Re: Custom debian-installer kernel - was: Re: Up cl

Hi Christian,

On 1/3/26 01:08, John Paul Adrian Glaubitz wrote:
> On Fri, 2026-01-02 at 22:01 +0100, Christian Groessler wrote:
>> Using the same kernel config and adding a --ramdisk in palo.conf (using
>> install/initrd.gz from 2025-12-06 CD) lets the kernel crash early, like
>> the other kernels I've tried. No further output after the palo loading
>> messages.
>>
>> So it crashes right at the start of the kernel, unlike without an initr=
d
>> where the kernel boot messages appear normally and it crashes because i=
t
>> has no rootfs and init program.
>=20
> Maybe debugging with earlyprintk can help?
>=20
> Try passing "earlyprintk=3Dserial,ttyS0,115200" to the kernel command li=
ne.
>=20
> See: https://wiki.gentoo.org/wiki/Kernel/Command-line_parameters

You mentioned, that "your" kernel does not crash, and that you run
on serial console.
Can you please capture the full output and post it here?
Then maybe I'm able to see what the difference between your machine
and all others is.
Do you maybe have some PCI cards, e.g. graphic cards or others added ?

You mentioned that the standard kernel failed to detect your CDROM drive.
In most (somewhat newer) HP machines the CDROM drive was ATA connected
to on-board controllers.

>> Then  I tried extracting the initrd.gz file system (replacing lib/modul=
es
>> with my kernel's version) and mounting it over NFS. I modifed the kerne=
l
>> config to compile in NFS client support and not get it from modules.
>=20
> Would be good to figure out what the actual problem is so we can fix the
> issue in Debian or in the upstream kernel.

Yes, for that the log would be good.

>> This worked so far that I could start the installer (it would find the
>> CD in the CD drive as installation media), but it complains that the
>> installer components (I believe kernel modules) are for a different
>> kernel than the booted version.
>=20
> Yes, that's expected.
> ...

Helge