drm/i915: Serialise concurrent calls to i915_gem_set_wedged()

Make i915_gem_set_wedged() and i915_gem_unset_wedged() behaviour more
consistent if called concurrently, and only do the wedging and reporting
once, curtailing any possible race where we start unwedging in the middle
of a wedge.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190114210408.4561-2-chris@chris-wilson.co.uk

drivers/gpu/drm/i915/i915_gem.c

@@ -3187,10 +3187,15 @@ static void nop_submit_request(struct i915_request *request)
 
 void i915_gem_set_wedged(struct drm_i915_private *i915)
 {
+	struct i915_gpu_error *error = &i915->gpu_error;
 	struct intel_engine_cs *engine;
 	enum intel_engine_id id;
 
-	GEM_TRACE("start\n");
+	mutex_lock(&error->wedge_mutex);
+	if (test_bit(I915_WEDGED, &error->flags)) {
+		mutex_unlock(&error->wedge_mutex);
+		return;
+	}
 
 	if (GEM_SHOW_DEBUG() && !intel_engines_are_idle(i915)) {
 		struct drm_printer p = drm_debug_printer(__func__);
@@ -3199,8 +3204,7 @@ void i915_gem_set_wedged(struct drm_i915_private *i915)
 			intel_engine_dump(engine, &p, "%s\n", engine->name);
 	}
 
-	if (test_and_set_bit(I915_WEDGED, &i915->gpu_error.flags))
-		goto out;
+	GEM_TRACE("start\n");
 
 	/*
 	 * First, stop submission to hw, but do not yet complete requests by
@@ -3236,23 +3240,31 @@ void i915_gem_set_wedged(struct drm_i915_private *i915)
 		intel_engine_wakeup(engine);
 	}
 
-out:
+	smp_mb__before_atomic();
+	set_bit(I915_WEDGED, &error->flags);
+
 	GEM_TRACE("end\n");
+	mutex_unlock(&error->wedge_mutex);
 
-	wake_up_all(&i915->gpu_error.reset_queue);
+	wake_up_all(&error->reset_queue);
 }
 
 bool i915_gem_unset_wedged(struct drm_i915_private *i915)
 {
+	struct i915_gpu_error *error = &i915->gpu_error;
 	struct i915_timeline *tl;
+	bool ret = false;
 
 	lockdep_assert_held(&i915->drm.struct_mutex);
-	if (!test_bit(I915_WEDGED, &i915->gpu_error.flags))
+
+	if (!test_bit(I915_WEDGED, &error->flags))
 		return true;
 
 	if (!i915->gt.scratch) /* Never full initialised, recovery impossible */
 		return false;
 
+	mutex_lock(&error->wedge_mutex);
+
 	GEM_TRACE("start\n");
 
 	/*
@@ -3286,7 +3298,7 @@ bool i915_gem_unset_wedged(struct drm_i915_private *i915)
 		 */
 		if (dma_fence_default_wait(&rq->fence, true,
 					   MAX_SCHEDULE_TIMEOUT) < 0)
-			return false;
+			goto unlock;
 	}
 	i915_retire_requests(i915);
 	GEM_BUG_ON(i915->gt.active_requests);
@@ -3309,8 +3321,11 @@ bool i915_gem_unset_wedged(struct drm_i915_private *i915)
 
 	smp_mb__before_atomic(); /* complete takeover before enabling execbuf */
 	clear_bit(I915_WEDGED, &i915->gpu_error.flags);
+	ret = true;
+unlock:
+	mutex_unlock(&i915->gpu_error.wedge_mutex);
 
-	return true;
+	return ret;
 }
 
 static void
@@ -5706,6 +5721,7 @@ int i915_gem_init_early(struct drm_i915_private *dev_priv)
 			  i915_gem_idle_work_handler);
 	init_waitqueue_head(&dev_priv->gpu_error.wait_queue);
 	init_waitqueue_head(&dev_priv->gpu_error.reset_queue);
+	mutex_init(&dev_priv->gpu_error.wedge_mutex);
 
 	atomic_set(&dev_priv->mm.bsd_engine_dispatch_index, 0);
 

drivers/gpu/drm/i915/i915_gpu_error.h

@@ -271,8 +271,8 @@ struct i915_gpu_error {
 #define I915_RESET_BACKOFF	0
 #define I915_RESET_HANDOFF	1
 #define I915_RESET_MODESET	2
+#define I915_RESET_ENGINE	3
 #define I915_WEDGED		(BITS_PER_LONG - 1)
-#define I915_RESET_ENGINE	(I915_WEDGED - I915_NUM_ENGINES)
 
 	/** Number of times an engine has been reset */
 	u32 reset_engine_count[I915_NUM_ENGINES];
@@ -283,6 +283,8 @@ struct i915_gpu_error {
 	/** Reason for the current *global* reset */
 	const char *reason;
 
+	struct mutex wedge_mutex; /* serialises wedging/unwedging */
+
 	/**
 	 * Waitqueue to signal when a hang is detected. Used to for waiters
 	 * to release the struct_mutex for the reset to procede.

drivers/gpu/drm/i915/selftests/mock_gem_device.c

@@ -188,6 +188,7 @@ struct drm_i915_private *mock_gem_device(void)
 
 	init_waitqueue_head(&i915->gpu_error.wait_queue);
 	init_waitqueue_head(&i915->gpu_error.reset_queue);
+	mutex_init(&i915->gpu_error.wedge_mutex);
 
 	i915->wq = alloc_ordered_workqueue("mock", 0);
 	if (!i915->wq)