drm/i915: Store the reset counter when constructing a request

As the request is only valid during the same global reset epoch, we can
record the current reset_counter when constructing the request and reuse
it when waiting upon that request in future. This removes a very hairy
atomic check serialised by the struct_mutex at the time of waiting and
allows us to transfer those waits to a central dispatcher for all
waiters and all requests.

PS: With per-engine resets, we obviously cannot assume a global reset
epoch for the requests - a per-engine epoch makes the most sense. The
challenge then is how to handle checking in the waiter for when to break
the wait, as the fine-grained reset may also want to requeue the
request (i.e. the assumption that just because the epoch changes the
request is completed may be broken - or we just avoid breaking that
assumption with the fine-grained resets).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1460565315-7748-7-git-send-email-chris@chris-wilson.co.uk

drivers/gpu/drm/i915/i915_drv.h

@@ -2250,6 +2250,7 @@ struct drm_i915_gem_request {
 	/** On Which ring this request was generated */
 	struct drm_i915_private *i915;
 	struct intel_engine_cs *engine;
+	unsigned reset_counter;
 
 	 /** GEM sequence number associated with the previous request,
 	  * when the HWS breadcrumb is equal to this the GPU is processing
@@ -3160,7 +3161,6 @@ void __i915_add_request(struct drm_i915_gem_request *req,
 #define i915_add_request_no_flush(req) \
 	__i915_add_request(req, NULL, false)
 int __i915_wait_request(struct drm_i915_gem_request *req,
-			unsigned reset_counter,
 			bool interruptible,
 			s64 *timeout,
 			struct intel_rps_client *rps);

drivers/gpu/drm/i915/i915_gem.c

@@ -1213,7 +1213,6 @@ static int __i915_spin_request(struct drm_i915_gem_request *req, int state)
 /**
  * __i915_wait_request - wait until execution of request has finished
  * @req: duh!
- * @reset_counter: reset sequence associated with the given request
  * @interruptible: do an interruptible wait (normally yes)
  * @timeout: in - how long to wait (NULL forever); out - how much time remaining
  *
@@ -1228,7 +1227,6 @@ static int __i915_spin_request(struct drm_i915_gem_request *req, int state)
  * errno with remaining time filled in timeout argument.
  */
 int __i915_wait_request(struct drm_i915_gem_request *req,
-			unsigned reset_counter,
 			bool interruptible,
 			s64 *timeout,
 			struct intel_rps_client *rps)
@@ -1290,7 +1288,7 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
 
 		/* We need to check whether any gpu reset happened in between
 		 * the caller grabbing the seqno and now ... */
-		if (reset_counter != i915_reset_counter(&dev_priv->gpu_error)) {
+		if (req->reset_counter != i915_reset_counter(&dev_priv->gpu_error)) {
 			/* ... but upgrade the -EAGAIN to an -EIO if the gpu
 			 * is truely gone. */
 			ret = i915_gem_check_wedge(&dev_priv->gpu_error, interruptible);
@@ -1460,13 +1458,7 @@ i915_wait_request(struct drm_i915_gem_request *req)
 
 	BUG_ON(!mutex_is_locked(&dev->struct_mutex));
 
-	ret = i915_gem_check_wedge(&dev_priv->gpu_error, interruptible);
-	if (ret)
-		return ret;
-
-	ret = __i915_wait_request(req,
-				  i915_reset_counter(&dev_priv->gpu_error),
-				  interruptible, NULL, NULL);
+	ret = __i915_wait_request(req, interruptible, NULL, NULL);
 	if (ret)
 		return ret;
 
@@ -1541,7 +1533,6 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
 	struct drm_device *dev = obj->base.dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_gem_request *requests[I915_NUM_ENGINES];
-	unsigned reset_counter;
 	int ret, i, n = 0;
 
 	BUG_ON(!mutex_is_locked(&dev->struct_mutex));
@@ -1550,12 +1541,6 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
 	if (!obj->active)
 		return 0;
 
-	ret = i915_gem_check_wedge(&dev_priv->gpu_error, true);
-	if (ret)
-		return ret;
-
-	reset_counter = i915_reset_counter(&dev_priv->gpu_error);
-
 	if (readonly) {
 		struct drm_i915_gem_request *req;
 
