[Intel-gfx] [PATCH] drm/i915: Skip repeated calls to i915_gem_set_wedged()

[Chris Wilson]

If we already wedged, i915_gem_set_wedged() becomes a complicated no-op.

v2: Make sure the double set-wedged is synchronous, a parallel call
should not return before the driver is indeed wedged.
v3: Use a real mutex (and fingers crossed that lockdep is ok!)
v4: Fix incorrect conversion to return false if unwedged.

References: https://bugs.freedesktop.org/show_bug.cgi?id=107343
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala at linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem.c               | 34 +++++++++++++------
 drivers/gpu/drm/i915/i915_gpu_error.h         |  4 ++-
 .../gpu/drm/i915/selftests/mock_gem_device.c  |  1 +
 3 files changed, 28 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 0453eb42a1a3..98a3e2601ab4 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3300,10 +3300,15 @@ static void nop_complete_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()) {
 		struct drm_printer p = drm_debug_printer(__func__);
@@ -3312,8 +3317,7 @@ 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
@@ -3373,20 +3377,28 @@ void i915_gem_set_wedged(struct drm_i915_private *i915)
 		i915_gem_reset_finish_engine(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;
 
+	mutex_lock(&error->wedge_mutex);
+
 	GEM_TRACE("start\n");
 
 	/*
@@ -3420,7 +3432,7 @@ 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);
@@ -3439,10 +3451,11 @@ bool i915_gem_unset_wedged(struct drm_i915_private *i915)
 
 	GEM_TRACE("end\n");
 
-	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
@@ -5764,6 +5777,7 @@ 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);
 
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h b/drivers/gpu/drm/i915/i915_gpu_error.h
index f893a4e8b783..49950b1231cf 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.h
+++ b/drivers/gpu/drm/i915/i915_gpu_error.h
@@ -267,8 +267,8 @@ 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];
@@ -279,6 +279,8 @@ 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.
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index 43ed8b28aeaa..6beb1f47e988 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -186,6 +186,7 @@ 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)
-- 
2.18.0