drm/i915/debug: Convert i915_verify_active() to scan all lists

... and check more regularly.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

4 files changed, 113 insertions, 39 deletions

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index cb4e9a63c83..5ccf9809538 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -77,7 +77,7 @@ enum plane {
 #define WATCH_COHERENCY	0
 #define WATCH_EXEC	0
 #define WATCH_RELOC	0
-#define WATCH_INACTIVE	0
+#define WATCH_LISTS	0
 #define WATCH_PWRITE	0
 
 #define I915_GEM_PHYS_CURSOR_0 1
@@ -1079,10 +1079,10 @@ bool i915_gem_object_fence_offset_ok(struct drm_gem_object *obj,
 /* i915_gem_debug.c */
 void i915_gem_dump_object(struct drm_gem_object *obj, int len,
 			  const char *where, uint32_t mark);
-#if WATCH_INACTIVE
-void i915_verify_inactive(struct drm_device *dev, char *file, int line);
+#if WATCH_LISTS
+int i915_verify_lists(struct drm_device *dev);
 #else
-#define i915_verify_inactive(dev, file, line)
+#define i915_verify_lists(dev) 0
 #endif
 void i915_gem_object_check_coherency(struct drm_gem_object *obj, int handle);
 void i915_gem_dump_object(struct drm_gem_object *obj, int len,
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index fe1424c6c3f..c3a7065947c 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -109,6 +109,7 @@ static int i915_mutex_lock_interruptible(struct drm_device *dev)
 		return -EAGAIN;
 	}
 
+	WARN_ON(i915_verify_lists(dev));
 	return 0;
 }
 
@@ -1612,7 +1613,6 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_gem_object *obj_priv = to_intel_bo(obj);
 
-	i915_verify_inactive(dev, __FILE__, __LINE__);
 	if (obj_priv->pin_count != 0)
 		list_move_tail(&obj_priv->list, &dev_priv->mm.pinned_list);
 	else
@@ -1626,7 +1626,7 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj)
 		obj_priv->active = 0;
 		drm_gem_object_unreference(obj);
 	}
-	i915_verify_inactive(dev, __FILE__, __LINE__);
+	WARN_ON(i915_verify_lists(dev));
 }
 
 static void
@@ -1821,6 +1821,8 @@ i915_gem_retire_requests_ring(struct drm_device *dev,
 	    list_empty(&ring->request_list))
 		return;
 
+	WARN_ON(i915_verify_lists(dev));
+
 	seqno = ring->get_seqno(dev, ring);
 	while (!list_empty(&ring->request_list)) {
 		struct drm_i915_gem_request *request;
@@ -1865,6 +1867,8 @@ i915_gem_retire_requests_ring(struct drm_device *dev,
 		ring->user_irq_put(dev, ring);
 		dev_priv->trace_irq_seqno = 0;
 	}
+
+	WARN_ON(i915_verify_lists(dev));
 }
 
 void
@@ -3690,8 +3694,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	if (ret)
 		goto pre_mutex_err;
 
-	i915_verify_inactive(dev, __FILE__, __LINE__);
-
 	if (dev_priv->mm.suspended) {
 		mutex_unlock(&dev->struct_mutex);
 		ret = -EBUSY;
@@ -3811,8 +3813,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		goto err;
 	}
 
-	i915_verify_inactive(dev, __FILE__, __LINE__);
-
 	/* Zero the global flush/invalidate flags. These
 	 * will be modified as new domains are computed
 	 * for each object
@@ -3828,8 +3828,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		i915_gem_object_set_to_gpu_domain(obj);
 	}
 
-	i915_verify_inactive(dev, __FILE__, __LINE__);
-
 	if (dev->invalidate_domains | dev->flush_domains) {
 #if WATCH_EXEC
 		DRM_INFO("%s: invalidate_domains %08x flush_domains %08x\n",
@@ -3860,8 +3858,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 						    old_write_domain);
 	}
 
-	i915_verify_inactive(dev, __FILE__, __LINE__);
-
 #if WATCH_COHERENCY
 	for (i = 0; i < args->buffer_count; i++) {
 		i915_gem_object_check_coherency(object_list[i],
@@ -3890,8 +3886,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	 */
 	i915_retire_commands(dev, ring);
 
