netdev - Re: [PATCH 2/2] af_unix: split 'u->readlock' into two: 'iolock' and 'bindlock'

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]

Date:   Fri, 2 Sep 2016 22:23:07 +0200
From:   Hannes Frederic Sowa <hannes@...essinduktion.org>
To:     David Miller <davem@...emloft.net>, torvalds@...ux-foundation.org
Cc:     rweikusat@...ileactivedefense.com, edumazet@...gle.com, w@....eu,
        netdev@...r.kernel.org
Subject: Re: [PATCH 2/2] af_unix: split 'u->readlock' into two: 'iolock' and
 'bindlock'

On 02.09.2016 21:15, David Miller wrote:
> From: Linus Torvalds <torvalds@...ux-foundation.org>
> Date: Fri, 2 Sep 2016 11:17:18 -0700
> 
>> Oh, this was missing a
>>
>>   Reported-and-tested-by: CAI Qian <caiqian@...hat.com>
>>
>> who found the new deadlock.
>>
>> There's now *another* lockdep deadlock report by him, but that one has
>> nothing to do with networking.
>>
>> (And neither of these deadlocks will actually deadlock the machine in
>> practice, but you can trigger the lockdep reports with some odd splice
>> patterns and overlayfs use)
> 
> I read over this and can't find any problems.
> 
> The main thing I was concerned about was an I/O path that really
> expects the socket's hash not to change for whatever reason, but even
> all of the unix_find_other() calls are done outside of the mutex
> already.

Furthermore we don't allow rehashing, if the socket is "published" once,
it can only be destroyed but can't change identity during its lifetime.
IIRC this was another bug we had a while ago.