Re: [PATCH-tip 0/5] locking/locktorture: Fix locktorture ww_mutex test problems

[Waiman Long]

On 3/19/21 7:16 AM, Ingo Molnar wrote:
> * Waiman Long <longman@xxxxxxxxxx> wrote:
>
> > This is a follow-up patch series for the previous patchset on fixing
> > locktorture ww_mutex test problem [1]. The first 3 patches of that
> > series were merged into tip. It turns out that the last one of the
> > three wasn't quite right. So this patch series revert the last patch.
> >
> > The rests of the patch series fix the ww_mutex testing problem in
> > locktorture as well as removing the DEFINE_WW_MUTEX() macro from
> > include/linux/ww_mutex.h.
> >
> > [1] https://lore.kernel.org/lkml/20210316153119.13802-1-longman@xxxxxxxxxx/
> >
> > Waiman Long (5):
> >    locking/ww_mutex: Revert "Treat ww_mutex_lock() like a trylock"
> >    locking/locktorture: Fix false positive circular locking splat in
> >      ww_mutex test
> >    locking/ww_mutex: Remove DEFINE_WW_MUTEX() macro
> >    locking/locktorture: Pass thread id to lock/unlock functions
> >    locking/locktorture: locking/locktorture: Fix incorrect use of
> >      ww_acquire_ctx in ww_mutex test
> Applied, thanks Waiman.
>
> I kept these two fixes in locking/urgent, for a v5.12 merge:
>
>    bee645788e07: ("locking/ww_mutex: Fix acquire/release imbalance in ww_acquire_init()/ww_acquire_fini()")
>    5de2055d31ea: ("locking/ww_mutex: Simplify use_ww_ctx & ww_ctx handling")
>
> As this bug could affect actual ww_mutex users.
>
> And queued up these four in locking/core, for a v5.13 merge:
>
>    8c52cca04f97: ("locking/locktorture: Fix incorrect use of ww_acquire_ctx in ww_mutex test")
>    aa3a5f31877e: ("locking/locktorture: Pass thread id to lock/unlock functions")
>    5261ced47f8e: ("locking/ww_mutex: Remove DEFINE_WW_MUTEX() macro")
>    2ea55bbba23e: ("locking/locktorture: Fix false positive circular locking splat in ww_mutex test")
>
> As these bugs are basically limited to a debugging facility.
>
> ( But we could also merge them into v5.12, if you think it's
>    justified. No strong opinions either way. )

I think v5.13 merge is fine. This problem exists for quite a while and 
no one notices it. So it is not really that urgent.

Thank a lot for the quick action.

Cheers,
Longman