-	i915_verify_inactive(dev, __FILE__, __LINE__);
-
 	for (i = 0; i < args->buffer_count; i++) {
 		struct drm_gem_object *obj = object_list[i];
 		obj_priv = to_intel_bo(obj);
@@ -3902,8 +3896,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	i915_add_request(dev, file_priv, request, ring);
 	request = NULL;
 
-	i915_verify_inactive(dev, __FILE__, __LINE__);
-
 err:
 	for (i = 0; i < pinned; i++)
 		i915_gem_object_unpin(object_list[i]);
@@ -4094,8 +4086,7 @@ i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment)
 	int ret;
 
 	BUG_ON(obj_priv->pin_count == DRM_I915_GEM_OBJECT_MAX_PIN_COUNT);
-
-	i915_verify_inactive(dev, __FILE__, __LINE__);
+	WARN_ON(i915_verify_lists(dev));
 
 	if (obj_priv->gtt_space != NULL) {
 		if (alignment == 0)
@@ -4129,8 +4120,8 @@ i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment)
 			list_move_tail(&obj_priv->list,
 				       &dev_priv->mm.pinned_list);
 	}
-	i915_verify_inactive(dev, __FILE__, __LINE__);
 
+	WARN_ON(i915_verify_lists(dev));
 	return 0;
 }
 
@@ -4141,7 +4132,7 @@ i915_gem_object_unpin(struct drm_gem_object *obj)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_gem_object *obj_priv = to_intel_bo(obj);
 
-	i915_verify_inactive(dev, __FILE__, __LINE__);
+	WARN_ON(i915_verify_lists(dev));
 	obj_priv->pin_count--;
 	BUG_ON(obj_priv->pin_count < 0);
 	BUG_ON(obj_priv->gtt_space == NULL);
@@ -4157,7 +4148,7 @@ i915_gem_object_unpin(struct drm_gem_object *obj)
 		atomic_dec(&dev->pin_count);
 		atomic_sub(obj->size, &dev->pin_memory);
 	}
-	i915_verify_inactive(dev, __FILE__, __LINE__);
+	WARN_ON(i915_verify_lists(dev));
 }
 
 int
diff --git a/drivers/gpu/drm/i915/i915_gem_debug.c b/drivers/gpu/drm/i915/i915_gem_debug.c
index 2732c909a94..48644b840a8 100644
--- a/drivers/gpu/drm/i915/i915_gem_debug.c
+++ b/drivers/gpu/drm/i915/i915_gem_debug.c
@@ -30,24 +30,107 @@
 #include "i915_drm.h"
 #include "i915_drv.h"
 
