Re: MD device not found on boot
04/03/2012 02:27, Daniel Bareiro wrote:
> On Saturday, 03 March 2012 10:43:52 +0100,
> tv.debian@googlemail.com wrote:
>
>>>> Are you using kernel-package "make-kpkg" or the "make deb-pkg"
>>>> target to build your kernel packages ? The former requires a bit of
>>>> manual configuration to work properly.
>
>>> I'm using make-kpkg:
>>>
>>> # cp /boot/config-`uname -r` .config
>>> # make menuconfig
>>> # make-kpkg --initrd --append_to_version=-layer7-imq-amd64 linux-image
>
>> What steps did you take to configure kernel-package ? Did you read
>> /usr/share/doc/kernel-package/README.gz ? Just a short citation:
>>
>> "With the new kernel-package conventions, you also need the example
>> scripts in /etc/kernel/postinst.d/ and /etc/kernel/postrm.d, to create
>> and remove the initramfs. Make sure that these scripts pay attention
>> to the INITRD env variable to determine whether or not to take any
>> action.
>>
>>
>> Let me repeat:
>> Since nothing is created automatically. you need to provide a hook
>> script for things to happen when you install the kernel image
>> package. The user provides such scripts. For example, to invoke
>> mkinitramfs, I did:
>>
>> --8<---------------cut here---------------start------------->8---
>> cp /usr/share/kernel-package/examples/etc/kernel/postinst.d/initramfs \
>> /etc/kernel/postinst.d/
>> cp /usr/share/kernel-package/examples/etc/kernel/postrm.d/initramfs \
>> /etc/kernel/postrm.d/
>> --8<---------------cut here---------------end--------------->8---
>> "
>>
>> Maybe that's part of the problem.
>
> Here I am not physically in front of the host to be tested, however I
> installed Debian GNU/Linux 5.0.9 Lenny in a KVM virtual machine in order
> to replicate the problem.
>
> I was reading the file /usr/share/doc/kernel-package/README.gz on the
> KVM virtual machine, but I didn't found the quote you mentioned, nor the
> two files.
Sorry, my bad, I forgot that kernel-package undertook a major overhaul
after Lenny, so it's no wonder why you can't find the quote, it doesn't
exist in your version. With Lenny kernel-package everything is supposed
to be configured automatically at install time.
However I tried the kernel compiled with the above steps
> (without additional steps) and I could not reproduce the problem on my
> virtual machine. i.e. it boots without problems. One test failed with a
> message that the initrd was too big, but then to expand the RAM of the
> VM, it boots without problems.
>
> So now the issue is finding the difference between these two scenarios.
>
> 1) Debian GNU/Linux Lenny 5.0.8:
>
> * 2.6.32-bpo.5-amd64 boots without problems.
> * 2.6.32-layer7-imq-amd64 fails to find the MD devices.
> * The environment is the same, so there were no changes in settings
> or metadata, which makes me think that the difference is something
> on the compiled kernel. But...
>
> 2) Debian GNU/Linux Lenny 5.0.9 (KVM):
>
> * 2.6.32-bpo.5-amd64 boots without problems.
> * 2.6.32-layer7-imq-amd64 compiled using the same procedure as in (1)
> boots without problems. This makes me think that then the problem
> is not the procedure used to compile.
>
> Then the problem is due to a difference between 5.0.8 and 5.0.9? But, if
> so, why precompiled kernel does not produce problems and the manually
> compiled kernel fails using a procedure that did not bring problems in
> another scenario?
Looks like you are triggering a corner case the 5.0.8, and maybe one of
the 5.0.9 updates fixed that ? I guess you checked that all
configuration files (/etc/kernel-img.conf, /etc/initramfs-tools/modules
etc...) are the same on the two hosts. Checked the changelogs for mdadm,
initramfs-tools and kernel-package between 5.08 and 5.09 ?
The virtual machine hardware could also influence the outcome, if you
could image the non working host and test it with the exact same setup
you could bisect the origin of the problem more easily. Could be in
initrd generation, but also simply in the boot loader or devices
addressing (something changed in grub in between 5.08 and 5.09 ?). The
problem occurs when the initrd is looking for the root fs on /dev/md2,
when dropped to a shell did you try to see if /dev/md2 really doesn't
exist, or is named differently. Can it be assembled in initrd shell and
the boot process resumed ?
Anyway, even if you find a bug in Lenny 5.0.9 (and all the more in
5.0.8) in initramfs-tools, mdadm or another essential package, it is not
going to be taken care of since Lenny reached it's end of life. Maybe
it's time for an upgrade ?
>
> Thanks for your reply.
>
>
> Regards,
> Daniel
Reply to: