[linux-lvm] Re: Upgrade from 2.6.10-ac8 to 2.6.12.5 broke lvm rootfs
Ville Herva
vherva at twilight.cs.hut.fi
Fri Jan 20 11:03:26 UTC 2006
On Wed, Aug 17, 2005 at 04:30:11PM +0300, you [Ville Herva] wrote:
> [repost: first post did not show up. sorry for the possible dupe]
>
> After upgrading the kernel from 2.6.10-ac8 to 2.6.12.5 the initramfs was no
> longer able to mount rootfs.
>
> mount: error 6 mounting ext3
>
> All the configuration options are identical, and upgrading lvm2 package:
> lvm2-2.00.25-1.01 -> lvm2-2.01.14-1.0
> device-mapper-1.00.19-2 -> device-mapper-1.01.04-1.0
>
> Did not change anything.
>
> Dm, ext3 and the relevant block device drivers statically compiled in.
>
> The vg has lvm1 format, fwiw.
>
> I enabled all the debug options I could think of in the nash-based initramfs
> init script. That did not appear to tell much: all I was able to tell was
> that lvm was succesfully called by the init script:
>
> mount -t proc /proc /proc
> mount -t sysfs none /sys
> insmod /lib/dm-snapshot.ko
> mkdevices /dev
> mkdmnod
> lvm vgscan -v
> # sleep 5
> lvm vgchange -ay
> # sleep 5
> lvm vgmknodes
> # sleep 5
> mkrootdev /dev/root
> umount /sys
> # sleep 5
> mount -o defaults --ro -t ext3 /dev/root /sysroot
> switchroot /sysroot
>
> but those didn't give any meaningful output (other than notices about
> setting log indentation level).
>
> Finally, I added "sleep 5" after each lvm command (commented out above),
> which appeared "solve" the problem.
>
> Apparently the lvm scripts somehow do their initialization asynchronously
> and the init script tries to mount root before it is available. I'm not sure
> why this is affected by the kernel version, though.
This happens with linux-2.6.15.1, lvm2-2.01.14-1.0 and mkinitrd-4.1.18-2.
I must manually patch the init script inside the initrd image to have those
sleep 5
lines after each and every lvm command, otherwise I get the
mount: error 6 mounting ext3
error.
Any ideas?
-- v --
v at iki.fi
More information about the linux-lvm
mailing list