ext4: drop s_umount over opening the log device
authorChristoph Hellwig <hch@lst.de>
Wed, 2 Aug 2023 15:41:28 +0000 (17:41 +0200)
committerChristian Brauner <brauner@kernel.org>
Fri, 11 Aug 2023 11:15:04 +0000 (13:15 +0200)
Just like get_tree_bdev needs to drop s_umount when opening the main
device, we need to do the same for the ext4 log device to avoid a
potential lock order reversal with s_unmount for the mark_dead path.

It might be preferable to just drop s_umount over ->fill_super entirely,
but that will require a fairly massive audit first, so we'll do the easy
version here first.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christian Brauner <brauner@kernel.org>
Message-Id: <20230802154131.2221419-10-hch@lst.de>
Signed-off-by: Christian Brauner <brauner@kernel.org>
fs/ext4/super.c

index fa8888f..47e3272 100644 (file)
@@ -5842,7 +5842,10 @@ static journal_t *ext4_get_dev_journal(struct super_block *sb,
        if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
                return NULL;
 
+       /* see get_tree_bdev why this is needed and safe */
+       up_write(&sb->s_umount);
        bdev = ext4_blkdev_get(j_dev, sb);
+       down_write(&sb->s_umount);
        if (bdev == NULL)
                return NULL;