Re: tasklist_lock lockdep warnings on 3.6

[Sasha Levin <levinsasha928@gmail.com>]

On 08/02/2012 04:16 PM, Sasha Levin wrote:
> On 08/02/2012 03:26 PM, Dave Jones wrote:
>> On Mon, Jul 30, 2012 at 04:50:46PM +0200, Sasha Levin wrote:
>>  > ping?
>>  > 
>>  > I'm still seeing this on linux-next.
>>
>> Likewise, except I'm seeing it in Linus' tree since shortly after this merge window
>> began. (https://lkml.org/lkml/2012/7/24/443).
>>
>> I've spent all of this last week doing multiple attempts at bisecting this without success.
>> (It doesn't always show up, and even after running the test for over an hour before
>> marking a build 'good')
>>
>> Anyone ?
>>
>> 	Dave
> 
> I've been able to hit this issue pretty much every time by hitting sysrq-t after about 10 seconds of fuzzing. It shows up in the middle of the sysrq-t spew.
> 
> Unfortunately I didn't have time yet to try and bisect it.
> 

So I've bisected it. The bisection was automatic most of the time, but I've confirmed that this is indeed the offending commit manually.

d35abdb28824cf74f0a106a0f9c6f3ff700a35bf is the first bad commit
commit d35abdb28824cf74f0a106a0f9c6f3ff700a35bf
Author: Al Viro <viro@zeniv.linux.org.uk>
Date:   Sat Jun 30 11:55:24 2012 +0400

    hold task_lock around checks in keyctl

    Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>


Relevant parties Cc'ed.