Re: [PATCH v2] blktrace: Fix potentail deadlock between delete & sysfs ops

[Waiman Long <longman@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1659 bytes --]

On 08/17/2017 07:23 PM, Steven Rostedt wrote:
> On Thu, 17 Aug 2017 18:18:18 -0400
> Steven Rostedt <rostedt@goodmis.org> wrote:
>
>> On Thu, 17 Aug 2017 18:13:20 -0400
>> Steven Rostedt <rostedt@goodmis.org> wrote:
>>
>>> On Thu, 17 Aug 2017 17:27:22 -0400
>>> Waiman Long <longman@redhat.com> wrote:
>>>
>>>   
>>>> It is actually what the patch is trying to do by checking for the
>>>> deletion flag in the mutex_trylock loop. Please note that mutex does not
>>>> guarantee FIFO ordering of lock acquisition. As a result, cpu1 may call
>>>> mutex_lock() and wait for it while cpu2 can set the deletion flag later
>>>> and get the mutex first before cpu1. So checking for the deletion flag
>>>> before taking the mutex isn't enough.    
>>> Yeah, I figured that out already (crossed emails). BTW, how did you
>>> trigger this warning. I'm playing around with adding loop devices,
>>> volume groups, and logical volumes, and reading the trace files
>>> created in the sysfs directory, then removing those items, but it's
>>> not triggering the "delete" path. What operation deletes the partition?  
>> I'm guessing that deleting an actual partition may work (unfortunately,
>> my test box has no partition to delete ;-) I'll find another box to
>> test on.
>>
> OK, deleting a partition doesn't trigger the lockdep splat. But I also
> added a printk in the BLKPG_DEL_PARTITION case switch, which also
> doesn't print. What command do I need to do to trigger that path?
>
> Thanks,
>
> -- Steve

Attached is a reproducer that was used to trigger the warning. Some
tuning may be needed depend on the actual configuration of the test machine.

Cheers,
Longman


[-- Attachment #2: run_test.sh --]
[-- Type: application/x-shellscript, Size: 198 bytes --]