3 files changed, 76 insertions, 16 deletions

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index cb4cf7ef4d1e..db809e034cc4 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -442,6 +442,43 @@ out_err:
 }
 
 /**
+ * Call bo::reserved and with the lru lock held.
+ * Will release GPU memory type usage on destruction.
+ * This is the place to put in driver specific hooks.
+ * Will release the bo::reserved lock and the
+ * lru lock on exit.
+ */
+
+static void ttm_bo_cleanup_memtype_use(struct ttm_buffer_object *bo)
+{
+	struct ttm_bo_global *glob = bo->glob;
+
+	if (bo->ttm) {
+
+		/**
+		 * Release the lru_lock, since we don't want to have
+		 * an atomic requirement on ttm_tt[unbind|destroy].
+		 */
+
+		spin_unlock(&glob->lru_lock);
+		ttm_tt_unbind(bo->ttm);
+		ttm_tt_destroy(bo->ttm);
+		bo->ttm = NULL;
+		spin_lock(&glob->lru_lock);
+	}
+
+	if (bo->mem.mm_node) {
+		drm_mm_put_block(bo->mem.mm_node);
+		bo->mem.mm_node = NULL;
+	}
+
+	atomic_set(&bo->reserved, 0);
+	wake_up_all(&bo->event_queue);
+	spin_unlock(&glob->lru_lock);
+}
+
+
+/**
  * If bo idle, remove from delayed- and lru lists, and unref.
  * If not idle, and already on delayed list, do nothing.
  * If not idle, and not on delayed list, put on delayed list,
@@ -456,6 +493,7 @@ static int ttm_bo_cleanup_refs(struct ttm_buffer_object *bo, bool remove_all)
 	int ret;
 
 	spin_lock(&bo->lock);
+retry:
 	(void) ttm_bo_wait(bo, false, false, !remove_all);
 
 	if (!bo->sync_obj) {
@@ -464,31 +502,52 @@ static int ttm_bo_cleanup_refs(struct ttm_buffer_object *bo, bool remove_all)
 		spin_unlock(&bo->lock);
 
 		spin_lock(&glob->lru_lock);
-		put_count = ttm_bo_del_from_lru(bo);
+		ret = ttm_bo_reserve_locked(bo, false, !remove_all, false, 0);
+
+		/**
+		 * Someone else has the object reserved. Bail and retry.
+		 */
 
-		ret = ttm_bo_reserve_locked(bo, false, false, false, 0);
-		BUG_ON(ret);
-		if (bo->ttm)
-			ttm_tt_unbind(bo->ttm);
+		if (unlikely(ret == -EBUSY)) {
+			spin_unlock(&glob->lru_lock);
+			spin_lock(&bo->lock);
+			goto requeue;
+		}
+
+		/**
+		 * We can re-check for sync object without taking
+		 * the bo::lock since setting the sync object requires
+		 * also bo::reserved. A busy object at this point may
+		 * be caused by another thread starting an accelerated
+		 * eviction.
+		 */
+
+		if (unlikely(bo->sync_obj)) {
+			atomic_set(&bo->reserved, 0);
+			wake_up_all(&bo->event_queue);
+			spin_unlock(&glob->lru_lock);
+			spin_lock(&bo->lock);
+			if (remove_all)
+				goto retry;
+			else
+				goto requeue;
+		}
+
+		put_count = ttm_bo_del_from_lru(bo);
 
 		if (!list_empty(&bo->ddestroy)) {
 			list_del_init(&bo->ddestroy);
 			++put_count;
 		}
-		if (bo->mem.mm_node) {
-			drm_mm_put_block(bo->mem.mm_node);
-			bo->mem.mm_node = NULL;
-		}
-		spin_unlock(&glob->lru_lock);
 
-		atomic_set(&bo->reserved, 0);
+		ttm_bo_cleanup_memtype_use(bo);
 
 		while (put_count--)
 			kref_put(&bo->list_kref, ttm_bo_ref_bug);
 
 		return 0;
 	}
-
+requeue:
 	spin_lock(&glob->lru_lock);
 	if (list_empty(&bo->ddestroy)) {
 		void *sync_obj = bo->sync_obj;
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 7cffb3e04232..3451a82adba7 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -351,6 +351,7 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
 	INIT_LIST_HEAD(&fbo->lru);
 	INIT_LIST_HEAD(&fbo->swap);
 	fbo->vm_node = NULL;
+	atomic_set(&fbo->cpu_writers, 0);
 
 	fbo->sync_obj = driver->sync_obj_ref(bo->sync_obj);
 	kref_init(&fbo->list_kref);
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
index ca904799f018..b1e02fffd3cc 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
@@ -69,7 +69,7 @@ struct ttm_page_pool {
 	spinlock_t		lock;
 	bool			fill_lock;
 	struct list_head	list;
-	int			gfp_flags;
+	gfp_t			gfp_flags;
 	unsigned		npages;
 	char			*name;
 	unsigned long		nfrees;
@@ -475,7 +475,7 @@ static void ttm_handle_caching_state_failure(struct list_head *pages,
  * This function is reentrant if caller updates count depending on number of
  * pages returned in pages array.
  */
-static int ttm_alloc_new_pages(struct list_head *pages, int gfp_flags,
+static int ttm_alloc_new_pages(struct list_head *pages, gfp_t gfp_flags,
 		int ttm_flags, enum ttm_caching_state cstate, unsigned count)
 {
 	struct page **caching_array;
@@ -666,7 +666,7 @@ int ttm_get_pages(struct list_head *pages, int flags,
 {
 	struct ttm_page_pool *pool = ttm_get_pool(flags, cstate);
 	struct page *p = NULL;
-	int gfp_flags = GFP_USER;
+	gfp_t gfp_flags = GFP_USER;
 	int r;
 
 	/* set zero flag for page allocation if required */
@@ -818,7 +818,7 @@ int ttm_page_alloc_init(struct ttm_mem_global *glob, unsigned max_pages)
 	return 0;
 }
 
-void ttm_page_alloc_fini()
+void ttm_page_alloc_fini(void)
 {
 	int i;