Re: Pre-bug-raising questions
2009/1/19 Mike Grice <mgrice@plus.net>:
> On Mon, Jan 19, 2009 at 12:46 PM, Bastian Blank <waldi@debian.org> wrote:
>> On Mon, Jan 19, 2009 at 11:52:59AM +0000, Mike Grice wrote:
>>> I'm pleased to say that this gets further than ever before:
>>> - it fails to detect the network device but i'm able to select the
>>> correct one from a list.
>>> - it fails to detect the disk device but i'm able to select the
>>> correct one from a list.
>>
>> Hmm. Which list do you mean?
>
> Ok. From a boot net:dhcp, the installer gives me:
>
> Select a language: English
> Choose Language: default
> Detect Network Hardware:
> "No Ethernet card was detected. If you know the name of the driver │
> │ needed by your Ethernet card, you can select it from the list."
>
> I'm then given a list of all the network drivers available. I can
> scroll down to the 'sunvnet' driver, and the installer then carries on
> as normal.
>
> I'm asked for a hostname and domain, which I provide. I select United
> Kingdom as the Debian archive mirror. I select the default mirror for
> that country 'ftp.uk.debian.org'. I provide HTTP proxy information,
> which for my network is http://192.168.107.1:3128 (basic squid
> install).
>
> The installer then 'loads additional components', which I think
> happens over the network. It mentions contacting a time server. I'm
> then on the 'detect disks' menu, which tells me:
> " │ No disk drive was detected. If you know the name of the driver
> needed │
> │ by your disk drive, you can select it from the list."
>
> I can scroll down to the 'sunvdc' driver, and the installer then
> carries on as normal.
>
> Those are the two menu things I referred to before.
>
> BTW, if I choose guided partitioning, it will select ext2 by default
> for the boot partition (I change this, but just thought you would like
> to know as its a bit odd, maybe a legacy SILO hangover).
>
>>> The problem I get now is that the
>>> kernel the installer installs as part of the setup is the etch-n-half
>>> kernel, which doesn't have the required drivers installed. So I'm
>>> stuck at 'waiting for root filesystem...' until I drop to the initrd,
>>> and I can't find the module to install.
>>
>> The daily installers defaults to installing unstable. However there is
>> no etch'n'half-kernel in unstable. So you override that decision
>> somewhere.
>
> Actually, I just checked this time from a clean install and its picked
> the new kernel without prompting. As I was messing with the network
> parameters the first time perhaps it defaulted to the older kernel?
> Either way, that part is a non-issue now. I will provide the install
> syslog (thanks for that idea FP) if you need it for diagnosis later...
>
>>> This time, the debian-installer installs a better sounding kernel
>>> (2.6.26-1-sparc64-smp), but I'm dumped to the initrd shell again.
>>> Looking through the filesystem, the modules are definitely there and I
>>> can insmod them and continue the install.
>>
>> Okay, so its a problem somewhere between udev and the kernel
>> definitions.
>
> Ok. Is this something I can fix myself without rebuilding the
> installer? (e.g., can I preseed it)?
>
>>> Is there anything I'm missing here? I can tell something is 'just a
>>> bit missing' and I'd love to be able to raise a correct bug report for
>>> this and get it squashed before Lenny.
>>
>> Add the modules to /etc/initramfs-tools/modules.
>
> Is that something I need to put in the bug report as a fix for this issue?
Adding those modules to that file and running update-initramfs -u does
fix this issue. I have the following problem, though:
The boot sequence now identifies all the modules gets as far as the
console login but my keystrokes don't actually get there (I can mash
the keyboard and no characters are echoed back, or responses given
from the host). This is from the 'console' as far as connecting over
telnet (which works for doing the install).
If I break out of the installer to the shell and make sure ssh-server
is running as part of the install, I can get to the box at the time
the login prompt is displayed and use the box as normal, so it's just
a console problem...
So I guess the salient points from all the above are:
1) the debian-installer is not detecting the correct modules for the
hdd and net even though theyre present in the udebs
2) the initramfs-tools config on the installed system doesn't load the
relevant modules by default
3) the text console doesn't work once the install is finished and the
system is up as normal
Thanks again,
Mike
Reply to: