[Intel-gfx] [PATCH v2] drm/i915/execlists: Track active elements during dequeue

[Chris Wilson]

Record the initial active element we use when building the next ELSP
submission, so that we can compare against it latter to see if there's
no change.

Fixes: 44d0a9c05bc0 ("drm/i915/execlists: Skip redundant resubmission")
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/gt/intel_lrc.c | 36 +++++++++++------------------
 1 file changed, 13 insertions(+), 23 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 0604b11a2842..15b81be08fcc 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1678,17 +1678,6 @@ static void virtual_xfer_breadcrumbs(struct virtual_engine *ve,
 	spin_unlock(&old->breadcrumbs.irq_lock);
 }
 
-static struct i915_request *
-last_active(const struct intel_engine_execlists *execlists)
-{
-	struct i915_request * const *last = READ_ONCE(execlists->active);
-
-	while (*last && i915_request_completed(*last))
-		last++;
-
-	return *last;
-}
-
 #define for_each_waiter(p__, rq__) \
 	list_for_each_entry_lockless(p__, \
 				     &(rq__)->sched.waiters_list, \
@@ -1827,14 +1816,9 @@ static void record_preemption(struct intel_engine_execlists *execlists)
 	(void)I915_SELFTEST_ONLY(execlists->preempt_hang.count++);
 }
 
-static unsigned long active_preempt_timeout(struct intel_engine_cs *engine)
+static unsigned long active_preempt_timeout(struct intel_engine_cs *engine,
+					    const struct i915_request *rq)
 {
-	struct i915_request *rq;
-
-	rq = last_active(&engine->execlists);
-	if (!rq)
-		return 0;
-
 	/* Force a fast reset for terminated contexts (ignoring sysfs!) */
 	if (unlikely(intel_context_is_banned(rq->context)))
 		return 1;
@@ -1842,13 +1826,14 @@ static unsigned long active_preempt_timeout(struct intel_engine_cs *engine)
 	return READ_ONCE(engine->props.preempt_timeout_ms);
 }
 
-static void set_preempt_timeout(struct intel_engine_cs *engine)
+static void set_preempt_timeout(struct intel_engine_cs *engine,
+				const struct i915_request *rq)
 {
 	if (!intel_engine_has_preempt_reset(engine))
 		return;
 
 	set_timer_ms(&engine->execlists.preempt,
-		     active_preempt_timeout(engine));
+		     active_preempt_timeout(engine, rq));
 }
 
 static inline void clear_ports(struct i915_request **ports, int count)
@@ -1861,6 +1846,7 @@ 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 * const *active;
 	struct i915_request *last;
 	struct rb_node *rb;
 	bool submit = false;
@@ -1915,7 +1901,11 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 	 * i.e. we will retrigger preemption following the ack in case
 	 * of trouble.
 	 */
-	last = last_active(execlists);
+	active = READ_ONCE(execlists->active);
+	while (*active && i915_request_completed(*active))
+		active++;
+
+	last = *active;
 	if (last) {
 		if (need_preempt(engine, last, rb)) {
 			ENGINE_TRACE(engine,
@@ -2200,7 +2190,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 		 * Skip if we ended up with exactly the same set of requests,
 		 * e.g. trying to timeslice a pair of ordered contexts
 		 */
-		if (!memcmp(execlists->active, execlists->pending,
+		if (!memcmp(active, execlists->pending,
 			    (port - execlists->pending + 1) * sizeof(*port))) {
 			do
 				execlists_schedule_out(fetch_and_zero(port));
@@ -2210,8 +2200,8 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 		}
 		clear_ports(port + 1, last_port - port);
 
+		set_preempt_timeout(engine, *execlists->pending);
 		execlists_submit_ports(engine);
-		set_preempt_timeout(engine);
 	} else {
 skip_submit:
 		ring_set_paused(engine, 0);
-- 
2.20.1