Commit 86e8aa0e authored by Filipe Manana's avatar Filipe Manana

Btrfs: unpin logs if rename exchange operation fails

If rename exchange operations fail at some point after we pinned any of
the logs, we end up aborting the current transaction but never unpin the
logs, which leaves concurrent tasks that are trying to sync the logs (as
part of an fsync request from user space) blocked forever and preventing
the filesystem from being unmountable.

Fix this by safely unpinning the log.
Signed-off-by: default avatarFilipe Manana <fdmanana@suse.com>
parent c9901618
......@@ -9412,6 +9412,8 @@ static int btrfs_rename_exchange(struct inode *old_dir,
u64 new_idx = 0;
u64 root_objectid;
int ret;
bool root_log_pinned = false;
bool dest_log_pinned = false;
/* we only allow rename subvolume link between subvolumes */
if (old_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
......@@ -9464,6 +9466,7 @@ static int btrfs_rename_exchange(struct inode *old_dir,
if (ret)
goto out_fail;
btrfs_pin_log_trans(root);
root_log_pinned = true;
}
/* And now for the dest. */
......@@ -9479,6 +9482,7 @@ static int btrfs_rename_exchange(struct inode *old_dir,
if (ret)
goto out_fail;
btrfs_pin_log_trans(dest);
dest_log_pinned = true;
}
/* Update inode version and ctime/mtime. */
......@@ -9557,17 +9561,47 @@ static int btrfs_rename_exchange(struct inode *old_dir,
if (new_inode->i_nlink == 1)
BTRFS_I(new_inode)->dir_index = new_idx;
if (old_ino != BTRFS_FIRST_FREE_OBJECTID) {
if (root_log_pinned) {
parent = new_dentry->d_parent;
btrfs_log_new_name(trans, old_inode, old_dir, parent);
btrfs_end_log_trans(root);
root_log_pinned = false;
}
if (new_ino != BTRFS_FIRST_FREE_OBJECTID) {
if (dest_log_pinned) {
parent = old_dentry->d_parent;
btrfs_log_new_name(trans, new_inode, new_dir, parent);
btrfs_end_log_trans(dest);
dest_log_pinned = false;
}
out_fail:
/*
* If we have pinned a log and an error happened, we unpin tasks
* trying to sync the log and force them to fallback to a transaction
* commit if the log currently contains any of the inodes involved in
* this rename operation (to ensure we do not persist a log with an
* inconsistent state for any of these inodes or leading to any
* inconsistencies when replayed). If the transaction was aborted, the
* abortion reason is propagated to userspace when attempting to commit
* the transaction. If the log does not contain any of these inodes, we
* allow the tasks to sync it.
*/
if (ret && (root_log_pinned || dest_log_pinned)) {
if (btrfs_inode_in_log(old_dir, root->fs_info->generation) ||
btrfs_inode_in_log(new_dir, root->fs_info->generation) ||
btrfs_inode_in_log(old_inode, root->fs_info->generation) ||
(new_inode &&
btrfs_inode_in_log(new_inode, root->fs_info->generation)))
btrfs_set_log_full_commit(root->fs_info, trans);
if (root_log_pinned) {
btrfs_end_log_trans(root);
root_log_pinned = false;
}
if (dest_log_pinned) {
btrfs_end_log_trans(dest);
dest_log_pinned = false;
}
}
ret = btrfs_end_transaction(trans, root);
out_notrans:
if (new_ino == BTRFS_FIRST_FREE_OBJECTID)
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment