Lilo fails with "sorry, don't know how to handle device 0xfe00" (LVM2)
Sorry, sending again in plain text. Also added note at the bottom about non-subscription.
From: Chris Carr
Sent: 07 February 2006 14:19
Subject: Lilo fails with "sorry, don't know how to handle device 0xfe00" (LVM2)
Hi folks - here is the system:
hda - 3GB IDE drive (all hda1)
hdb - ATAPI CDROM
hdc - 3.6GB IDE drive (all hdc1)
hda1 & hdc1 make LVM2 volume group VG1 (6.6GB). VG1 has four logical volumes:
VG1-Root 2.7G (my original system which I am trying to rescue)
VG1-NewRoot 1.5G (new Debian installation which won't work either)
What happened is this: everything was fine, I was booting into VG1-Root using lilo
(boot=/dev/hda, root=/dev/mapper/VG1-Root). This was under a 2.6.10 kernel. The
system was originally installed from the Debian installer (version 25 Jan 05) with a
2.6.8 kernel, but I compiled 2.6.10 some time last spring. I remember I had to
re-use the 2.6.8 initrd to activate the LVs because something was wrong with
initrd-tools at the time - but it was all working ok.
On Saturday I compiled a 188.8.131.52 kernel only to find that it didn't work because
devfs has been replaced by udev and my system relies on devfs to boot (something
to do with device mapper). I didn't have time to mess about with udev so I compiled
a new 2.6.10 kernel (identical to the existing one but with USB support added). I
stupidly forgot that it needed the old ramdisk, so I ran lilo without the initrd= line
and I now can't boot the system from HD.
So I burned Knoppix 4.0.2 and followed the instructions at www.knoppix.net/wiki/LVM2.
As an aside, I would STRONGLY URGE the maintainers of Knoppix to include
lvm-common and lvm2 on the Knoppix CD - they are only ~700k and are completely
essential for those of us using LVM2. My machine only has 64MB RAM so it ran
out of memory during the apt-get command, and I was nearly stuffed (until a friend
suggested using my USB keyfob as swap space).
I managed to activate and mount my LVs under Knoppix (kernel 2.6.12), and I edited
my VG1-Root/lilo.conf to restore the use of the initrd. But when I ran lilo it bombed
/proc/devices: fopen failed: No such file or directory
Is device-manager driver missing from kernel?
Warning: /proc/partitions does not exist: disc scan bypassed
Fatal: sorry, don't know how to handle device 0xfe00
... I ran lilo with "chroot /mnt/VG1-Root /sbin/lilo". 0xfe00 is device 254,0 which is in
/dev/mapper as VG1-Root - so far so good. But in /mnt/VG1-Root/dev/mapper the
major number is 253.
So I wondered if there had been some change to lilo and/or libdevmapper which caused
that mismatch. It seemed obvious to me that that's why it was failing - because my
original system had the LVs with major 253, and the Knoppix kernel thought they
should be 254.
So I downloaded the Debian Etch installer (20051111) and installed a completely new
system on VG1-NewRoot. The installer failed to install lilo (see bug#351768 if you're
interested) but I thought, never mind, I'll do it with Knoppix.
So I fire up Knoppix again (taking another ten minutes to do the apt-get install of lvm2)
and mount VG1-NewRoot - and sure enough, the dev/mapper entries there are 254,x
which looks good. So I run lilo again (chroot /mnt/VG1-NewRoot /sbin/lilo) and lo, I get
the same error! This time I really don't understand it, since ALL the major numbers are
254 (in /proc/devices and in /dev/mapper AND in /mnt/VG1-NewRoot/dev/mapper) so it
Can anyone tell me what's going on, and how to fix it? All versions of lilo are 22.6.1
released Nov 04 (the one in VG1-Root/sbin, the one in VG1-NewRoot/sbin, the one on
the Knoppix 4.0.2 CD and the one on the Etch installer CD). I can't find any reference
to this error (on google or Debian BTS) later than 2003, which is when lilo was patched
to work with libdevmapper (latest is
Sorry about the length - thanks for reading, and grateful for any help. I cannot subscribe
to this list from my work email address, and I currently have no email access at home,
so cc replies appreciated.
(also posted to comp.os.linux.setup)