[Intel-gfx] [PATCH] drm/i915/execlists: Assert tasklet is locked for process_csb()

[Chris Wilson]

Quoting Chris Wilson (2019-10-14 10:27:59)
> Quoting Tvrtko Ursulin (2019-10-14 10:25:04)
> > 
> > On 13/10/2019 20:31, Chris Wilson wrote:
> > > We rely on only the tasklet being allowed to call into process_csb(), so
> > > assert that is locked when we do. As the tasklet uses a simple bitlock,
> > > there is no strong lockdep checking so we must make do with a plain
> > > assertion that the tasklet is running and assume that we are the
> > > tasklet!
> > > 
> > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > > Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> > > ---
> > >   drivers/gpu/drm/i915/gt/intel_lrc.c | 1 +
> > >   drivers/gpu/drm/i915/i915_gem.h     | 5 +++++
> > >   2 files changed, 6 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > > index 8be9e69d5718..ab20433182d1 100644
> > > --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> > > +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > > @@ -1984,6 +1984,7 @@ static void process_csb(struct intel_engine_cs *engine)
> > >       u8 head, tail;
> > >   
> > >       GEM_BUG_ON(USES_GUC_SUBMISSION(engine->i915));
> > > +     GEM_BUG_ON(!tasklet_is_locked(&execlists->tasklet));
> > 
> > I see some reset paths calling process_csb which looks like a problem 
> > for this assert.
> 
> Oh, no it's not :)
> 
> execlists_reset() is surrounded by reset_prepare and reset_finish who
> are responsible for disabling the tasklet and serialising with direct
> submission around the reset.

Same being true for cancel_requests, on wedging we have to disable the
tasklet before taking control of the state.
-Chris