Is anyone using XFS with LILO to boot off?
I've been using XFS for a short while now, but always
just on the big data
partitions on various servers I've built, keeping root,
/usr on ext2 so I
can use my debian recovery CD if ever required, but this
time I thought
I'd try an all-XFS system and have been having major
headaches trying to
get it to boot off an XFS partition.
Heres the scoop:
create a raid1 set, mkfs -t xfs it, then mount it
under /mnt
copy / onto it via cd / ; find -xdev | cpio -pm /mnt
Edit /mnt/etc/fstab and /mnt/etc/lilo.conf to do the
right things
run lilo -r /mnt
This is something I've done many times to get / onto a
raid0 device with
ext2. (Debian install doesn't support this directly -
yet)
At this point, the XFS is corrupt on it:
lion:/# umount /mnt
lion:/# mount /dev/md1 /mnt
mount: you must specify the filesystem type
lion:/# mount -t xfs /dev/md1 /mnt
mount: wrong fs type, bad option, bad superblock on
/dev/md1,
or too many mounted file systems
lion:/# xfs_check /dev/md1
xfs_check: unexpected XFS SB magic number 0xfae84200
bad superblock magic number fae84200, giving up
Is Lilo incompatable with XFS?
Complete sequence of events:
lion:/# uname -a
Linux lion 2.4.21-ac4 #3 Sun Jul 27 18:48:19 BST 2003
i686 unknown
lion:/# cat /etc/issue
Debian GNU/\s 3.0 \n \l
lion:/# cat /usr/local/src/xfsprogs/VERSION
#
# This file is used by configure to get version
information
#
PKG_MAJOR=2
PKG_MINOR=3
PKG_REVISION=9
PKG_BUILD=0
lion:/# mkfs -t xfs -f /dev/md1
meta-data=/dev/md1 isize=256 agcount=8,
agsize=31121 blks
= sectsz=512
data = bsize=4096
blocks=248968, imaxpct=0
= sunit=1 swidth=2
blks, unwritten=0
naming =version 2 bsize=4096
log =internal log bsize=4096
blocks=1200, version=1
= sectsz=512 sunit=1
blks
realtime =none extsz=8192 blocks=0,
rtextents=0
lion:/# mount /dev/md1 /mnt
lion:/# find . -xdev | cpio -pm /mnt
240549 blocks
lion:/# vi /mnt/etc/lilo.conf
lion:/# vi /mnt/etc/fstab
(in lilo.conf:
boot=/dev/md1
root=/dev/md1
raid-extra-boot=/dev/hda,/dev/hdc
in fstab:
/dev/md1 / xfs errors=remount-ro
0 1
lion:/# lilo -r /mnt
Warning: using BIOS device code 0x80 for RAID boot blocks
Added Linux *
Added Linux-1
Added Linux-2.4.21-0
Added Linux-2.4.18
Added Linux-Orig
The boot record of /dev/md1 has been updated.
The boot record of /dev/hda has been updated.
Warning: /dev/hdc is not on the first disk
The boot record of /dev/hdc has been updated.
lion:/# ls /mnt
archive boot dev floppy initrd lost+found opt
root tmp var
bin cdrom etc home lib mnt proc
sbin usr
lion:/# umount /mnt
lion:/# mount /dev/md1 /mnt
mount: you must specify the filesystem type
lion:/# mount -t xfs /dev/md1 /mnt
mount: wrong fs type, bad option, bad superblock on
/dev/md1,
or too many mounted file systems
I don't think that raid1 has anything to do with it, as I
can duplicate
the results exactly when I stopped the RAID and just used
/dev/hda2 on its
own. I've also tried it the other way round - I moved
root onto /dev/md1,
(hda2, hdc2) and tired to make it work on the very first
partition - both
with raid1 and just a single drive and I get the same
results every time.
I feel I'm missing something obvious, but according to
http://www.tldp.org/HOWTO/Linux+XFS-HOWTO/x154.html
and
http://oss.sgi.com/projects/xfs/xfsroot.html
it ought to "just work".
Any clues or observations would be appreciated
Thanks,
Gordon