Btrfs: Avoid orphan inodes cleanup while replaying log

We do log replay in a single transaction, so it's not good to do unbound
operations. This patch cleans up orphan inodes cleanup after replaying
the log. It also avoids doing other unbound operations such as truncating
a file during replaying log. These unbound operations are postponed to
the orphan inode cleanup stage.

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

5 files changed, 55 insertions, 36 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index ae5b0aa..fcfbefb 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -859,8 +859,9 @@ struct btrfs_fs_info {
 	struct mutex ordered_operations_mutex;
 	struct rw_semaphore extent_commit_sem;
 
-	struct rw_semaphore subvol_sem;
+	struct rw_semaphore cleanup_work_sem;
 
+	struct rw_semaphore subvol_sem;
 	struct srcu_struct subvol_srcu;
 
 	struct list_head trans_list;
@@ -1034,12 +1035,12 @@ struct btrfs_root {
 	int ref_cows;
 	int track_dirty;
 	int in_radix;
+	int clean_orphans;
 
 	u64 defrag_trans_start;
 	struct btrfs_key defrag_progress;
 	struct btrfs_key defrag_max;
 	int defrag_running;
-	int defrag_level;
 	char *name;
 	int in_sysfs;
 
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 101940f..c1e59e3 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -892,6 +892,8 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->stripesize = stripesize;
 	root->ref_cows = 0;
 	root->track_dirty = 0;
+	root->in_radix = 0;
+	root->clean_orphans = 0;
 
 	root->fs_info = fs_info;
 	root->objectid = objectid;
@@ -928,7 +930,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->defrag_trans_start = fs_info->generation;
 	init_completion(&root->kobj_unregister);
 	root->defrag_running = 0;
-	root->defrag_level = 0;
 	root->root_key.objectid = objectid;
 	root->anon_super.s_root = NULL;
 	root->anon_super.s_dev = 0;
@@ -1210,8 +1211,10 @@ again:
 	ret = radix_tree_insert(&fs_info->fs_roots_radix,
 				(unsigned long)root->root_key.objectid,
 				root);
-	if (ret == 0)
+	if (ret == 0) {
 		root->in_radix = 1;
+		root->clean_orphans = 1;
+	}
 	spin_unlock(&fs_info->fs_roots_radix_lock);
 	radix_tree_preload_end();
 	if (ret) {
@@ -1225,10 +1228,6 @@ again:
 	ret = btrfs_find_dead_roots(fs_info->tree_root,
 				    root->root_key.objectid);
 	WARN_ON(ret);
-
-	if (!(fs_info->sb->s_flags & MS_RDONLY))
-		btrfs_orphan_cleanup(root);
-
 	return root;
 fail:
 	free_fs_root(root);
@@ -1689,6 +1688,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	mutex_init(&fs_info->cleaner_mutex);
 	mutex_init(&fs_info->volume_mutex);
 	init_rwsem(&fs_info->extent_commit_sem);
+	init_rwsem(&fs_info->cleanup_work_sem);
 	init_rwsem(&fs_info->subvol_sem);
 
 	btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
@@ -2388,6 +2388,11 @@ int btrfs_commit_super(struct btrfs_root *root)
 	mutex_lock(&root->fs_info->cleaner_mutex);
 	btrfs_clean_old_snapshots(root);
 	mutex_unlock(&root->fs_info->cleaner_mutex);
+
+	/* wait until ongoing cleanup work done */
+	down_write(&root->fs_info->cleanup_work_sem);
+	up_write(&root->fs_info->cleanup_work_sem);
+
 	trans = btrfs_start_transaction(root, 1);
 	ret = btrfs_commit_transaction(trans, root);
 	BUG_ON(ret);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index fa57247..eb2db3b 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -2093,16 +2093,17 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
 	struct inode *inode;
 	int ret = 0, nr_unlink = 0, nr_truncate = 0;
 
-	path = btrfs_alloc_path();
-	if (!path)
+	if (!xchg(&root->clean_orphans, 0))
 		return;
+
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
 	path->reada = -1;
 
 	key.objectid = BTRFS_ORPHAN_OBJECTID;
 	btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
 	key.offset = (u64)-1;
 
