drm/i915: Prepare object synchronisation for asynchronicity

We are about to specialize object synchronisation to enable nonblocking
execbuf submission. First we make a copy of the current object
synchronisation for execbuffer. The general i915_gem_object_sync() will
be removed following the removal of CS flips in the near future.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: John Harrison <john.c.harrison@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-16-chris@chris-wilson.co.uk

drivers/gpu/drm/i915/i915_drv.h

@@ -3221,8 +3221,6 @@ i915_gem_obj_finish_shmem_access(struct drm_i915_gem_object *obj)
 }
 
 int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
-int i915_gem_object_sync(struct drm_i915_gem_object *obj,
-			 struct drm_i915_gem_request *to);
 void i915_vma_move_to_active(struct i915_vma *vma,
 			     struct drm_i915_gem_request *req,
 			     unsigned int flags);

drivers/gpu/drm/i915/i915_gem.c

@@ -2818,97 +2818,6 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 	return ret;
 }
 
-static int
-__i915_gem_object_sync(struct drm_i915_gem_request *to,
-		       struct drm_i915_gem_request *from)
-{
-	int ret;
-
-	if (to->engine == from->engine)
-		return 0;
-
-	if (!i915.semaphores) {
-		ret = i915_wait_request(from,
-					from->i915->mm.interruptible |
-					I915_WAIT_LOCKED,
-					NULL,
-					NO_WAITBOOST);
-		if (ret)
-			return ret;
-	} else {
-		int idx = intel_engine_sync_index(from->engine, to->engine);
-		if (from->fence.seqno <= from->engine->semaphore.sync_seqno[idx])
-			return 0;
-
-		trace_i915_gem_ring_sync_to(to, from);
-		ret = to->engine->semaphore.sync_to(to, from);
-		if (ret)
-			return ret;
-
-		from->engine->semaphore.sync_seqno[idx] = from->fence.seqno;
-	}
-
-	return 0;
-}
-
-/**
- * i915_gem_object_sync - sync an object to a ring.
- *
- * @obj: object which may be in use on another ring.
- * @to: request we are wishing to use
- *
- * This code is meant to abstract object synchronization with the GPU.
- * Conceptually we serialise writes between engines inside the GPU.
- * We only allow one engine to write into a buffer at any time, but
- * multiple readers. To ensure each has a coherent view of memory, we must:
- *
- * - If there is an outstanding write request to the object, the new
- *   request must wait for it to complete (either CPU or in hw, requests
- *   on the same ring will be naturally ordered).
- *
- * - If we are a write request (pending_write_domain is set), the new
- *   request must wait for outstanding read requests to complete.
- *
- * Returns 0 if successful, else propagates up the lower layer error.
- */
-int
-i915_gem_object_sync(struct drm_i915_gem_object *obj,
-		     struct drm_i915_gem_request *to)
-{
-	struct i915_gem_active *active;
-	unsigned long active_mask;
-	int idx;
-
-	lockdep_assert_held(&obj->base.dev->struct_mutex);
-
-	active_mask = i915_gem_object_get_active(obj);
-	if (!active_mask)
-		return 0;
-
-	if (obj->base.pending_write_domain) {
-		active = obj->last_read;
-	} else {
-		active_mask = 1;
-		active = &obj->last_write;
-	}
-
-	for_each_active(active_mask, idx) {
-		struct drm_i915_gem_request *request;
-		int ret;
-
-		request = i915_gem_active_peek(&active[idx],
-					       &obj->base.dev->struct_mutex);
-		if (!request)
-			continue;
-
-		ret = __i915_gem_object_sync(to, request);
-		if (ret)
-			return ret;
-	}
-
-	return 0;
-}
-
 static void __i915_vma_iounmap(struct i915_vma *vma)
 {
 	GEM_BUG_ON(i915_vma_is_pinned(vma));

drivers/gpu/drm/i915/i915_gem_execbuffer.c

@@ -1133,7 +1133,8 @@ i915_gem_execbuffer_move_to_gpu(struct drm_i915_gem_request *req,
 		struct drm_i915_gem_object *obj = vma->obj;
 
 		if (obj->flags & other_rings) {
-			ret = i915_gem_object_sync(obj, req);
+			ret = i915_gem_request_await_object
+				(req, obj, obj->base.pending_write_domain);
 			if (ret)
 				return ret;
 		}

drivers/gpu/drm/i915/i915_gem_request.c

@@ -460,6 +460,93 @@ i915_gem_request_alloc(struct intel_engine_cs *engine,
 	return ERR_PTR(ret);
 }
 
