[PATCH 13/34] prune-fences

[Chris Wilson]

---
 drivers/dma-buf/reservation.c   | 76 +++++++++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_gem.c | 12 +++----
 drivers/gpu/drm/i915/i915_vma.c |  6 ++--
 include/linux/reservation.h     |  2 ++
 4 files changed, 88 insertions(+), 8 deletions(-)

diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c
index de7b6e709a68..0fc2b8644b24 100644
--- a/drivers/dma-buf/reservation.c
+++ b/drivers/dma-buf/reservation.c
@@ -219,6 +219,82 @@ void reservation_object_add_excl_fence(struct reservation_object *obj,
 }
 EXPORT_SYMBOL(reservation_object_add_excl_fence);
 
+int reservation_object_prune_fences(struct reservation_object *obj, gfp_t gfp)
+{
+	struct reservation_object_list *old, *fobj = NULL;
+	struct dma_fence *check, *old_fence = NULL;
+	u32 uninitialized_var(discard);
+
+	old = reservation_object_get_list(obj);
+	if (old) {
+		u32 i, j;
+
+		fobj = obj->staged;
+		obj->staged = NULL;
+		if (!fobj) {
+			fobj = kmalloc(offsetof(typeof(*fobj),
+						shared[old->shared_max]),
+				       gfp);
+			if (!fobj)
+				return -ENOMEM;
+
+			fobj->shared_max = old->shared_max;
+		}
+
+		for (i = j = 0, discard = fobj->shared_max;
+		     i < old->shared_count; ++i) {
+			u32 idx;
+
+			check = rcu_dereference_protected(old->shared[i],
+							  reservation_object_held(obj));
+
+			if (dma_fence_is_signaled(check))
+				idx = --discard;
+			else
+				idx = j++;
+			RCU_INIT_POINTER(fobj->shared[idx], check);
+		}
+		fobj->shared_count = j;
+
+		if (j == old->shared_count) {
+			kfree(fobj);
+			fobj = old;
+		}
+	}
+
+	check = rcu_dereference_protected(obj->fence_excl,
+					  reservation_object_held(obj));
+	if (check && dma_fence_is_signaled(check)) {
+		old_fence = check;
+		check = NULL;
+	}
+
+	if (rcu_access_pointer(obj->fence_excl) == check &&
+	    rcu_access_pointer(obj->fence) == fobj)
+		return 0;
+
+	write_seqcount_begin(&obj->seq);
+
+	RCU_INIT_POINTER(obj->fence_excl, check);
+	RCU_INIT_POINTER(obj->fence, fobj);
+
+	/* write_seqcount_end() provides the necessary mb for the RCU writes */
+	write_seqcount_end(&obj->seq);
+
+	dma_fence_put(old_fence);
+	if (fobj != old) {
+		for (; discard < fobj->shared_max; discard++) {
+			check = rcu_dereference_protected(fobj->shared[discard],
+							  reservation_object_held(obj));
+			dma_fence_put(check);
+		}
+		kfree_rcu(old, rcu);
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(reservation_object_prune_fences);
+
 /**
 * reservation_object_copy_fences - Copy all fences from src to dst.
 * @dst: the destination reservation object
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index de8f5fe45e38..58dd7bbad290 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -450,12 +450,12 @@ i915_gem_object_wait_reservation(struct reservation_object *resv,
 	 * signaled and that the reservation object has not been changed (i.e.
 	 * no new fences have been added).
 	 */
-	if (prune_fences && !__read_seqcount_retry(&resv->seq, seq)) {
-		if (reservation_object_trylock(resv)) {
-			if (!__read_seqcount_retry(&resv->seq, seq))
-				reservation_object_add_excl_fence(resv, NULL);
-			reservation_object_unlock(resv);
-		}
+	if (prune_fences && reservation_object_trylock(resv)) {
+		if (!__read_seqcount_retry(&resv->seq, seq))
+			reservation_object_add_excl_fence(resv, NULL);
+		else
+			reservation_object_prune_fences(resv, GFP_KERNEL);
+		reservation_object_unlock(resv);
 	}
 
 	return timeout;
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index e0e7c48f45dc..e8752fd7e047 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -56,8 +56,10 @@ i915_vma_retire(struct i915_gem_active *active,
 
 	/* Prune the shared fence arrays iff completely idle (inc. external) */
 	if (reservation_object_trylock(obj->resv)) {
-		if (reservation_object_test_signaled_rcu(obj->resv, true))
-			reservation_object_add_excl_fence(obj->resv, NULL);
+		reservation_object_prune_fences(obj->resv,
+						GFP_KERNEL |
+					       	__GFP_NORETRY |
+					       	__GFP_NOWARN);
 		reservation_object_unlock(obj->resv);
 	}
 
diff --git a/include/linux/reservation.h b/include/linux/reservation.h
index 02166e815afb..c95531d05788 100644
--- a/include/linux/reservation.h
+++ b/include/linux/reservation.h
@@ -272,6 +272,8 @@ void reservation_object_add_shared_fence(struct reservation_object *obj,
 void reservation_object_add_excl_fence(struct reservation_object *obj,
 				       struct dma_fence *fence);
 
+int reservation_object_prune_fences(struct reservation_object *obj, gfp_t gfp);
+
 int reservation_object_get_fences_rcu(struct reservation_object *obj,
 				      struct dma_fence **pfence_excl,
 				      unsigned *pshared_count,
-- 
2.15.1