[Intel-gfx] [PATCH] drm/i915: Reduce nested prepare_remote_context() to a trylock

[Chris Wilson]

Quoting Chris Wilson (2019-11-26 12:18:07)
> Quoting Tvrtko Ursulin (2019-11-26 12:15:58)
> > 
> > On 26/11/2019 06:55, Chris Wilson wrote:
> > > On context retiring, we may invoke the kernel_context to unpin this
> > > context. Elsewhere, we may use the kernel_context to modify this
> > > context. This currently leads to an AB-BA lock inversion, so we need to
> > > back-off from the contended lock, and repeat.
> > > 
> > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111732
> > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > > ---
> > >   drivers/gpu/drm/i915/gt/intel_context.c | 6 ++----
> > >   1 file changed, 2 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
> > > index ee9d2bcd2c13..4fcb98f96da6 100644
> > > --- a/drivers/gpu/drm/i915/gt/intel_context.c
> > > +++ b/drivers/gpu/drm/i915/gt/intel_context.c
> > > @@ -310,10 +310,8 @@ int intel_context_prepare_remote_request(struct intel_context *ce,
> > >       GEM_BUG_ON(rq->hw_context == ce);
> > >   
> > >       if (rcu_access_pointer(rq->timeline) != tl) { /* timeline sharing! */
> > > -             err = mutex_lock_interruptible_nested(&tl->mutex,
> > > -                                                   SINGLE_DEPTH_NESTING);
> > > -             if (err)
> > > -                     return err;
> > > +             if (!mutex_trylock(&tl->mutex))
> > > +                     return -EAGAIN;
> > >   
> > >               /* Queue this switch after current activity by this context. */
> > >               err = i915_active_fence_set(&tl->last_request, rq);
> > > 
> > 
> > Please just drop a short comment above the trylock since with git blame 
> > it is often very hard to find the commit.
> 
> Ok. I'm still hoping to find another way to provide the serialisation
> cleanly, but with engine_retire() being more aggressive, the rate of
> contention has increased :(

Hmm. Staring at i915_active_fence_set()... That could be made to be
atomic with only a small amount of hassle. (By small, I mean by the
usual RCU standards.)
-Chris