Re: trying to repair/recover xfs filesystem after system crash

[Nathan Scott <nathans@sgi.com>]

On Fri, Aug 08, 2003 at 12:18:07AM -0500, djoneill@gmx.net wrote:
> On Fri, Aug 08, 2003 at 01:14:33PM +1000, Nathan Scott wrote:
> > On Fri, Aug 08, 2003 at 04:57:39AM +0200, onedj@gmx.net wrote:
> > > A couple weeks ago my system powered down suddenly.  One XFS filesystem was
> > > affected somehow and I have not been able to repair it with xfs_repair. 
> > > xfs_repair immediately freezes when I run it and gives no output.  The same is
> > > true for xfs_check.  xfs_info gives the following:
> > 
> > Do you get any syslog errors when running repair/check (eg.
> > device errors?).  They shouldn't just hang... what do they
> > display?  If nothing at all, strace output would be useful.
> 
> No output whatsoever to syslog or messages.  Here is the strace output:
> 
> 
> think# strace xfs_repair /dev/hda4
> execve("/sbin/xfs_repair", ["xfs_repair", "/dev/hda4"], [/* 21 vars */])
> = 0
> ...
> stat64("/dev/hda4", {st_mode=S_IFBLK|0660, st_rdev=makedev(3, 4), ...})
> = 0
> stat64("/dev/hda4", {st_mode=S_IFBLK|0660, st_rdev=makedev(3, 4), ...})
> = 0
> ustat(0x304, 0xbfffe724
> 

That's pretty wierd - looks like ustat is not returning from the
kernel... huh?  xfs_repair is sitting in libxfs code which tries
to determine whether there is a filesystem mounted on the device,
I've never seen anything like this hang before (its unlikely to
be an XFS problem, we're in VFS code there, in the kernel).

cheers.

-- 
Nathan