btrfs: add missing setup of log for full commit at add_conflicting_inode()
authorFilipe Manana <fdmanana@suse.com>
Tue, 10 Jan 2023 14:56:36 +0000 (14:56 +0000)
committerDavid Sterba <dsterba@suse.com>
Thu, 12 Jan 2023 14:43:30 +0000 (15:43 +0100)
When logging conflicting inodes, if we reach the maximum limit of inodes,
we return BTRFS_LOG_FORCE_COMMIT to force a transaction commit. However
we don't mark the log for full commit (with btrfs_set_log_full_commit()),
which means that once we leave the log transaction and before we commit
the transaction, some other task may sync the log, which is incomplete
as we have not logged all conflicting inodes, leading to some inconsistent
in case that log ends up being replayed.

So also call btrfs_set_log_full_commit() at add_conflicting_inode().

Fixes: e09d94c9e448 ("btrfs: log conflicting inodes without holding log mutex of the initial inode")
CC: stable@vger.kernel.org # 6.1
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/tree-log.c

index c09daab3f19e8ffcaa1c764f9e1e11b05ce85382..afad44a0becfe9e5becea4fbbb63d58c5c147d36 100644 (file)
@@ -5598,8 +5598,10 @@ static int add_conflicting_inode(struct btrfs_trans_handle *trans,
         * LOG_INODE_EXISTS mode) and slow down other fsyncs or transaction
         * commits.
         */
-       if (ctx->num_conflict_inodes >= MAX_CONFLICT_INODES)
+       if (ctx->num_conflict_inodes >= MAX_CONFLICT_INODES) {
+               btrfs_set_log_full_commit(trans);
                return BTRFS_LOG_FORCE_COMMIT;
+       }
 
        inode = btrfs_iget(root->fs_info->sb, ino, root);
        /*