drm/i915: Wait for object idle without locks in atomic_commit, v2.

Make pinning and waiting a separate step, and wait for object idle
without struct_mutex held.

Changes since v1:
- Do not wait when a reset is in progress.
- Remove call to i915_gem_object_wait_rendering for
  intel_overlay_do_put_image (Chris Wilson)

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drivers/gpu/drm/i915/i915_drv.h

@@ -3017,8 +3017,6 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write);
 int __must_check
 i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
 				     u32 alignment,
-				     struct intel_engine_cs *pipelined,
-				     struct drm_i915_gem_request **pipelined_request,
 				     const struct i915_ggtt_view *view);
 void i915_gem_object_unpin_from_display_plane(struct drm_i915_gem_object *obj,
 					      const struct i915_ggtt_view *view);

drivers/gpu/drm/i915/i915_gem.c

@@ -3883,17 +3883,11 @@ int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
 int
 i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
 				     u32 alignment,
-				     struct intel_engine_cs *pipelined,
-				     struct drm_i915_gem_request **pipelined_request,
 				     const struct i915_ggtt_view *view)
 {
 	u32 old_read_domains, old_write_domain;
 	int ret;
 
-	ret = i915_gem_object_sync(obj, pipelined, pipelined_request);
-	if (ret)
-		return ret;
-
 	/* Mark the pin_display early so that we account for the
 	 * display coherency whilst setting up the cache domains.
 	 */

drivers/gpu/drm/i915/intel_atomic_plane.c

@@ -84,6 +84,7 @@ intel_plane_duplicate_state(struct drm_plane *plane)
 	state = &intel_state->base;
 
 	__drm_atomic_helper_plane_duplicate_state(plane, state);
+	intel_state->wait_req = NULL;
 
 	return state;
 }
@@ -100,6 +101,7 @@ void
 intel_plane_destroy_state(struct drm_plane *plane,
 			  struct drm_plane_state *state)
 {
+	WARN_ON(state && to_intel_plane_state(state)->wait_req);
 	drm_atomic_helper_plane_destroy_state(plane, state);
 }
 

drivers/gpu/drm/i915/intel_display.c

@@ -2320,9 +2320,7 @@ static unsigned int intel_linear_alignment(struct drm_i915_private *dev_priv)
 int
 intel_pin_and_fence_fb_obj(struct drm_plane *plane,
 			   struct drm_framebuffer *fb,
-			   const struct drm_plane_state *plane_state,
+			   const struct drm_plane_state *plane_state)
-			   struct intel_engine_cs *pipelined,
-			   struct drm_i915_gem_request **pipelined_request)
 {
 	struct drm_device *dev = fb->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -2378,8 +2376,8 @@ intel_pin_and_fence_fb_obj(struct drm_plane *plane,
 	 */
 	intel_runtime_pm_get(dev_priv);
 
-	ret = i915_gem_object_pin_to_display_plane(obj, alignment, pipelined,
+	ret = i915_gem_object_pin_to_display_plane(obj, alignment,
-						   pipelined_request, &view);
+						   &view);
 	if (ret)
 		goto err_pm;
 
@@ -11426,9 +11424,14 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
 	 * synchronisation, so all we want here is to pin the framebuffer
 	 * into the display plane and skip any waits.
 	 */
+	if (!mmio_flip) {
+		ret = i915_gem_object_sync(obj, ring, &request);
+		if (ret)
+			goto cleanup_pending;
+	}
+
 	ret = intel_pin_and_fence_fb_obj(crtc->primary, fb,
-					 crtc->primary->state,
+					 crtc->primary->state);
-					 mmio_flip ? i915_gem_request_get_ring(obj->last_write_req) : ring, &request);
 	if (ret)
 		goto cleanup_pending;
 
@@ -13150,7 +13153,10 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
 				       struct drm_atomic_state *state,
 				       bool async)
 {
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_plane_state *plane_state;
 	struct drm_crtc_state *crtc_state;
+	struct drm_plane *plane;
 	struct drm_crtc *crtc;
 	int i, ret;
 
@@ -13163,6 +13169,9 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
 		ret = intel_crtc_wait_for_pending_flips(crtc);
 		if (ret)
 			return ret;
+
+		if (atomic_read(&to_intel_crtc(crtc)->unpin_work_count) >= 2)
+			flush_workqueue(dev_priv->wq);
 	}
 
 	ret = mutex_lock_interruptible(&dev->struct_mutex);
@@ -13170,6 +13179,37 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
 		return ret;
 
 	ret = drm_atomic_helper_prepare_planes(dev, state);
+	if (!ret && !async && !i915_reset_in_progress(&dev_priv->gpu_error)) {
+		u32 reset_counter;
+
+		reset_counter = atomic_read(&dev_priv->gpu_error.reset_counter);
+		mutex_unlock(&dev->struct_mutex);
+
+		for_each_plane_in_state(state, plane, plane_state, i) {
+			struct intel_plane_state *intel_plane_state =
+				to_intel_plane_state(plane_state);
+
+			if (!intel_plane_state->wait_req)
+				continue;
+
+			ret = __i915_wait_request(intel_plane_state->wait_req,
+						  reset_counter, true,
+						  NULL, NULL);
+
+			/* Swallow -EIO errors to allow updates during hw lockup. */
+			if (ret == -EIO)
+				ret = 0;
+
+			if (ret)
+				break;
+		}
+
+		if (!ret)
+			return 0;
+
+		mutex_lock(&dev->struct_mutex);
+		drm_atomic_helper_cleanup_planes(dev, state);
+	}
 
