1 files changed, 33 insertions, 47 deletions

diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c b/drivers/gpu/drm/nouveau/nouveau_fence.c
index 05d0b3eb3690..a3a04e0d76ec 100644
--- a/drivers/gpu/drm/nouveau/nouveau_fence.c
+++ b/drivers/gpu/drm/nouveau/nouveau_fence.c
@@ -339,68 +339,54 @@ nouveau_fence_wait(struct nouveau_fence *fence, bool lazy, bool intr)
 }
 
 int
-nouveau_fence_sync(struct nouveau_bo *nvbo, struct nouveau_channel *chan, bool exclusive, bool intr)
+nouveau_fence_sync(struct nouveau_bo *nvbo, struct nouveau_channel *chan,
+		   bool exclusive, bool intr)
 {
 	struct nouveau_fence_chan *fctx = chan->fence;
-	struct dma_fence *fence;
 	struct dma_resv *resv = nvbo->bo.base.resv;
-	struct dma_resv_list *fobj;
-	struct nouveau_fence *f;
-	int ret = 0, i;
+	int i, ret;
 
 	if (!exclusive) {
 		ret = dma_resv_reserve_shared(resv, 1);
-
 		if (ret)
 			return ret;
 	}
 
-	fobj = dma_resv_shared_list(resv);
-	fence = dma_resv_excl_fence(resv);
-
-	if (fence) {
-		struct nouveau_channel *prev = NULL;
-		bool must_wait = true;
-
-		f = nouveau_local_fence(fence, chan->drm);
-		if (f) {
-			rcu_read_lock();
-			prev = rcu_dereference(f->channel);
-			if (prev && (prev == chan || fctx->sync(f, prev, chan) == 0))
-				must_wait = false;
-			rcu_read_unlock();
-		}
+	/* Waiting for the exclusive fence first causes performance regressions
+	 * under some circumstances. So manually wait for the shared ones first.
+	 */
+	for (i = 0; i < 2; ++i) {
+		struct dma_resv_iter cursor;
+		struct dma_fence *fence;
+
+		dma_resv_for_each_fence(&cursor, resv, exclusive, fence) {
+			struct nouveau_fence *f;
+
+			if (i == 0 && dma_resv_iter_is_exclusive(&cursor))
+				continue;
+
+			f = nouveau_local_fence(fence, chan->drm);
+			if (f) {
+				struct nouveau_channel *prev;
+				bool must_wait = true;
+
+				rcu_read_lock();
+				prev = rcu_dereference(f->channel);
+				if (prev && (prev == chan ||
+					     fctx->sync(f, prev, chan) == 0))
+					must_wait = false;
+				rcu_read_unlock();
+				if (!must_wait)
+					continue;
+			}
 
-		if (must_wait)
 			ret = dma_fence_wait(fence, intr);
-
-		return ret;
-	}
-
-	if (!exclusive || !fobj)
-		return ret;
-
-	for (i = 0; i < fobj->shared_count && !ret; ++i) {
-		struct nouveau_channel *prev = NULL;
-		bool must_wait = true;
-
-		fence = rcu_dereference_protected(fobj->shared[i],
-						dma_resv_held(resv));
-
-		f = nouveau_local_fence(fence, chan->drm);
-		if (f) {
-			rcu_read_lock();
-			prev = rcu_dereference(f->channel);
-			if (prev && (prev == chan || fctx->sync(f, prev, chan) == 0))
-				must_wait = false;
-			rcu_read_unlock();
+			if (ret)
+				return ret;
 		}
-
-		if (must_wait)
-			ret = dma_fence_wait(fence, intr);
 	}
 
-	return ret;
+	return 0;
 }
 
 void