Bug#401778: (installation-report: Successfull installation on sunblade 2000 with downloaded firmware plus video and initrd tricks)

[Daniel Dehennin]

> According to Maximilian this is probably in the mdadm part of the
> initramfs-tools scripts. Please consult with him if needed.

I don't think so, the mdadm initramfs script don't know that is must
wait for the devices to be up by the firmware.

Sometime, the initialisation of the firmware is slow enough for the
udev script to finish before the /dev/sd[ab] are created.

The init script then runs local-top after mounting the root
filesystem, mdadm is run but the hard drive devices are not present.

Then, when the firmware is ok, it send events and udev create the
devices in /dev but it's too late, the system hang for ever, need to
reboot manually.

It can be the same without the use of mdadm, the system can try to
mount the root files system from a device that doesn't exist.

Maybe the initramfs-tools can integrate an optional script which can
be activated just before local-top (after udev) ?

A parameter could configure the devices the script must wait for

I'm not sure it's mdadm related.

> I'm also a bit confused about the need for the extra parameters the
> submitter added in silo.conf
>     append="md=0,/dev/sda2,/dev/sdb2,md=1,/dev/sda3,/dev/sdb3
> AFAIK mdadm should take care of that by itself.

Right, I remove this from my silo.conf, it was in a howto I found but
it's useless.

Regards.
-- 
Daniel 'NebuchadnezzaR' Dehennin
Récupérer ma clef GPG:
gpg --keyserver pgp.mit.edu --recv-keys 0x2A408F69