drm/i915: Cancel pending execlists irq handler upon idling
authorChris Wilson <chris@chris-wilson.co.uk>
Tue, 27 Jun 2017 15:25:10 +0000 (16:25 +0100)
committerChris Wilson <chris@chris-wilson.co.uk>
Wed, 28 Jun 2017 10:30:33 +0000 (11:30 +0100)
Due to the slight asynchronicity in handling the execlists interrupts
(i.e. we defer the work to a handler that may consume more than one
interrupt event), when the engine is idle we may still have an irq
tasklet queued (especially when it has been deferred to a ksoftirqd).
At the beginning of the tasklet, we assert that we do hold a device
wakeref for the access we are about to perform. This assumes that when
we idle and release the GT wakeref, all execlists work has been
completed (since the elsp tracking says the hw is idle). However, there
may still be a tasklet queued, so as we mark the engine idle, also
cancel any pending tasklet.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170627152510.28589-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
drivers/gpu/drm/i915/intel_engine_cs.c

index 3b46c1f7b88b3894ff308eea34cf5a480121a128..49e875c46c966ad16901fcc4c0965dfd7c12038d 100644 (file)
@@ -1328,6 +1328,7 @@ void intel_engines_mark_idle(struct drm_i915_private *i915)
        for_each_engine(engine, i915, id) {
                intel_engine_disarm_breadcrumbs(engine);
                i915_gem_batch_pool_fini(&engine->batch_pool);
+               tasklet_kill(&engine->irq_tasklet);
                engine->no_priolist = false;
        }
 }