@@ -1577,9 +1562,9 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
 	}
 
 	mutex_unlock(&dev->struct_mutex);
+	ret = 0;
 	for (i = 0; ret == 0 && i < n; i++)
-		ret = __i915_wait_request(requests[i], reset_counter, true,
-					  NULL, rps);
+		ret = __i915_wait_request(requests[i], true, NULL, rps);
 	mutex_lock(&dev->struct_mutex);
 
 	for (i = 0; i < n; i++) {
@@ -2735,6 +2720,7 @@ __i915_gem_request_alloc(struct intel_engine_cs *engine,
 			 struct drm_i915_gem_request **req_out)
 {
 	struct drm_i915_private *dev_priv = to_i915(engine->dev);
+	unsigned reset_counter = i915_reset_counter(&dev_priv->gpu_error);
 	struct drm_i915_gem_request *req;
 	int ret;
 
@@ -2743,6 +2729,11 @@ __i915_gem_request_alloc(struct intel_engine_cs *engine,
 
 	*req_out = NULL;
 
+	ret = i915_gem_check_wedge(&dev_priv->gpu_error,
+				   dev_priv->mm.interruptible);
+	if (ret)
+		return ret;
+
 	req = kmem_cache_zalloc(dev_priv->requests, GFP_KERNEL);
 	if (req == NULL)
 		return -ENOMEM;
@@ -2754,6 +2745,7 @@ __i915_gem_request_alloc(struct intel_engine_cs *engine,
 	kref_init(&req->ref);
 	req->i915 = dev_priv;
 	req->engine = engine;
+	req->reset_counter = reset_counter;
 	req->ctx  = ctx;
 	i915_gem_context_reference(req->ctx);
 
@@ -3132,11 +3124,9 @@ retire:
 int
 i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 {
-	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_gem_wait *args = data;
 	struct drm_i915_gem_object *obj;
 	struct drm_i915_gem_request *req[I915_NUM_ENGINES];
-	unsigned reset_counter;
 	int i, n = 0;
 	int ret;
 
@@ -3170,7 +3160,6 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 	}
 
 	drm_gem_object_unreference(&obj->base);
-	reset_counter = i915_reset_counter(&dev_priv->gpu_error);
 
 	for (i = 0; i < I915_NUM_ENGINES; i++) {
 		if (obj->last_read_req[i] == NULL)
@@ -3183,7 +3172,7 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 
 	for (i = 0; i < n; i++) {
 		if (ret == 0)
-			ret = __i915_wait_request(req[i], reset_counter, true,
+			ret = __i915_wait_request(req[i], true,
 						  args->timeout_ns > 0 ? &args->timeout_ns : NULL,
 						  to_rps_client(file));
 		i915_gem_request_unreference__unlocked(req[i]);
@@ -3215,7 +3204,6 @@ __i915_gem_object_sync(struct drm_i915_gem_object *obj,
 	if (!i915_semaphore_is_enabled(obj->base.dev)) {
 		struct drm_i915_private *i915 = to_i915(obj->base.dev);
 		ret = __i915_wait_request(from_req,
-					  i915_reset_counter(&i915->gpu_error),
 					  i915->mm.interruptible,
 					  NULL,
 					  &i915->rps.semaphores);
@@ -4171,7 +4159,6 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
 	struct drm_i915_file_private *file_priv = file->driver_priv;
 	unsigned long recent_enough = jiffies - DRM_I915_THROTTLE_JIFFIES;
 	struct drm_i915_gem_request *request, *target = NULL;
-	unsigned reset_counter;
 	int ret;
 
 	ret = i915_gem_wait_for_error(&dev_priv->gpu_error);
@@ -4196,7 +4183,6 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
 
 		target = request;
 	}
-	reset_counter = i915_reset_counter(&dev_priv->gpu_error);
 	if (target)
 		i915_gem_request_reference(target);
 	spin_unlock(&file_priv->mm.lock);
@@ -4204,7 +4190,7 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
 	if (target == NULL)
 		return 0;
 
