[Intel-gfx] [PATCH] drm/i915/gt: Be defensive in the face of false CS events

[Chris Wilson]

Quoting Tvrtko Ursulin (2020-07-10 13:30:09)
> 
> On 10/07/2020 13:16, Chris Wilson wrote:
> > If the HW throws a curve ball and reports either en event before it is
> > possible, or just a completely impossible event, we have to grin and
> > bear it. The first few events, we will likely not notice as we would be
> > expecting some event, but as soon as we stop expecting an event and yet
> > they still keep coming, then we enter into undefined state territory.
> > In which case, bail out, stop processing the events, and reset the
> > engine and our set of queued requests to recover.
> > 
> > The sporadic hangs and warnings will continue to plague CI, but at least
> > system stability should not be compromised.
> > 
> > Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2045
> > 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 | 8 ++++++--
> >   1 file changed, 6 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > index fbcfeaed6441..c86324d2d2bb 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > @@ -2567,6 +2567,7 @@ static void process_csb(struct intel_engine_cs *engine)
> >       tail = READ_ONCE(*execlists->csb_write);
> >       if (unlikely(head == tail))
> >               return;
> > +     execlists->csb_head = tail;
> 
> This deserves a comment...
> 
> >   
> >       /*
> >        * Hopefully paired with a wmb() in HW!
> > @@ -2613,6 +2614,9 @@ static void process_csb(struct intel_engine_cs *engine)
> >               if (promote) {
> >                       struct i915_request * const *old = execlists->active;
> >   
> > +                     if (GEM_WARN_ON(!*execlists->pending))
> > +                             break;
> > +
> 
> ... but why not continue? You think nothing good can come out of trying 
> further and break simply expedites the hang? We have to be confident we 
> can cope with any random i915 state caused by skipping maybe valid entries.

We are already past the point of no return as the events coming from HW
do not correspond to our events; continuing on cannot recover, we will
already have made mistakes.
 
> Conclusion will define what kind of comment to put above. "Assume we 
> always consume all CSB entries, or things are really bad and we mark all 
> as invalid upon finding first bad entry"?

It's dead, Jim.

We escape out, reset the engine/GPU, consign the port tracking to the bin,
and reload with the next set of requests.
-Chris