1 files changed, 62 insertions, 49 deletions

diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c
index 7696beec4c21..2decb9fff445 100644
--- a/fs/btrfs/dev-replace.c
+++ b/fs/btrfs/dev-replace.c
@@ -45,7 +45,7 @@
  *
  * - Copy existing extents
  *
- *   This happens by re-using scrub facility, as scrub also iterates through
+ *   This happens by reusing scrub facility, as scrub also iterates through
  *   existing extents from commit root.
  *
  *   Location:		scrub_write_block_to_dev_replace() from
@@ -76,7 +76,7 @@ int btrfs_init_dev_replace(struct btrfs_fs_info *fs_info)
 	struct extent_buffer *eb;
 	int slot;
 	int ret = 0;
-	struct btrfs_path *path = NULL;
+	BTRFS_PATH_AUTO_FREE(path);
 	int item_size;
 	struct btrfs_dev_replace_item *ptr;
 	u64 src_devid;
@@ -85,10 +85,8 @@ int btrfs_init_dev_replace(struct btrfs_fs_info *fs_info)
 		return 0;
 
 	path = btrfs_alloc_path();
-	if (!path) {
-		ret = -ENOMEM;
-		goto out;
-	}
+	if (!path)
+		return -ENOMEM;
 
 	key.objectid = 0;
 	key.type = BTRFS_DEV_REPLACE_KEY;
@@ -103,10 +101,8 @@ no_valid_dev_replace_entry_found:
 		if (btrfs_find_device(fs_info->fs_devices, &args)) {
 			btrfs_err(fs_info,
 			"found replace target device without a valid replace item");
-			ret = -EUCLEAN;
-			goto out;
+			return -EUCLEAN;
 		}
-		ret = 0;
 		dev_replace->replace_state =
 			BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED;
 		dev_replace->cont_reading_from_srcdev_mode =
@@ -123,7 +119,7 @@ no_valid_dev_replace_entry_found:
 		dev_replace->tgtdev = NULL;
 		dev_replace->is_valid = 0;
 		dev_replace->item_needs_writeback = 0;
-		goto out;
+		return 0;
 	}
 	slot = path->slots[0];
 	eb = path->nodes[0];
@@ -226,8 +222,6 @@ no_valid_dev_replace_entry_found:
 		break;
 	}
 
-out:
-	btrfs_free_path(path);
 	return ret;
 }
 
@@ -316,7 +310,7 @@ static int btrfs_init_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
 	set_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &device->dev_state);
 	set_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state);
 	device->dev_stats_valid = 1;
-	set_blocksize(device->bdev, BTRFS_BDEV_BLOCKSIZE);
+	set_blocksize(bdev_file, BTRFS_BDEV_BLOCKSIZE);
 	device->fs_devices = fs_devices;
 
 	ret = btrfs_get_dev_zone_info(device, false);
@@ -346,7 +340,7 @@ int btrfs_run_dev_replace(struct btrfs_trans_handle *trans)
 	struct btrfs_fs_info *fs_info = trans->fs_info;
 	int ret;
 	struct btrfs_root *dev_root = fs_info->dev_root;
-	struct btrfs_path *path;
+	BTRFS_PATH_AUTO_FREE(path);
 	struct btrfs_key key;
 	struct extent_buffer *eb;
 	struct btrfs_dev_replace_item *ptr;
@@ -365,16 +359,15 @@ int btrfs_run_dev_replace(struct btrfs_trans_handle *trans)
 	key.offset = 0;
 
 	path = btrfs_alloc_path();
-	if (!path) {
-		ret = -ENOMEM;
-		goto out;
-	}
+	if (!path)
+		return -ENOMEM;
+
 	ret = btrfs_search_slot(trans, dev_root, &key, path, -1, 1);
 	if (ret < 0) {
 		btrfs_warn(fs_info,
 			   "error %d while searching for dev_replace item!",
 			   ret);
-		goto out;
+		return ret;
 	}
 
 	if (ret == 0 &&
@@ -395,7 +388,7 @@ int btrfs_run_dev_replace(struct btrfs_trans_handle *trans)
 			btrfs_warn(fs_info,
 				   "delete too small dev_replace item failed %d!",
 				   ret);
-			goto out;
+			return ret;
 		}
 		ret = 1;
 	}
@@ -408,7 +401,7 @@ int btrfs_run_dev_replace(struct btrfs_trans_handle *trans)
 		if (ret < 0) {
 			btrfs_warn(fs_info,
 				   "insert dev_replace item failed %d!", ret);
-			goto out;
+			return ret;
 		}
 	}
 
@@ -441,11 +434,6 @@ int btrfs_run_dev_replace(struct btrfs_trans_handle *trans)
 	dev_replace->item_needs_writeback = 0;
 	up_write(&dev_replace->rwsem);
 
-	btrfs_mark_buffer_dirty(trans, eb);
-
-out:
-	btrfs_free_path(path);
-
 	return ret;
 }
 
