Commit 0e446673 authored by Christoph Hellwig's avatar Christoph Hellwig Committed by Niv Sardi

[XFS] fix error handling in xlog_recover_process_one_iunlink

If we fail after xfs_iget we have to drop the reference count, spotted
by Dave Chinner.  Also remove some useless asserts and stop trying to
deal with di_mode == 0 inodes because never gets those without passing
the IGET_CREATE flag to xfs_iget.
Signed-off-by: default avatarChristoph Hellwig <hch@lst.de>
Reviewed-by: default avatarEric Sandeen <sandeen@sandeen.net>
Signed-off-by: default avatarNiv Sardi <xaiki@sgi.com>
parent 24f211ba
...@@ -3147,13 +3147,12 @@ xlog_recover_process_one_iunlink( ...@@ -3147,13 +3147,12 @@ xlog_recover_process_one_iunlink(
/* /*
* Get the on disk inode to find the next inode in the bucket. * Get the on disk inode to find the next inode in the bucket.
*/ */
ASSERT(ip != NULL);
error = xfs_itobp(mp, NULL, ip, &dip, &ibp, XFS_BUF_LOCK); error = xfs_itobp(mp, NULL, ip, &dip, &ibp, XFS_BUF_LOCK);
if (error) if (error)
goto fail; goto fail_iput;
ASSERT(dip != NULL);
ASSERT(ip->i_d.di_nlink == 0); ASSERT(ip->i_d.di_nlink == 0);
ASSERT(ip->i_d.di_mode != 0);
/* setup for the next pass */ /* setup for the next pass */
agino = be32_to_cpu(dip->di_next_unlinked); agino = be32_to_cpu(dip->di_next_unlinked);
...@@ -3165,18 +3164,11 @@ xlog_recover_process_one_iunlink( ...@@ -3165,18 +3164,11 @@ xlog_recover_process_one_iunlink(
*/ */
ip->i_d.di_dmevmask = 0; ip->i_d.di_dmevmask = 0;
/* IRELE(ip);
* If this is a new inode, handle it specially. Otherwise, just
* drop our reference to the inode. If there are no other
* references, this will send the inode to xfs_inactive() which
* will truncate the file and free the inode.
*/
if (ip->i_d.di_mode == 0)
xfs_iput_new(ip, 0);
else
IRELE(ip);
return agino; return agino;
fail_iput:
IRELE(ip);
fail: fail:
/* /*
* We can't read in the inode this bucket points to, or this inode * We can't read in the inode this bucket points to, or this inode
......
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