-#if WATCH_INACTIVE
-void
-i915_verify_inactive(struct drm_device *dev, char *file, int line)
+#if WATCH_LISTS
+int
+i915_verify_lists(struct drm_device *dev)
 {
+	static int warned;
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	struct drm_gem_object *obj;
-	struct drm_i915_gem_object *obj_priv;
-
-	list_for_each_entry(obj_priv, &dev_priv->mm.inactive_list, list) {
-		obj = &obj_priv->base;
-		if (obj_priv->pin_count || obj_priv->active ||
-		    (obj->write_domain & ~(I915_GEM_DOMAIN_CPU |
-					   I915_GEM_DOMAIN_GTT)))
-			DRM_ERROR("inactive %p (p %d a %d w %x)  %s:%d\n",
+	struct drm_i915_gem_object *obj;
+	int err = 0;
+
+	if (warned)
+		return 0;
+
+	list_for_each_entry(obj, &dev_priv->render_ring.active_list, list) {
+		if (obj->base.dev != dev ||
+		    !atomic_read(&obj->base.refcount.refcount)) {
+			DRM_ERROR("freed render active %p\n", obj);
+			err++;
+			break;
+		} else if (!obj->active ||
+			   (obj->base.read_domains & I915_GEM_GPU_DOMAINS) == 0) {
+			DRM_ERROR("invalid render active %p (a %d r %x)\n",
+				  obj,
+				  obj->active,
+				  obj->base.read_domains);
+			err++;
+		} else if (obj->base.write_domain && list_empty(&obj->gpu_write_list)) {
+			DRM_ERROR("invalid render active %p (w %x, gwl %d)\n",
+				  obj,
+				  obj->base.write_domain,
+				  !list_empty(&obj->gpu_write_list));
+			err++;
+		}
+	}
+
+	list_for_each_entry(obj, &dev_priv->mm.flushing_list, list) {
+		if (obj->base.dev != dev ||
+		    !atomic_read(&obj->base.refcount.refcount)) {
+			DRM_ERROR("freed flushing %p\n", obj);
+			err++;
+			break;
+		} else if (!obj->active ||
+			   (obj->base.write_domain & I915_GEM_GPU_DOMAINS) == 0 ||
+			   list_empty(&obj->gpu_write_list)){
+			DRM_ERROR("invalid flushing %p (a %d w %x gwl %d)\n",
+				  obj,
+				  obj->active,
+				  obj->base.write_domain,
+				  !list_empty(&obj->gpu_write_list));
+			err++;
+		}
+	}
+
+	list_for_each_entry(obj, &dev_priv->mm.gpu_write_list, gpu_write_list) {
+		if (obj->base.dev != dev ||
+		    !atomic_read(&obj->base.refcount.refcount)) {
+			DRM_ERROR("freed gpu write %p\n", obj);
+			err++;
+			break;
+		} else if (!obj->active ||
+			   (obj->base.write_domain & I915_GEM_GPU_DOMAINS) == 0) {
+			DRM_ERROR("invalid gpu write %p (a %d w %x)\n",
 				  obj,
-				  obj_priv->pin_count, obj_priv->active,
-				  obj->write_domain, file, line);
+				  obj->active,
+				  obj->base.write_domain);
+			err++;
+		}
+	}
+
+	list_for_each_entry(obj, &dev_priv->mm.inactive_list, list) {
+		if (obj->base.dev != dev ||
+		    !atomic_read(&obj->base.refcount.refcount)) {
+			DRM_ERROR("freed inactive %p\n", obj);
+			err++;
+			break;
+		} else if (obj->pin_count || obj->active ||
+			   (obj->base.write_domain & I915_GEM_GPU_DOMAINS)) {
+			DRM_ERROR("invalid inactive %p (p %d a %d w %x)\n",
+				  obj,
+				  obj->pin_count, obj->active,
+				  obj->base.write_domain);
+			err++;
+		}
 	}
+
+	list_for_each_entry(obj, &dev_priv->mm.pinned_list, list) {
+		if (obj->base.dev != dev ||
+		    !atomic_read(&obj->base.refcount.refcount)) {
+			DRM_ERROR("freed pinned %p\n", obj);
+			err++;
+			break;
+		} else if (!obj->pin_count || obj->active ||
+			   (obj->base.write_domain & I915_GEM_GPU_DOMAINS)) {
+			DRM_ERROR("invalid pinned %p (p %d a %d w %x)\n",
+				  obj,
+				  obj->pin_count, obj->active,
+				  obj->base.write_domain);
+			err++;
+		}
+	}
+
+	return warned = err;
 }
 #endif /* WATCH_INACTIVE */
 
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 5103b95cea9..d89b88791aa 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -579,6 +579,8 @@ int intel_init_ring_buffer(struct drm_device *dev,
 	int ret;
 
 	ring->dev = dev;
+	INIT_LIST_HEAD(&ring->active_list);
+	INIT_LIST_HEAD(&ring->request_list);
 
 	if (I915_NEED_GFX_HWS(dev)) {
 		ret = init_status_page(dev, ring);
@@ -627,8 +629,6 @@ int intel_init_ring_buffer(struct drm_device *dev,
 		if (ring->space < 0)
 			ring->space += ring->size;
 	}
-	INIT_LIST_HEAD(&ring->active_list);
-	INIT_LIST_HEAD(&ring->request_list);
 	return ret;
 
 err_unmap: