Re: [PATCH] workqueue/lockdep: Explicitly initialize wq_barrier::done::map
From: Tejun Heo
Date: Thu Aug 17 2017 - 09:26:45 EST
On Thu, Aug 17, 2017 at 05:46:12PM +0800, Boqun Feng wrote:
> With CROSSRELEASE feature introduced, such a potential deadlock is
> reported:
>
> > Worker A : acquired of wfc.work -> wait for cpu_hotplug_lock to be released
> > Task B : acquired of cpu_hotplug_lock -> wait for lock#3 to be released
> > Task C : acquired of lock#3 -> wait for completion of barr->done
> > (Task C is in lru_add_drain_all_cpuslocked())
> > Worker D : wait for wfc.work to be released -> will complete barr->done
>
> However, given this very case, such a dead lock could not happen because
> Task C's barr->done and Worker D's barr->done can not be the same
> instance. So this is a false positive.
>
> The reason of this false positive is we initialize all wq_barrier::done
> at insert_wq_barrier() via init_completion(), which makes them belong to
> the same lock class, therefore, impossible circles are reported.
>
> To fix this, explicitly initialize the lockdep map for wq_barrier::done
> in insert_wq_barrier(), so that the lock class key of wq_barrier::done
> is a subkey of the corresponding work_struct, as a result we won't build
> a dependency between a wq_barrier with a unrelated work, and we can
> differ wq barriers based on the related works, so the false positive
> above is avoided.
>
> Also define the empty lockdep_init_map_crosslock() for !CROSSRELEASE
> to make the code simple and away from unnecessary #ifdefs.
>
> Reported-by: Ingo Molnar <mingo@xxxxxxxxxx>
> Signed-off-by: Boqun Feng <boqun.feng@xxxxxxxxx>
> Cc: Byungchul Park <byungchul.park@xxxxxxx>
Acked-by: Tejun Heo <tj@xxxxxxxxxx>
Thanks.
--
tejun