On Tue, 2002-09-10 at 13:22, Chris Wedgwood wrote:
> On Tue, Sep 10, 2002 at 03:58:14PM +0200, thomas wrote:
> > i've got a RAID1 running that seems to not umount itself properly. so
> > when i try to mount the system after a reboot i see:
> > now i don't have /dev/md0 in fstab just to be sure that it
> > doesn't try to mount itself before the raid array is
> > started. but it doesn't matter. whatever i try the only thing
> > that helps is xfs_repair -L /dev/md0, after that it mounts
> > without any problems.
> Is this / on raid1?
> If so, I also see this... I put some stack dump stuff in to verify
> that /dev/md0 get opened *twice* as / is mounted, once I assume for
> the auto-raid stuff (or something?) and then again to mount /
> The raid1 code won't switch to RO in there is more than one reference
> and thus it's never able to be mounted cleanly.
> Sorry this doesn't provide a solution, I was tracking this down under
> time pressure and had to fix the problem using another means.
Are you doing this with current cvs code? The xfs log used to
effectively open the device again. I think this has been reworked
in the current code to avoid the double open.
I have not checked into this, so I might be wrong.
Steve Lord voice: +1-651-683-3511
Principal Engineer, Filesystem Software email: lord@xxxxxxx