+static int
+i915_gem_request_await_request(struct drm_i915_gem_request *to,
+			       struct drm_i915_gem_request *from)
+{
+	int idx, ret;
+
+	GEM_BUG_ON(to == from);
+
+	if (to->engine == from->engine)
+		return 0;
+
+	idx = intel_engine_sync_index(from->engine, to->engine);
+	if (from->fence.seqno <= from->engine->semaphore.sync_seqno[idx])
+		return 0;
+
+	trace_i915_gem_ring_sync_to(to, from);
+	if (!i915.semaphores) {
+		ret = i915_wait_request(from,
+					I915_WAIT_INTERRUPTIBLE |
+					I915_WAIT_LOCKED,
+					NULL, NO_WAITBOOST);
+		if (ret)
+			return ret;
+	} else {
+		ret = to->engine->semaphore.sync_to(to, from);
+		if (ret)
+			return ret;
+	}
+
+	from->engine->semaphore.sync_seqno[idx] = from->fence.seqno;
+	return 0;
+}
+
+/**
+ * i915_gem_request_await_object - set this request to (async) wait upon a bo
+ *
+ * @to: request we are wishing to use
+ * @obj: object which may be in use on another ring.
+ *
+ * This code is meant to abstract object synchronization with the GPU.
+ * Conceptually we serialise writes between engines inside the GPU.
+ * We only allow one engine to write into a buffer at any time, but
+ * multiple readers. To ensure each has a coherent view of memory, we must:
+ *
+ * - If there is an outstanding write request to the object, the new
+ *   request must wait for it to complete (either CPU or in hw, requests
+ *   on the same ring will be naturally ordered).
+ *
+ * - If we are a write request (pending_write_domain is set), the new
+ *   request must wait for outstanding read requests to complete.
+ *
+ * Returns 0 if successful, else propagates up the lower layer error.
+ */
+int
+i915_gem_request_await_object(struct drm_i915_gem_request *to,
+			      struct drm_i915_gem_object *obj,
+			      bool write)
+{
+	struct i915_gem_active *active;
+	unsigned long active_mask;
+	int idx;
+
+	if (write) {
+		active_mask = i915_gem_object_get_active(obj);
+		active = obj->last_read;
+	} else {
+		active_mask = 1;
+		active = &obj->last_write;
+	}
+
+	for_each_active(active_mask, idx) {
+		struct drm_i915_gem_request *request;
+		int ret;
+
+		request = i915_gem_active_peek(&active[idx],
+					       &obj->base.dev->struct_mutex);
+		if (!request)
+			continue;
+
+		ret = i915_gem_request_await_request(to, request);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
 static void i915_gem_mark_busy(const struct intel_engine_cs *engine)
 {
 	struct drm_i915_private *dev_priv = engine->i915;

drivers/gpu/drm/i915/i915_gem_request.h

@@ -209,6 +209,11 @@ static inline void i915_gem_request_assign(struct drm_i915_gem_request **pdst,
 	*pdst = src;
 }
 
+int
+i915_gem_request_await_object(struct drm_i915_gem_request *to,
+			      struct drm_i915_gem_object *obj,
+			      bool write);
+
 void __i915_add_request(struct drm_i915_gem_request *req, bool flush_caches);
 #define i915_add_request(req) \
 	__i915_add_request(req, true)

drivers/gpu/drm/i915/intel_display.c

@@ -12273,7 +12273,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
 			goto cleanup_unpin;
 		}
 
-		ret = i915_gem_object_sync(obj, request);
+		ret = i915_gem_request_await_object(request, obj, false);
 		if (ret)
 			goto cleanup_request;