linux-kernel - Re: [patch] mm, oom: prevent additional oom kills before memory is freed

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: <alpine.DEB.2.10.1706151534170.140219@chino.kir.corp.google.com>
Date:   Thu, 15 Jun 2017 15:42:23 -0700 (PDT)
From:   David Rientjes <rientjes@...gle.com>
To:     Michal Hocko <mhocko@...nel.org>
cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>,
        linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [patch] mm, oom: prevent additional oom kills before memory is
 freed

On Fri, 16 Jun 2017, Michal Hocko wrote:

> I am sorry but I have really hard to make the oom reaper a reliable way
> to stop all the potential oom lockups go away. I do not want to
> reintroduce another potential lockup now.

Please show where this "potential lockup" ever existed in a bug report or 
a testcase?  I have never seen __mmput() block when trying to free the 
memory it maps.

> I also do not see why any
> solution should be rushed into. I have proposed a way to go and unless
> it is clear that this is not a way forward then I simply do not agree
> with any partial workarounds or shortcuts.

This is not a shortcut, it is a bug fix.  4.12 kills 1-4 processes 
unnecessarily as a result of setting MMF_OOM_SKIP incorrectly before the 
mm's memory can be freed.  If you have not seen this issue before, which 
is why you asked if I ever observed it in practice, then you have not 
stress tested oom reaping.  It is very observable and reproducible.  I do 
not agree that adding additional and obscure locking into __mmput() is the 
solution to what is plainly and obviously fixed with this simple patch.

4.12 needs to stop killing 2-5 processes on every oom condition instead of 
1.