Bug#755119: debian-installer: wrong numeration of dmraid device
On 17 July 2014 23:11, Dima Veselov <kab00m@lich.phys.spbu.ru> wrote:
> Package: debian-installer
> Severity: normal
> Tags: d-i
>
> Bug was caught on preseeding Debian 7.6.0/i386, but I beleive it will
> work for any other installation on dmraid enabled box. In my case
> dmraid=true was added to kernel options of latest netboot kernel.
>
> Installer insert dmraid module and create device, say /dev/mapper/asr_raid
> When partitions are created this device split into /dev/mapper/asr_raidp1
> asr_raidp2 and others.
>
> I have two v60x servers with AIC-7902 HostRAID and they behave different.
>
> In the real kernel booting devices will be named /dev/mapper/asr_raid1
> asr_raid2 and others. (without p). So initramfs will fail to boot if
> d-i will use device names as boot names. this is first bug and it is
> most serious. As a workaround I used ln -s asr_raid2 asr_raidp2,
> booted, refreshed fstab and update-grub.
>
> On the other server d-i used UUIDs as boot names and it fail to load
> because kernel find and link /dev/s* devices to UUID devices, then
> load dmraid, which make sd devices busy and initramfs fail to load
> root (by the way it will still has asr_raidp device in resume
> environment variable). As a workaround I edited grub boot list,
> booted, edited fstab and update-grub. This is bug number two.
>
> Server intallations differs: first used dmraid=true in kernel boot
> options, second had 'd-i disk-detect/dmraid/enable boolean true'
> in preseed file. Don't know if it make sense.
>
> Also initramfs complain about dm_raid45 module and I don't know
> if its a bug number three or not.
>
I believe both of these problems (wrong enumeration & dm_raid45
module) can be solved by me uploading newer version of dmraid into
Debian. I still hold ITA on that package and should act upon it.
After it lands in unstable & testing, a minimal stable backport can be provided.
--
Regards,
Dimitri.
Reply to: