From: Chris Wilson <chris@chris-wilson.co.uk>
Date: Wed, 27 Jun 2018 21:13:01 +0100
Subject: drm/i915: Reduce spinlock hold time during notify_ring() interrupt
Git-commit: 3f88325c2e396658a17632a19b59958750d7ad89
Patch-mainline: v4.19-rc1
References: FATE#326289 FATE#326079 FATE#326049 FATE#322398 FATE#326166

By taking advantage of the RCU protection of the task struct, we can find
the appropriate signaler under the spinlock and then release the spinlock
before waking the task and signaling the fence.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180627201304.15817-1-chris@chris-wilson.co.uk

Acked-by: Petr Tesarik <ptesarik@suse.com>
---
 drivers/gpu/drm/i915/i915_irq.c |   32 +++++++++++++++++++++-----------
 1 file changed, 21 insertions(+), 11 deletions(-)

--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -1154,21 +1154,23 @@ static void ironlake_rps_change_irq_hand
 
 static void notify_ring(struct intel_engine_cs *engine)
 {
+	const u32 seqno = intel_engine_get_seqno(engine);
 	struct i915_request *rq = NULL;
+	struct task_struct *tsk = NULL;
 	struct intel_wait *wait;
 
-	if (!engine->breadcrumbs.irq_armed)
+	if (unlikely(!engine->breadcrumbs.irq_armed))
 		return;
 
 	atomic_inc(&engine->irq_count);
-	set_bit(ENGINE_IRQ_BREADCRUMB, &engine->irq_posted);
+
+	rcu_read_lock();
 
 	spin_lock(&engine->breadcrumbs.irq_lock);
 	wait = engine->breadcrumbs.irq_wait;
 	if (wait) {
-		bool wakeup = engine->irq_seqno_barrier;
-
-		/* We use a callback from the dma-fence to submit
+		/*
+		 * We use a callback from the dma-fence to submit
 		 * requests after waiting on our own requests. To
 		 * ensure minimum delay in queuing the next request to
 		 * hardware, signal the fence now rather than wait for
@@ -1179,19 +1181,22 @@ static void notify_ring(struct intel_eng
 		 * and to handle coalescing of multiple seqno updates
 		 * and many waiters.
 		 */
-		if (i915_seqno_passed(intel_engine_get_seqno(engine),
-				      wait->seqno)) {
+		if (i915_seqno_passed(seqno, wait->seqno)) {
 			struct i915_request *waiter = wait->request;
 
-			wakeup = true;
 			if (!test_bit(DMA_FENCE_FLAG_SIGNALED_BIT,
 				      &waiter->fence.flags) &&
 			    intel_wait_check_request(wait, waiter))
 				rq = i915_request_get(waiter);
-		}
 
-		if (wakeup)
-			wake_up_process(wait->tsk);
+			tsk = wait->tsk;
+		} else {
+			if (engine->irq_seqno_barrier) {
+				set_bit(ENGINE_IRQ_BREADCRUMB,
+					&engine->irq_posted);
+				tsk = wait->tsk;
+			}
+		}
 	} else {
 		if (engine->breadcrumbs.irq_armed)
 			__intel_engine_disarm_breadcrumbs(engine);
@@ -1204,6 +1209,11 @@ static void notify_ring(struct intel_eng
 		i915_request_put(rq);
 	}
 
+	if (tsk && tsk->state & TASK_NORMAL)
+		wake_up_process(tsk);
+
+	rcu_read_unlock();
+
 	trace_intel_engine_notify(engine, wait);
 }