btrfs: skip transaction commit after failure to create subvolume

At ioctl.c:create_subvol(), when we fail to create a subvolume we always
commit the transaction. In most cases this is a no-op, since all the error
paths, except for one, abort the transaction - the only exception is when
we fail to insert the new root item into the root tree, in that case we
don't abort the transaction because we didn't do anything that is
irreversible - however we end up committing the transaction which although
is not a functional problem, it adds unnecessary rotation of the backup
roots in the superblock and unnecessary work.

So change that to commit a transaction only when no error happened,
otherwise just call btrfs_end_transaction() to release our reference on
the transaction.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 7565b667f4fc0ccf4b308bab6eb88c9b17579102..a5bd6926f7ffe9741eeddcbd5ce246019728423e 100644 (file)
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -544,7 +544,6 @@ static noinline int create_subvol(struct user_namespace *mnt_userns,
        struct timespec64 cur_time = current_time(dir);
        struct inode *inode;
        int ret;
-       int err;
        dev_t anon_dev = 0;
        u64 objectid;
        u64 index = 0;
@@ -724,9 +723,10 @@ fail:
        trans->bytes_reserved = 0;
        btrfs_subvolume_release_metadata(root, &block_rsv);
 
-       err = btrfs_commit_transaction(trans);
-       if (err && !ret)
-               ret = err;
+       if (ret)
+               btrfs_end_transaction(trans);
+       else
+               ret = btrfs_commit_transaction(trans);
 
        if (!ret) {
                inode = btrfs_lookup_dentry(dir, dentry);