linux-kernel - Re: mm, oom_reaper: How to handle race with oom_killer

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20160622120843.GE9208@dhcp22.suse.cz>
Date:	Wed, 22 Jun 2016 14:08:43 +0200
From:	Michal Hocko <mhocko@...nel.org>
To:	Tetsuo Handa <penguin-kernel@...ove.SAKURA.ne.jp>
Cc:	linux-mm@...ck.org, rientjes@...gle.com, oleg@...hat.com,
	vdavydov@...allels.com, mgorman@...hsingularity.net,
	hughd@...gle.com, riel@...hat.com, akpm@...ux-foundation.org,
	linux-kernel@...r.kernel.org
Subject: Re: mm, oom_reaper: How to handle race with oom_killer_disable() ?

On Wed 22-06-16 19:57:17, Tetsuo Handa wrote:
> Michal Hocko wrote:
[...]
> > That being said I guess the patch to try_to_freeze_tasks after
> > oom_killer_disable should be simple enough to go for now and stable
> > trees and we can come up with something less hackish later. I do not
> > like the fact that oom_killer_disable doesn't act as a full "barrier"
> > anymore.
> > 
> > What do you think?
> 
> I'm OK with calling try_to_freeze_tasks(true) again for Linux 4.6 and 4.7 kernels.

OK, I will resend the patch CC Rafael and stable.
 
> But if free memory is little such that oom_killer_disable() can not expect TIF_MEMDIE
> threads to clear TIF_MEMDIE by themselves (and therefore has to depend on the OOM
> reaper to clear TIF_MEMDIE on behalf of them after the OOM reaper reaped some memory),
> subsequent operations would be as well blocked waiting for an operation which cannot
> make any forward progress because it cannot proceed with an allocation. Then,
> oom_killer_disable() returns false after some timeout (i.e. "do not try to suspend
> when the system is almost OOM") will be a safer reaction.

Yes that is exactly what I meant by "oom_killer_disable has to give up"
alternative. pm suspend already has a notion of timeout for back off
and oom_killer_disable can use wait_even_timeout. But let's do that
separately.

-- 
Michal Hocko
SUSE Labs