1 files changed, 89 insertions, 25 deletions

diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index bbd71af00a91..765bec89fc0d 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -300,11 +300,11 @@ void i915_request_retire_upto(struct i915_request *rq)
 }
 
 static int
-__i915_request_await_execution(struct i915_request *rq,
-			       struct i915_request *signal,
-			       void (*hook)(struct i915_request *rq,
-					    struct dma_fence *signal),
-			       gfp_t gfp)
+__await_execution(struct i915_request *rq,
+		  struct i915_request *signal,
+		  void (*hook)(struct i915_request *rq,
+			       struct dma_fence *signal),
+		  gfp_t gfp)
 {
 	struct execute_cb *cb;
 
@@ -341,6 +341,8 @@ __i915_request_await_execution(struct i915_request *rq,
 	}
 	spin_unlock_irq(&signal->lock);
 
+	/* Copy across semaphore status as we need the same behaviour */
+	rq->sched.flags |= signal->sched.flags;
 	return 0;
 }
 
@@ -811,31 +813,21 @@ already_busywaiting(struct i915_request *rq)
 }
 
 static int
-emit_semaphore_wait(struct i915_request *to,
-		    struct i915_request *from,
-		    gfp_t gfp)
+__emit_semaphore_wait(struct i915_request *to,
+		      struct i915_request *from,
+		      u32 seqno)
 {
 	const int has_token = INTEL_GEN(to->i915) >= 12;
 	u32 hwsp_offset;
-	int len;
+	int len, err;
 	u32 *cs;
 
 	GEM_BUG_ON(INTEL_GEN(to->i915) < 8);
 
-	/* Just emit the first semaphore we see as request space is limited. */
-	if (already_busywaiting(to) & from->engine->mask)
-		goto await_fence;
-
-	if (i915_request_await_start(to, from) < 0)
-		goto await_fence;
-
-	/* Only submit our spinner after the signaler is running! */
-	if (__i915_request_await_execution(to, from, NULL, gfp))
-		goto await_fence;
-
 	/* We need to pin the signaler's HWSP until we are finished reading. */
-	if (intel_timeline_read_hwsp(from, to, &hwsp_offset))
-		goto await_fence;
+	err = intel_timeline_read_hwsp(from, to, &hwsp_offset);
+	if (err)
+		return err;
 
 	len = 4;
 	if (has_token)
@@ -858,7 +850,7 @@ emit_semaphore_wait(struct i915_request *to,
 		 MI_SEMAPHORE_POLL |
 		 MI_SEMAPHORE_SAD_GTE_SDD) +
 		has_token;
-	*cs++ = from->fence.seqno;
+	*cs++ = seqno;
 	*cs++ = hwsp_offset;
 	*cs++ = 0;
 	if (has_token) {
@@ -867,6 +859,28 @@ emit_semaphore_wait(struct i915_request *to,
 	}
 
 	intel_ring_advance(to, cs);
+	return 0;
+}
+
+static int
+emit_semaphore_wait(struct i915_request *to,
+		    struct i915_request *from,
+		    gfp_t gfp)
+{
+	/* Just emit the first semaphore we see as request space is limited. */
+	if (already_busywaiting(to) & from->engine->mask)
+		goto await_fence;
+
+	if (i915_request_await_start(to, from) < 0)
+		goto await_fence;
+
+	/* Only submit our spinner after the signaler is running! */
+	if (__await_execution(to, from, NULL, gfp))
+		goto await_fence;
+
+	if (__emit_semaphore_wait(to, from, from->fence.seqno))
+		goto await_fence;
+
 	to->sched.semaphores |= from->engine->mask;
 	to->sched.flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN;
 	return 0;
@@ -980,6 +994,57 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence)
 	return 0;
 }
 
+static bool intel_timeline_sync_has_start(struct intel_timeline *tl,
+					  struct dma_fence *fence)
+{
+	return __intel_timeline_sync_is_later(tl,
+					      fence->context,
+					      fence->seqno - 1);
+}
+
+static int intel_timeline_sync_set_start(struct intel_timeline *tl,
+					 const struct dma_fence *fence)
+{
+	return __intel_timeline_sync_set(tl, fence->context, fence->seqno - 1);
+}
+
+static int
+__i915_request_await_execution(struct i915_request *to,
+			       struct i915_request *from,
+			       void (*hook)(struct i915_request *rq,
+					    struct dma_fence *signal))
+{
+	int err;
+
+	/* Submit both requests at the same time */
+	err = __await_execution(to, from, hook, I915_FENCE_GFP);
+	if (err)
+		return err;
+
+	/* Squash repeated depenendices to the same timelines */
+	if (intel_timeline_sync_has_start(i915_request_timeline(to),
+					  &from->fence))
+		return 0;
+
+	/* Ensure both start together [after all semaphores in signal] */
+	if (intel_engine_has_semaphores(to->engine))
+		err = __emit_semaphore_wait(to, from, from->fence.seqno - 1);
+	else
+		err = i915_request_await_start(to, from);
+	if (err < 0)
+		return err;
+
+	/* Couple the dependency tree for PI on this exposed to->fence */
+	if (to->engine->schedule) {
+		err = i915_sched_node_add_dependency(&to->sched, &from->sched);
+		if (err < 0)
+			return err;
+	}
+
+	return intel_timeline_sync_set_start(i915_request_timeline(to),
+					     &from->fence);
+}
+
 int
 i915_request_await_execution(struct i915_request *rq,
 			     struct dma_fence *fence,
@@ -1013,8 +1078,7 @@ i915_request_await_execution(struct i915_request *rq,
 		if (dma_fence_is_i915(fence))
 			ret = __i915_request_await_execution(rq,
 							     to_request(fence),
-							     hook,
-							     I915_FENCE_GFP);
+							     hook);
 		else
 			ret = i915_sw_fence_await_dma_fence(&rq->submit, fence,
 							    I915_FENCE_TIMEOUT,