Quoting Michel Thierry (2018-04-06 23:35:43)
> On 4/6/2018 3:03 PM, Chris Wilson wrote:
> > -static u32 fake_hangcheck(struct i915_request *rq)
> > +static u32 fake_hangcheck(struct i915_request *rq, u32 mask)
> > {
> > - u32 reset_count;
> > + struct i915_gpu_error *error = &rq->i915->gpu_error;
> > + u32 reset_count = i915_reset_count(error);
> >
> > - rq->engine->hangcheck.stalled = true;
> > - rq->engine->hangcheck.seqno = intel_engine_get_seqno(rq->engine);
> > + error->stalled_mask = mask;
> >
> > - reset_count = i915_reset_count(&rq->i915->gpu_error);
> > + smp_mb__before_atomic();
> checkpatch is going to complain about the lack of comment
checkpatch is seeing mb where there is only a barrier(). ;)
/* No really, set_bit must be after setting stalled_mask */
The other side is handled via wake_up_all() being a full barrier between
us and the other process.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
