linux-kernel - Re: [3.2-rc3] OOM killer doesn't kill the obvious memory hog

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: <20111204220436.GB7046@dastard>
Date:	Mon, 5 Dec 2011 09:04:36 +1100
From:	Dave Chinner <david@...morbit.com>
To:	KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
Cc:	David Rientjes <rientjes@...gle.com>, linux-kernel@...r.kernel.org,
	linux-mm@...ck.org
Subject: Re: [3.2-rc3] OOM killer doesn't kill the obvious memory hog

On Fri, Dec 02, 2011 at 02:44:41PM +0900, KAMEZAWA Hiroyuki wrote:
> On Fri, 2 Dec 2011 14:31:48 +1100
> Dave Chinner <david@...morbit.com> wrote:
> 
> > So, it's a distro bug - sshd should never be started from from udev
> > context because of this inherited oom_score_adj thing.
> > Interestingly, the ifup ssh restart script says this:
> > 
> > # We'd like to use 'reload' here, but it has some problems; see #502444.
> > if [ -x /usr/sbin/invoke-rc.d ]; then
> >         invoke-rc.d ssh restart >/dev/null 2>&1 || true
> > else
> >         /etc/init.d/ssh restart >/dev/null 2>&1 || true
> > fi
> > 
> > Bug 502444 describes the exact startup race condition that I've just
> > found. It does a ssh server restart because reload causes the sshd
> > server to fail to start if a start is currently in progress.  So,
> > rather than solving the start vs reload race condition, it got a
> > bandaid (use restart to restart sshd from the reload context) and
> > left it as a landmine.....
> > 
> 
> Thank you for chasing. 
> Hm, BTW, do you think this kind of tracepoint is useful for debugging ?
> This patch is just an example.

Definitely a good idea, because not all applications have logging
like sshd does.  Besides, the first thing I went looking for was
tracepoints. ;)

> 
> ==
> From ed565cbf842e0b30827fba7bfdbc724fe21d9d2d Mon Sep 17 00:00:00 2001
> From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
> Date: Fri, 2 Dec 2011 14:10:51 +0900
> Subject: [PATCH] oom_score_adj trace point.
> 
> oom_score_adj is set by some daemon and launch tasks ans inherited
> to applications, sometimes unexpectedly.
> 
> This patch is for debugging oom_score_adj inheritance. This
> adds trace points for oom_score_adj inheritance.
> 
>     bash-2501  [002]   448.860197: oom_score_adj_update: task 2501[bash] updates oom_score_adj=-1000
>     bash-2501  [002]   455.678190: oom_score_adj_inherited: new task 2527 inherited oom_score_adj -1000
>     ls-2527  [007]   455.678683: oom_score_task_rename: task 2527[bash] to [ls] oom_score_adj=-1000
>     bash-2501  [007]   461.632103: oom_score_adj_inherited: new task 2528 inherited oom_score_adj -1000
>     bash-2501  [007]   461.632335: oom_score_adj_inherited: new task 2529 inherited oom_score_adj -1000
>     ls-2528  [003]   461.632983: oom_score_task_rename: task 2528[bash] to [ls] oom_score_adj=-1000
>     less-2529  [005]   461.633086: oom_score_task_rename: task 2529[bash] to [less] oom_score_adj=-1000
>     bash-2501  [004]   474.888710: oom_score_adj_update: task 2501[bash] updates oom_score_adj=0
> 
> Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>

Acked-by: Dave Chinner <dchinner@...hat.com>

Cheers,

Dave.
-- 
Dave Chinner
david@...morbit.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/