Steve Lord wrote:
>
> OK, I had a problem on a system here doing some extreme testing, and
> had to attempt a repair on a root filesystem myself - from a remote
> location.
>
> I have managed to make xfs_repair -n run, here is what I did:
>
> 1. Boot the system single user, I used this:
>
> xfs-2.4.6 ro root=/dev/hda1 init=/bin/sh
>
> 2. remount the root filesystem read/write:
>
> /bin/mount -o remount,rw /dev/hda1
>
> 3. edit /etc/mtab and make sure that / shows up with a ro mount:
>
> /dev/hda1 / xfs ro 0 0
>
> 4. remount the filesystem readonly:
>
> /bin/mount -o remount,ro /dev/hda1
>
> You can now run xfs_repair -n on the filesystem, it will still not let
> you do an actual repair on it.
>
> Steve
>
> p.s.
>
> I faked the header here, so it will not show up as a followup.
I'm being delayed by a gnome problem in doing more testing. Basically
X11 died (hard lockup, no ping, magic sysrq dead) during a regular user
login, and now it is giving me security grief (it messed up something in
the Xauth stuff). I can no longer login my regular user with gnome, I
have to use KDE or something else other than gnome.
I see the advantage of the above approach is that it marks the drive ro
from the start, so inaccurate mtab listings will always be inaccurate in
the direction of marking ro. I'll add this as part of my experiment list
when I get gnome working right again.
D. Stimits, stimits@xxxxxxxxxx
|