Red Hat Bugzilla – Bug 239522
mkinitrd copies only 20 chars from root=
Last modified: 2007-11-30 17:12:04 EST
Description of problem:
After at some point the kernel stopped booting with root=LABEL=/ I switched to
/dev/VG12345/LVroot1 (I have FC6 too). When a new kernel was available and I
rebooted the new one loaded my FC6 (/dev/VG12345/LVroot).
Version-Release number of selected component (if applicable):
always, for the last 3 kernels I installed at least.
Steps to Reproduce:
1. install F7
2. edit grub.conf so root=/dev/VG12345/LVroot1 instead of root=LABEL=/
3. update the kernel rpm
4. look at grub.conf
the root device gets truncated to 19 characters.
Using device paths instead of labels should work, haven't tested with LABEL>19
mount shows me that I have /dev/mapper/VG12345-LVroot on / type ext3 (rw), but
that is not true. Maybe that's what confuses mkinitrd...
My grub.conf starts with:
# grub.conf generated by anaconda
# Note that you do not have to rerun grub after making changes to this file
# NOTICE: You have a /boot partition. This means that
# all kernel and initrd paths are relative to /boot/, eg.
# root (hd0,1)
# kernel /vmlinuz-version ro root=/dev/VG12345/LVroot1
# initrd /initrd-version.img
and every time new kernel is installed I get:
kernel /vmlinuz-version ro root=/dev/VG12345/LVroot
(no 1 at the end)
The option root=LABEL=/ now works, didn't with one kernel, and I have to
modify this line every new kernel (even xen one) gets installed.
I'm changing even the comment to see what will happen next time...
PS: I have /dev/VG12345/LVroot and it is FC6, while /dev/VG12345/LVroot1 is F7.
Ops. I searched in /etc and I think I found the problem. When one of the
development kernels did not boot I changed all LABEL=/ with /dev/VG (and typed
LVroot instead of LVroot1 for '/')... After that I have forgotten to revert
/etc/fstab, and that's from where new kernels have been getting their root=
option. My '/' entry in fstab was wrong (LVroot, not LVroot1), but that seems to
make no difference anywhere else ?!? Even mount shows LVroot, but the real
contents is from LVroot1 ?!? cat /proc/mounts is correct of course.
It is clear that the problem is not what I thought, but have no idea what to do
with this bug report now.
In short, my fault - not a bug.