Re: Installation reporting on HP Proliant ML350 with Smartarray

[Francesco Paolo Lovergine <frankie@debian.org>]

On Sat, Dec 13, 2003 at 10:31:29PM -0500, Joey Hess wrote:
> I apologise for the delay in responding to your installation report. As
> far as I can see nobody else did, so here goes..
> 
> Francesco P. Lovergine wrote:
> > Some problems in installation:
> > 
> > * d-i does not recognize ethernet (tg3) and controller (cciss for 641 series)
> > 	I loaded by hand modules and proceeded. I could send pci ids
> > 	in a couple of days.
> 
> It'd be good if you'd send those pci ids, this should be easy to fix.
> 

02:02.0 Class 0104: 0e11:0046 (rev 01)
        Subsystem: 0e11:409a

See also:

http://pciids.sourceforge.net/iii/?i=0e110046

for other subsystems.

> > * kernel installation failed
> > 	The problem is that the root /dev/cciss/disc/disc0/part1 path is not
> > 	correctly managed by mkinitrd and it creates a 0byte file.
> > 	The installation succeded by hand using a traditional
> > 	device name -r /dev/cciss/c0d0p1, customizing in chroot /target
> > 	the /etc/mkinitrd/modules to add cciss support, creating
> > 	lilo.conf by hand (with non devfs device names) and running 
> > 	lilo. 
> 
> Sounds like you have a good handle on what was broken and why, so why
> don't you file a bug report on the kernel-image-2.4.22-i386 package (or
> whatever kernel package it installed) with the full details of what
> mkinitrd does wrong.
> 

I cannot re-run the installer now, the problem related to devfs/mkinitrd.
I think mkinitrd has a poor management of devfs remapping to standard devices.
In fact /dev/cciss/disc/disc0/part1 -> /dev/cciss/c0d0p1
is not so evident. Probably a mkinitrd bugreport is more adequate.

> > 	Also, manual disk partition failed and presented a
> > 	strange list of choices:
> > 
> > 	.
> > 	.
> > 	0
> > 	0
> > 	1
> > 	1
> > 	/
> > 	/
> > 	Finished
> > 
> > 	or so.
> 
> I think it's having a hard time converting the smartarray's nonstandard
> devices into the easily readable format it tries to use. Can you file a
> bug against the "partitioner" package?
> 

I'll do both soon.

-- 
Francesco P. Lovergine