Bug#733574: partially breaks initramfs when modules are not included

[Eduard Bloch]

Hallo,
* Michael Tokarev [Tue, Dec 31 2013, 09:08:53PM]:

> >  - without md* modules
> >  - where another initramfs script exists in local-top which is required
> >    to boot (say: for cryptoroot)
> > 
> > Result: startup fails.
> 
> Well. I'm not sure it is a valid combination.  If you're using mdadm,
> why do you have kernel without md support?  Either remove mdadm, or
> fix your kernel config.  I can move this test into mdadm initramfs
> creation hook, so it will prevent creating initramfs if md can't be
> used in there, -- this way, you'll see the problem much sooner than
> when booting actually fails.

I don't use it. It was just installed for historical reasons.
Furthermore, the rootfs is not on an MD array, therefore the initramfs
script should not even TRY to configure them.

> Arguable, this `exit 1' can be removed entirely and we can just run
> mdadm to try to do its work anyway.  But the root question stands
> still: why to use invalid combination in the first place?

Even invalid combinations should be handled in a delicate way when it
comes to the boot process.

Regards,
Eduard.