fc5 seek errors - nash mount (built-in)

Peter Jones pjones at redhat.com
Wed Mar 29 20:31:12 UTC 2006


On Wed, 2006-03-29 at 09:45 -0500, James Olson wrote:
> # cd /bin
> # mv mount mount.jim  /* rename /bin/mount so nash will have to use it's own */

FWIW, you an call nash's as "nash-mount" instead of this.

> # nash
> (running in test mode).
> Red Hat nash version 5.0.32 starting
> mount -t ext3 /dev/sdb3 /mnt/sea23
> mount -t 'ext3' '/dev/sdb3' '/mnt/sea23' ()
> 
> This generates a whole bunch of seek errors on my dmraid drive (hde).
> The internet nash mount function is definately causing the seek errors
> (even in test mode).   If you let it find and use /bin/mount it works
> fine and mounts with no seek errors on any drives (only with nash
> --force option, test mode just prints how it parsed the mount command
> line).

Yup, that's it.  nash's mount tries to populate the whole blkid.tab when
you tell it to mount something by device name, just like mount(8) does.

So yeah, that probably needs to be fixed.  Longer term (hopefully FC6)
we want to be only probing disks we expect to use, by first checking the
hardware unique identifiers, and then only checking other drives as a
last resort.

Short term I'm not sure what the best answer is.  Arguably, we should be
assembling all of the dmraids very early on; before we ever look for
filesystems anywhere at all.  But initrd generally tries to only
touch / , which is contrary to that.

Lemme think on this for a day or two, and see how far away the uuid
scanning plan really is.

-- 
  Peter




More information about the Ataraid-list mailing list