-
 	while (1) {
 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 		if (ret < 0) {
@@ -3298,6 +3299,11 @@ void btrfs_delete_inode(struct inode *inode)
 	}
 	btrfs_wait_ordered_range(inode, 0, (u64)-1);
 
+	if (root->fs_info->log_root_recovering) {
+		BUG_ON(!list_empty(&BTRFS_I(inode)->i_orphan));
+		goto no_delete;
+	}
+
 	if (inode->i_nlink > 0) {
 		BUG_ON(btrfs_root_refs(&root->root_item) != 0);
 		goto no_delete;
@@ -3705,6 +3711,13 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
 	}
 	srcu_read_unlock(&root->fs_info->subvol_srcu, index);
 
+	if (root != sub_root) {
+		down_read(&root->fs_info->cleanup_work_sem);
+		if (!(inode->i_sb->s_flags & MS_RDONLY))
+			btrfs_orphan_cleanup(sub_root);
+		up_read(&root->fs_info->cleanup_work_sem);
+	}
+
 	return inode;
 }
 
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index cfcc93c..975fdd3 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -3755,6 +3755,7 @@ out:
 				       BTRFS_DATA_RELOC_TREE_OBJECTID);
 		if (IS_ERR(fs_root))
 			err = PTR_ERR(fs_root);
+		btrfs_orphan_cleanup(fs_root);
 	}
 	return err;
 }
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 6bbaa10..4a9434b 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -930,6 +930,17 @@ out_nowrite:
 	return 0;
 }
 
+static int insert_orphan_item(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root, u64 offset)
+{
+	int ret;
+	ret = btrfs_find_orphan_item(root, offset);
+	if (ret > 0)
+		ret = btrfs_insert_orphan_item(trans, root, offset);
+	return ret;
+}
+
+
 /*
  * There are a few corners where the link count of the file can't
  * be properly maintained during replay.  So, instead of adding
@@ -997,9 +1008,13 @@ static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
 	}
 	BTRFS_I(inode)->index_cnt = (u64)-1;
 
-	if (inode->i_nlink == 0 && S_ISDIR(inode->i_mode)) {
-		ret = replay_dir_deletes(trans, root, NULL, path,
-					 inode->i_ino, 1);
+	if (inode->i_nlink == 0) {
+		if (S_ISDIR(inode->i_mode)) {
+			ret = replay_dir_deletes(trans, root, NULL, path,
+						 inode->i_ino, 1);
+			BUG_ON(ret);
+		}
+		ret = insert_orphan_item(trans, root, inode->i_ino);
 		BUG_ON(ret);
 	}
 	btrfs_free_path(path);
@@ -1587,7 +1602,6 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
 		/* inode keys are done during the first stage */
 		if (key.type == BTRFS_INODE_ITEM_KEY &&
 		    wc->stage == LOG_WALK_REPLAY_INODES) {
-			struct inode *inode;
 			struct btrfs_inode_item *inode_item;
 			u32 mode;
 
@@ -1603,31 +1617,16 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
 					     eb, i, &key);
 			BUG_ON(ret);
 
-			/* for regular files, truncate away
-			 * extents past the new EOF
+			/* for regular files, make sure corresponding
+			 * orhpan item exist. extents past the new EOF
+			 * will be truncated later by orphan cleanup.
 			 */
 			if (S_ISREG(mode)) {
-				inode = read_one_inode(root,
-						       key.objectid);
-				BUG_ON(!inode);
-
-				ret = btrfs_truncate_inode_items(wc->trans,
-					root, inode, inode->i_size,
-					BTRFS_EXTENT_DATA_KEY);
+				ret = insert_orphan_item(wc->trans, root,
+							 key.objectid);
 				BUG_ON(ret);
-
-				/* if the nlink count is zero here, the iput
-				 * will free the inode.  We bump it to make
-				 * sure it doesn't get freed until the link
-				 * count fixup is done
-				 */
-				if (inode->i_nlink == 0) {
-					btrfs_inc_nlink(inode);
-					btrfs_update_inode(wc->trans,
-							   root, inode);
-				}
-				iput(inode);
 			}
+
 			ret = link_to_fixup_dir(wc->trans, root,
 						path, key.objectid);
 			BUG_ON(ret);