-	ret = __i915_wait_request(target, reset_counter, true, NULL, NULL);
+	ret = __i915_wait_request(target, true, NULL, NULL);
 	if (ret == 0)
 		queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work, 0);
 

drivers/gpu/drm/i915/i915_gem_userptr.c

@@ -65,7 +65,6 @@ static void wait_rendering(struct drm_i915_gem_object *obj)
 {
 	struct drm_device *dev = obj->base.dev;
 	struct drm_i915_gem_request *requests[I915_NUM_ENGINES];
-	unsigned reset_counter;
 	int i, n;
 
 	if (!obj->active)
@@ -82,12 +81,10 @@ static void wait_rendering(struct drm_i915_gem_object *obj)
 		requests[n++] = i915_gem_request_reference(req);
 	}
 
-	reset_counter = atomic_read(&to_i915(dev)->gpu_error.reset_counter);
 	mutex_unlock(&dev->struct_mutex);
 
 	for (i = 0; i < n; i++)
-		__i915_wait_request(requests[i], reset_counter, false,
-				    NULL, NULL);
+		__i915_wait_request(requests[i], false, NULL, NULL);
 
 	mutex_lock(&dev->struct_mutex);
 

drivers/gpu/drm/i915/intel_display.c

@@ -11365,7 +11365,6 @@ static void intel_mmio_flip_work_func(struct work_struct *work)
 
 	if (mmio_flip->req) {
 		WARN_ON(__i915_wait_request(mmio_flip->req,
-					    mmio_flip->crtc->reset_counter,
 					    false, NULL,
 					    &mmio_flip->i915->rps.mmioflips));
 		i915_gem_request_unreference__unlocked(mmio_flip->req);
@@ -13426,9 +13425,6 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
 
 	ret = drm_atomic_helper_prepare_planes(dev, state);
 	if (!ret && !async && !i915_reset_in_progress_or_wedged(&dev_priv->gpu_error)) {
-		u32 reset_counter;
-
-		reset_counter = i915_reset_counter(&dev_priv->gpu_error);
 		mutex_unlock(&dev->struct_mutex);
 
 		for_each_plane_in_state(state, plane, plane_state, i) {
@@ -13439,8 +13435,7 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
 				continue;
 
 			ret = __i915_wait_request(intel_plane_state->wait_req,
-						  reset_counter, true,
-						  NULL, NULL);
+						  true, NULL, NULL);
 
 			/* Swallow -EIO errors to allow updates during hw lockup. */
 			if (ret == -EIO)

drivers/gpu/drm/i915/intel_lrc.c

@@ -890,16 +890,9 @@ static int logical_ring_prepare(struct drm_i915_gem_request *req, int bytes)
  */
 int intel_logical_ring_begin(struct drm_i915_gem_request *req, int num_dwords)
 {
-	struct drm_i915_private *dev_priv;
 	int ret;
 
 	WARN_ON(req == NULL);
-	dev_priv = req->i915;
-
-	ret = i915_gem_check_wedge(&dev_priv->gpu_error,
-				   dev_priv->mm.interruptible);
-	if (ret)
-		return ret;
 
 	ret = logical_ring_prepare(req, num_dwords * sizeof(uint32_t));
 	if (ret)

drivers/gpu/drm/i915/intel_ringbuffer.c

@@ -2364,7 +2364,6 @@ int intel_engine_idle(struct intel_engine_cs *engine)
 
 	/* Make sure we do not trigger any retires */
 	return __i915_wait_request(req,
-				   i915_reset_counter(&req->i915->gpu_error),
 				   req->i915->mm.interruptible,
 				   NULL, NULL);
 }
@@ -2495,11 +2494,6 @@ int intel_ring_begin(struct drm_i915_gem_request *req,
 	engine = req->engine;
 	dev_priv = req->i915;
 
-	ret = i915_gem_check_wedge(&dev_priv->gpu_error,
-				   dev_priv->mm.interruptible);
-	if (ret)
-		return ret;
-
 	ret = __intel_ring_prepare(engine, num_dwords * sizeof(uint32_t));
 	if (ret)
 		return ret;