@@ -641,6 +629,7 @@ static int btrfs_dev_replace_start(struct btrfs_fs_info *fs_info,
 		return ret;
 
 	down_write(&dev_replace->rwsem);
+	dev_replace->replace_task = current;
 	switch (dev_replace->replace_state) {
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
@@ -648,7 +637,7 @@ static int btrfs_dev_replace_start(struct btrfs_fs_info *fs_info,
 		break;
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
-		ASSERT(0);
+		DEBUG_WARN("unexpected STARTED ot SUSPENDED dev-replace state");
 		ret = BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED;
 		up_write(&dev_replace->rwsem);
 		goto leave;
@@ -684,7 +673,7 @@ static int btrfs_dev_replace_start(struct btrfs_fs_info *fs_info,
 	if (ret)
 		btrfs_err(fs_info, "kobj add dev failed %d", ret);
 
-	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
+	btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
 
 	/*
 	 * Commit dev_replace state and reserve 1 item for it.
@@ -805,17 +794,17 @@ static int btrfs_set_target_alloc_state(struct btrfs_device *srcdev,
 
 	lockdep_assert_held(&srcdev->fs_info->chunk_mutex);
 
-	while (find_first_extent_bit(&srcdev->alloc_state, start,
-				     &found_start, &found_end,
-				     CHUNK_ALLOCATED, &cached_state)) {
-		ret = set_extent_bit(&tgtdev->alloc_state, found_start,
-				     found_end, CHUNK_ALLOCATED, NULL);
+	while (btrfs_find_first_extent_bit(&srcdev->alloc_state, start,
+					   &found_start, &found_end,
+					   CHUNK_ALLOCATED, &cached_state)) {
+		ret = btrfs_set_extent_bit(&tgtdev->alloc_state, found_start,
+					   found_end, CHUNK_ALLOCATED, NULL);
 		if (ret)
 			break;
 		start = found_end + 1;
 	}
 
-	free_extent_state(cached_state);
+	btrfs_free_extent_state(cached_state);
 	return ret;
 }
 
@@ -824,22 +813,45 @@ static void btrfs_dev_replace_update_device_in_mapping_tree(
 						struct btrfs_device *srcdev,
 						struct btrfs_device *tgtdev)
 {
-	u64 start = 0;
-	int i;
+	struct rb_node *node;
+
+	/*
+	 * The chunk mutex must be held so that no new chunks can be created
+	 * while we are updating existing chunks. This guarantees we don't miss
+	 * any new chunk that gets created for a range that falls before the
+	 * range of the last chunk we processed.
+	 */
+	lockdep_assert_held(&fs_info->chunk_mutex);
 
 	write_lock(&fs_info->mapping_tree_lock);
-	do {
+	node = rb_first_cached(&fs_info->mapping_tree);
+	while (node) {
+		struct rb_node *next = rb_next(node);
 		struct btrfs_chunk_map *map;
+		u64 next_start;
 
-		map = btrfs_find_chunk_map_nolock(fs_info, start, U64_MAX);
-		if (!map)
-			break;
-		for (i = 0; i < map->num_stripes; i++)
+		map = rb_entry(node, struct btrfs_chunk_map, rb_node);
+		next_start = map->start + map->chunk_len;
+
+		for (int i = 0; i < map->num_stripes; i++)
 			if (srcdev == map->stripes[i].dev)
 				map->stripes[i].dev = tgtdev;
-		start = map->start + map->chunk_len;
-		btrfs_free_chunk_map(map);
-	} while (start);
+
+		if (cond_resched_rwlock_write(&fs_info->mapping_tree_lock)) {
+			map = btrfs_find_chunk_map_nolock(fs_info, next_start, U64_MAX);
+			if (!map)
+				break;
+			node = &map->rb_node;
+			/*
+			 * Drop the lookup reference since we are holding the
+			 * lock in write mode and no one can remove the chunk
+			 * map from the tree and drop its tree reference.
+			 */
+			btrfs_free_chunk_map(map);
+		} else {
+			node = next;
+		}
+	}
 	write_unlock(&fs_info->mapping_tree_lock);
 }
 
@@ -880,7 +892,7 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
 		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
 		return ret;
 	}
-	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
+	btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
 
 	/*
 	 * We have to use this loop approach because at this point src_device
@@ -971,6 +983,7 @@ error:
 	list_add(&tgt_device->dev_alloc_list, &fs_devices->alloc_list);
 	fs_devices->rw_devices++;
 
+	dev_replace->replace_task = NULL;
 	up_write(&dev_replace->rwsem);
 	btrfs_rm_dev_replace_blocked(fs_info);
 
@@ -1252,16 +1265,16 @@ static int btrfs_dev_replace_kthread(void *data)
 	return 0;
 }
 
-int __pure btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace)
+bool __pure btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace)
 {
 	if (!dev_replace->is_valid)
-		return 0;
+		return false;
 
 	switch (dev_replace->replace_state) {
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
-		return 0;
+		return false;
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
 	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
 		/*
@@ -1276,7 +1289,7 @@ int __pure btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace)
 		 */
 		break;
 	}
-	return 1;
+	return true;
 }
 
 void btrfs_bio_counter_sub(struct btrfs_fs_info *fs_info, s64 amount)