On Fri, Apr 30, 2004 at 07:38:53PM +0200, Karol Kozimor wrote:
> Hi,
Hi there,
> I suffered from major filesystem corruption while testing third party
> patches. The filesystem shut down and was handled with xfs_repair. After
> that, several directories were found in /lost+found that could no longer be
> removed, e.g.:
> # ls -la /lost+found/12616156/
> total 8
> drwxrwxrwx 2 root root 4096 2004-04-29 15:22 ./
> drwxr-xr-x 12 root root 155 2004-04-30 10:53 ../
> # rmdir /lost+found/12616156
> rmdir: `/lost+found/12616156': Directory not empty
> # # rm -rf /lost+found/12616156
> rm: cannot remove directory `/lost+found/12616156': Directory not empty
> ...
> rmdir("/lost+found/12616156") = -1 ENOTEMPTY (Directory not empty)
>
> Subsequent xfs_repair runs do not fix the problem, the directories are
> simply unlinked and relinked in /lost+found.
This is a known bug in xfs_repair.
> Was it ext2, I would have used debugfs. What can I do to rectify this and
The XFS equivalent is xfs_db(8) and you should be able to clear
these broken inodes with that tool, then re-run xfs_repair.
> / or provide any useful information?
Can you open a bug on oss.sgi.com and add in the information
you provided above, xfs_info output, and xfs_bmap -vv on one
of the inodes in question.
thanks.
--
Nathan
|