drm/i915/execlists: Remove incorrect BUG_ON for schedule-out
authorChris Wilson <chris@chris-wilson.co.uk>
Sat, 7 Sep 2019 10:50:46 +0000 (11:50 +0100)
committerChris Wilson <chris@chris-wilson.co.uk>
Mon, 9 Sep 2019 10:28:06 +0000 (11:28 +0100)
As we may unwind incomplete requests (for preemption) prior to
processing the CSB and the schedule-out events, we may update rq->engine
(resetting it to point back to the parent virtual engine) prior to
calling execlists_schedule_out(), invalidating the assertion that the
request still points to the inflight engine. (The likelihood of this is
increased if the CSB interrupt processing is pushed to the ksoftirqd for
being too slow and direct submission overtakes it.)

Tvrtko summarised it as:
"So unwind from direct submission resets rq->engine and races with
process_csb from the tasklet which notices request has actually
completed."

Reported-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Fixes: df403069029d ("drm/i915/execlists: Lift process_csb() out of the irq-off spinlock")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190907105046.19934-1-chris@chris-wilson.co.uk
drivers/gpu/drm/i915/gt/intel_lrc.c

index 0ddfbeb..f073aea 100644 (file)
@@ -631,7 +631,6 @@ execlists_schedule_out(struct i915_request *rq)
        struct intel_engine_cs *cur, *old;
 
        trace_i915_request_out(rq);
-       GEM_BUG_ON(intel_context_inflight(ce) != rq->engine);
 
        old = READ_ONCE(ce->inflight);
        do