Ocfs2 devel - Aug 2015 - [patch 01/28] ocfs2: set filesytem read-only when ocfs2_delete_entry failed.

From: jiangyiwen <jiangyiwen at huawei.com>
Subject: ocfs2: set filesytem read-only when ocfs2_delete_entry failed.

In ocfs2_rename, it will lead to an inode with two entried(old and new) if
ocfs2_delete_entry(old) failed.  Thus, filesystem will be inconsistent.

The case is described below:

ocfs2_rename
    -> ocfs2_start_trans
    -> ocfs2_add_entry(new)
    -> ocfs2_delete_entry(old)
        -> __ocfs2_journal_access *failed* because of -ENOMEM
    -> ocfs2_commit_trans

So filesystem should be set to read-only at the moment.

Signed-off-by: Yiwen Jiang <jiangyiwen at huawei.com>
Cc: Joseph Qi <joseph.qi at huawei.com>
Cc: Joel Becker <jlbec at evilplan.org>
Cc: Mark Fasheh <mfasheh at suse.com>
Signed-off-by: Andrew Morton <akpm at linux-foundation.org>
---

 fs/ocfs2/namei.c |   15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff -puN
fs/ocfs2/namei.c~ocfs2-set-filesytem-read-only-when-ocfs2_delete_entry-failed
fs/ocfs2/namei.c
---
a/fs/ocfs2/namei.c~ocfs2-set-filesytem-read-only-when-ocfs2_delete_entry-failed
+++ a/fs/ocfs2/namei.c
@@ -1544,12 +1544,25 @@ static int ocfs2_rename(struct inode *ol
 	status = ocfs2_find_entry(old_dentry->d_name.name,
 				  old_dentry->d_name.len, old_dir,
 				  &old_entry_lookup);
-	if (status)
+	if (status) {
+		if (!is_journal_aborted(osb->journal->j_journal)) {
+			ocfs2_error(osb->sb, "new entry %.*s is added, but old entry %.*s
"
+					"is not deleted.",
+					new_dentry->d_name.len, new_dentry->d_name.name,
+					old_dentry->d_name.len, old_dentry->d_name.name);
+		}
 		goto bail;
+	}
 
 	status = ocfs2_delete_entry(handle, old_dir, &old_entry_lookup);
 	if (status < 0) {
 		mlog_errno(status);
+		if (!is_journal_aborted(osb->journal->j_journal)) {
+			ocfs2_error(osb->sb, "new entry %.*s is added, but old entry %.*s
"
+					"is not deleted.",
+					new_dentry->d_name.len, new_dentry->d_name.name,
+					old_dentry->d_name.len, old_dentry->d_name.name);
+		}
 		goto bail;
 	}
 
_

On Wed, Aug 26, 2015 at 03:11:21PM -0700, Andrew Morton
wrote:
> From: jiangyiwen <jiangyiwen at huawei.com>
> Subject: ocfs2: set filesytem read-only when ocfs2_delete_entry failed.
> 
> In ocfs2_rename, it will lead to an inode with two entried(old and new) if
> ocfs2_delete_entry(old) failed.  Thus, filesystem will be inconsistent.
> 
> The case is described below:
> 
> ocfs2_rename
>     -> ocfs2_start_trans
>     -> ocfs2_add_entry(new)
>     -> ocfs2_delete_entry(old)
>         -> __ocfs2_journal_access *failed* because of -ENOMEM
>     -> ocfs2_commit_trans
> 
> So filesystem should be set to read-only at the moment.
> 
> Signed-off-by: Yiwen Jiang <jiangyiwen at huawei.com>
> Cc: Joseph Qi <joseph.qi at huawei.com>
> Cc: Joel Becker <jlbec at evilplan.org>
> Cc: Mark Fasheh <mfasheh at suse.com>
> Signed-off-by: Andrew Morton <akpm at linux-foundation.org>
Reviewed-by: Mark Fasheh <mfasheh at suse.de>

--
Mark Fasheh