drm/i915: Flush submission tasklet after bumping priority

When called from process context tasklet_schedule() defers itself to
ksoftirqd. From experience this may cause unacceptable latencies of over
200ms in executing the submission tasklet, our goal is to reprioritise
the HW execution queue and trigger HW preemption immediately, so disable
bh over the call to schedule and force the tasklet to run afterwards if
scheduled.

v2: Keep rcu_read_lock() around for PREEMPT_RCU

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180507135731.10587-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 5ece6ae..89bf5d6 100644 (file)
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -578,10 +578,12 @@ static void __fence_set_priority(struct dma_fence *fence,
        rq = to_request(fence);
        engine = rq->engine;
 
-       rcu_read_lock();
+       local_bh_disable();
+       rcu_read_lock(); /* RCU serialisation for set-wedged protection */
        if (engine->schedule)
                engine->schedule(rq, attr);
        rcu_read_unlock();
+       local_bh_enable(); /* kick the tasklets if queues were reprioritised */
 }
 
 static void fence_set_priority(struct dma_fence *fence,