Bug#566192: initramfs-tools: md0 device not recognized at boot after upgrade to kernel 2.6.30

Mon Feb 1 21:10:36 UTC 2010

Hello Martin

Please find here the report of what I did about this issue. In one word, my
installation was missing the mdadm package...

I compared both initrd.img-2.6.26-2-686 and initrd.img-2.6.30-2-686. I used
cpio -t < initrd.img-2.6.[26-30]-2-686  > content[26-30], sorted each
content file, edited both file to remove 2.6.26 or 2.6.30 and compared it
using diffuse...

I found quite a few differences in the modules loaded... Concentrating only
on the module related to md, I modified the /etc/initramfs-tools/modules to
include all the modules present in initrd.img-2.6.26-2-686 and NOT present
in initrd.img-2.6.30-2-686

/etc/initramfs-tools/modules now reads

# List of modules that you want to include in your initramfs.
#
# Syntax:  module_name [args ...]
#
# You must run update-initramfs(8) to effect this change.
#
# Examples:
#
raid0
raid10
raid456
multipath
raid1
md-mod
linear

I created a new initrd.img-2.6.30-2-686 by running # update-initramfs -c -k
2.6.30-2-686.. Same problem...

Time to compare more in detail...

I also noted some other differences that related to the use of my raid
devices. My initrd.img-2.6.30-2-686 is missing
conf/conf.d/md
sbin/mdadm
script/local-top/mdadm
unrelated to my problem because I do not use it, I also noted that
script/local-top/lvm was missing too

Time to think now... Where is that sbin/mdadm coming from...
The command # find / -name mdadm finds NOTHING...
Synaptic shows that mdadm is NOT INSTALLED...
I do have a raid 5 for /home and a mirror for / and the swap.. but I have no
mdadm???

I reinstalled my system in octobre using the official DVD for stable, using
the non expert mode with questions and answer, the only exception being that
I did NOT let the installation initialize and format my /home on a raid5.
After the installation I modified synaptic to use the test. I am the only
one to use this machine and the only time I go into root is when I run
synaptic to follow the evolution of the test Debian... I do not recall that
I ever played with mdadm since the installation... (Was it ever
installed???, was it removed by synaptic at some stage)? I am sure that I
never tried to remove it as I have plenty of space on my disks...

All that is VERY VERY STRANGE :-) (Computer is no magic... Most of the time)

I installed mdadm via synaptic, restored /etc/initramfs-tools/modules, ran
update-initramfs -c -k 2.6.30-2-686 that gave me an error message (Known
issue I think)
update-initramfs: Generating /boot/initrd.img-2.6.30-2-686
W: mdadm: You appear to be missing /dev/md/2 from /etc/mdadm/mdadm.conf !
W: mdadm: If this array is required to boot, you must add the following
line:
W: mdadm: ARRAY /dev/md/2 UUID=2e58b3c7:72f2231e:f25f8b76:ced30a76

I ran zcat initrd.img-2.6.30-2-686 | cpio -t | grep md and found all needed
modules and quite a few expected files:
lib/udev/rules.d/64-md-raid.rules
lib/modules/2.6.30-2-686/kernel/drivers/ide/amd74xx.ko
lib/modules/2.6.30-2-686/kernel/drivers/ide/cmd64x.ko
lib/modules/2.6.30-2-686/kernel/drivers/ide/cmd640.ko
lib/modules/2.6.30-2-686/kernel/drivers/md
lib/modules/2.6.30-2-686/kernel/drivers/md/linear.ko
lib/modules/2.6.30-2-686/kernel/drivers/md/raid0.ko
lib/modules/2.6.30-2-686/kernel/drivers/md/raid10.ko
lib/modules/2.6.30-2-686/kernel/drivers/md/raid456.ko
lib/modules/2.6.30-2-686/kernel/drivers/md/multipath.ko
lib/modules/2.6.30-2-686/kernel/drivers/md/raid1.ko
lib/modules/2.6.30-2-686/kernel/drivers/md/raid6_pq.ko
lib/modules/2.6.30-2-686/kernel/drivers/md/md-mod.ko
lib/modules/2.6.30-2-686/kernel/drivers/char/agp/amd-k7-agp.ko
lib/modules/2.6.30-2-686/kernel/drivers/char/agp/amd64-agp.ko
lib/modules/2.6.30-2-686/kernel/drivers/video/console/mdacon.ko
35415 blocs
etc/mdadm
etc/mdadm/mdadm.conf
conf/conf.d/md
scripts/local-top/mdadm
sbin/mdadm

I tried to boot... the boot process now goes past the point I was blocked
at, but stops further...The last message ( I recorded on a paper and typed
it here by hand) is
[21.588641] sd 0:0:2:0: [sdc] attached SCSI disk
I did not touch the machine for 60 seconds at my watch, then issued
Cont-alt-del and got something like (only 2 sec before the screen vanishes)
stopping all md disks   followed by a list with md0 md1 md2

I guess I have hit another bug here... If and when I have more time, I will
maybe have a deeper look :-)... There is a bug report (#549287) that seems
to imply udev or mdadm... How do I go about removing a package and reinstall
an old one... (Some more reading in perspective, unless I find a cook book
about that on google...) :-)

Anyway, thanks for your attention. I guess you can close this bug or learn
from it that one should make sure to verify if mdadm is present before to
try the kernel update on a raid disk.

Regards.

    Louis Richard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-mdadm-devel/attachments/20100201/511ddea8/attachment.htm>