[Intel-gfx] [PATCH] drm/i915/selftests: Try to show where the pulse went

[Chris Wilson]

Quoting Mika Kuoppala (2019-11-28 11:27:30)
> Chris Wilson <chris at chris-wilson.co.uk> writes:
> 
> > We have a case of a mysteriously absent pulse, so dump the engine
> > details to see if we can find out what happened to it.
> >
> > References: https://bugs.freedesktop.org/show_bug.cgi?id=112405
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> >  drivers/gpu/drm/i915/gt/intel_engine_cs.c           | 2 ++
> >  drivers/gpu/drm/i915/gt/selftest_engine_heartbeat.c | 5 +++++
> >  2 files changed, 7 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> > index 8f6e353caa66..df3369c3f330 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> > @@ -1479,6 +1479,8 @@ void intel_engine_dump(struct intel_engine_cs *engine,
> >               drm_printf(m, "*** WEDGED ***\n");
> >  
> >       drm_printf(m, "\tAwake? %d\n", atomic_read(&engine->wakeref.count));
> > +     drm_printf(m, "\tBarriers?: %s\n",
> > +                yesno(!llist_empty(&engine->barrier_tasks)));
> 
> A random thought arise when reading this and it was that
> should we do a memory barrier before dump. But there
> would be no pairing and it could actually hide something
> more sinister.

Yeah, exactly who do we want to serialise with -- and we will always
need a snapshot just in case we screw that up :)

It's debug; as async as we can without oopsing, and keep on staring
until it starts to make sense.
-Chris