 	mutex_unlock(&dev->struct_mutex);
 	return ret;
@@ -13196,15 +13236,17 @@ static int intel_atomic_commit(struct drm_device *dev,
 			       bool async)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	struct drm_crtc *crtc;
 	struct drm_crtc_state *crtc_state;
+	struct drm_crtc *crtc;
 	int ret = 0;
 	int i;
 	bool any_ms = false;
 
 	ret = intel_atomic_prepare_commit(dev, state, async);
-	if (ret)
+	if (ret) {
+		DRM_DEBUG_ATOMIC("Preparing state failed with %i\n", ret);
 		return ret;
+	}
 
 	drm_atomic_helper_swap_state(dev, state);
 	dev_priv->wm.config = to_intel_atomic_state(state)->wm_config;
@@ -13495,11 +13537,20 @@ intel_prepare_plane_fb(struct drm_plane *plane,
 		if (ret)
 			DRM_DEBUG_KMS("failed to attach phys object\n");
 	} else {
-		ret = intel_pin_and_fence_fb_obj(plane, fb, new_state, NULL, NULL);
+		ret = intel_pin_and_fence_fb_obj(plane, fb, new_state);
 	}
 
-	if (ret == 0)
+	if (ret == 0) {
+		if (obj) {
+			struct intel_plane_state *plane_state =
+				to_intel_plane_state(new_state);
+
+			i915_gem_request_assign(&plane_state->wait_req,
+						obj->last_write_req);
+		}
+
 		i915_gem_track_fb(old_obj, obj, intel_plane->frontbuffer_bit);
+	}
 
 	return ret;
 }
@@ -13519,9 +13570,12 @@ intel_cleanup_plane_fb(struct drm_plane *plane,
 {
 	struct drm_device *dev = plane->dev;
 	struct intel_plane *intel_plane = to_intel_plane(plane);
+	struct intel_plane_state *old_intel_state;
 	struct drm_i915_gem_object *old_obj = intel_fb_obj(old_state->fb);
 	struct drm_i915_gem_object *obj = intel_fb_obj(plane->state->fb);
 
+	old_intel_state = to_intel_plane_state(old_state);
+
 	if (!obj && !old_obj)
 		return;
 
@@ -13533,6 +13587,9 @@ intel_cleanup_plane_fb(struct drm_plane *plane,
 	if ((old_obj && (old_obj->frontbuffer_bits & intel_plane->frontbuffer_bit)) ||
 	    (obj && !(obj->frontbuffer_bits & intel_plane->frontbuffer_bit)))
 		i915_gem_track_fb(old_obj, obj, intel_plane->frontbuffer_bit);
+
+	i915_gem_request_assign(&old_intel_state->wait_req, NULL);
+
 }
 
 int
@@ -15498,8 +15555,7 @@ void intel_modeset_gem_init(struct drm_device *dev)
 		mutex_lock(&dev->struct_mutex);
 		ret = intel_pin_and_fence_fb_obj(c->primary,
 						 c->primary->fb,
-						 c->primary->state,
+						 c->primary->state);
-						 NULL, NULL);
 		mutex_unlock(&dev->struct_mutex);
 		if (ret) {
 			DRM_ERROR("failed to pin boot fb on pipe %d\n",

drivers/gpu/drm/i915/intel_drv.h

@@ -281,6 +281,9 @@ struct intel_plane_state {
 	int scaler_id;
 
 	struct drm_intel_sprite_colorkey ckey;
+
+	/* async flip related structures */
+	struct drm_i915_gem_request *wait_req;
 };
 
 struct intel_initial_plane_config {
@@ -1084,9 +1087,7 @@ void intel_release_load_detect_pipe(struct drm_connector *connector,
 				    struct drm_modeset_acquire_ctx *ctx);
 int intel_pin_and_fence_fb_obj(struct drm_plane *plane,
 			       struct drm_framebuffer *fb,
-			       const struct drm_plane_state *plane_state,
+			       const struct drm_plane_state *plane_state);
-			       struct intel_engine_cs *pipelined,
-			       struct drm_i915_gem_request **pipelined_request);
 struct drm_framebuffer *
 __intel_framebuffer_create(struct drm_device *dev,
 			   struct drm_mode_fb_cmd2 *mode_cmd,

drivers/gpu/drm/i915/intel_fbdev.c

@@ -161,7 +161,7 @@ static int intelfb_alloc(struct drm_fb_helper *helper,
 	}
 
 	/* Flush everything out, we'll be doing GTT only from now on */
-	ret = intel_pin_and_fence_fb_obj(NULL, fb, NULL, NULL, NULL);
+	ret = intel_pin_and_fence_fb_obj(NULL, fb, NULL);
 	if (ret) {
 		DRM_ERROR("failed to pin obj: %d\n", ret);
 		goto out_fb;

drivers/gpu/drm/i915/intel_overlay.c

@@ -749,7 +749,7 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
 	if (ret != 0)
 		return ret;
 
-	ret = i915_gem_object_pin_to_display_plane(new_bo, 0, NULL, NULL,
+	ret = i915_gem_object_pin_to_display_plane(new_bo, 0,
 						   &i915_ggtt_view_normal);
 	if (ret != 0)
 		return ret;