[Intel-gfx] [PATCH 17/56] drm/i915/gt: Skip over completed active execlists, again

[Chris Wilson]

Now that we are careful to always force-restore contexts upon rewinding
(where necessary), we can restore our optimisation to skip over
completed active execlists when dequeuing.

Referenecs: 35f3fd8182ba ("drm/i915/execlists: Workaround switching back to a completed context")
References: 8ab3a3812aa9 ("drm/i915/gt: Incrementally check for rewinding")
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala at linux.intel.com>
---
 .../drm/i915/gt/intel_execlists_submission.c  | 33 ++++++++++---------
 1 file changed, 17 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
index a8544692e7e2..e695852c1f61 100644
--- a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
+++ b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
@@ -1227,12 +1227,20 @@ static void set_preempt_timeout(struct intel_engine_cs *engine,
 		     active_preempt_timeout(engine, rq));
 }
 
+static bool completed(const struct i915_request *rq)
+{
+	if (i915_request_has_sentinel(rq))
+		return false;
+
+	return __i915_request_is_complete(rq);
+}
+
 static void execlists_dequeue(struct intel_engine_cs *engine)
 {
 	struct intel_engine_execlists * const execlists = &engine->execlists;
 	struct i915_request **port = execlists->pending;
 	struct i915_request ** const last_port = port + execlists->port_mask;
-	struct i915_request *last = *execlists->active;
+	struct i915_request *last, * const *active;
 	struct virtual_engine *ve;
 	struct rb_node *rb;
 	bool submit = false;
@@ -1269,21 +1277,15 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 	 * i.e. we will retrigger preemption following the ack in case
 	 * of trouble.
 	 *
-	 * In theory we can skip over completed contexts that have not
-	 * yet been processed by events (as those events are in flight):
-	 *
-	 * while ((last = *active) && i915_request_completed(last))
-	 *	active++;
-	 *
-	 * However, the GPU cannot handle this as it will ultimately
-	 * find itself trying to jump back into a context it has just
-	 * completed and barf.
 	 */
+	active = execlists->active;
+	while ((last = *active) && completed(last))
+		active++;
 
 	if (last) {
-		if (__i915_request_is_complete(last)) {
-			goto check_secondary;
-		} else if (need_preempt(engine, last)) {
+		GEM_BUG_ON(!i915_request_is_active(last));
+
+		if (need_preempt(engine, last)) {
 			ENGINE_TRACE(engine,
 				     "preempting last=%llx:%lld, prio=%d, hint=%d\n",
 				     last->fence.context,
@@ -1346,7 +1348,6 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 			 * we hopefully coalesce several updates into a single
 			 * submission.
 			 */
-check_secondary:
 			if (!list_is_last(&last->sched.link,
 					  &engine->active.requests)) {
 				/*
@@ -1551,7 +1552,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 	 * of ordered contexts.
 	 */
 	if (submit &&
-	    memcmp(execlists->active,
+	    memcmp(active,
 		   execlists->pending,
 		   (port - execlists->pending) * sizeof(*port))) {
 		*port = NULL;
@@ -1559,7 +1560,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 			execlists_schedule_in(*port, port - execlists->pending);
 
 		WRITE_ONCE(execlists->yield, -1);
-		set_preempt_timeout(engine, *execlists->active);
+		set_preempt_timeout(engine, *active);
 		execlists_submit_ports(engine);
 	} else {
 		ring_set_paused(engine, 0);
-- 
2.20.1