(). Not to mention
that task_will_free_mem() called find_lock_task_mm() right before
that, so this is sub-optimal in any case.

IOW, this should explain that we only need this for mark_oom_victim(),
and only if CONFIG_MMU=n. And this leads to other questions:

	- Why we can livelock in this case? This should be documented
	  too imo,

	  I have to admit I don't understand why. But yes, yes, sorry,
	  I ignored a lot of emails in this area :/

	- Why mark_oom_victim() can't check ->mm != NULL itself?

Oleg.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2016-06-23 22:57 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-06-23 15:58 [PATCH] " Tetsuo Handa
2016-06-23 16:24 ` [PATCH v2] " Tetsuo Handa
2016-06-23 22:58   ` Oleg Nesterov [this message]
2016-06-24  9:54   ` Michal Hocko
2016-06-24 10:56     ` Tetsuo Handa
2016-06-24 12:04       ` Michal Hocko
2016-06-24 16:19         ` Tetsuo Handa
2016-06-27 11:37           ` Michal Hocko
2016-06-27 13:32             ` Tetsuo Handa
2016-06-27 14:06               ` Michal Hocko

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=20160623225809.GB17186@redhat.com \
    --to=oleg@redhat.com \
    --cc=linux-mm@kvack.org \
    --cc=mhocko@kernel.org \
    --cc=mhocko@suse.com \
    --cc=penguin-kernel@I-love.SAKURA.ne.jp \
    --cc=rientjes@google.com \
    --cc=vdavydov@virtuozzo.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