So, I finally got the system booting at least, with the edd=off option. Now, on to the next problem....

The system:
The system is an older IWill KK266-R (VIA KT133A chipset) motherboard with an Athlon XP 1600+ processor. It has an integrated CMD649 RAID controller. The system also has an Intel PRO/100 network adapter, and a SiI 3512A SATA disk controller. But they don't seem to be part of the problem.

The KT133A chipset has its quirks, but I believe they've been ironed out since long, and indeed, this machine has been running Linux for a long, long time in the past, happily serving files.

The OS currently installed is the Annvix 3.0 distro, updated with the latest kernel (2.6.22-something I believe).

The CMD649 controller has 4 disks connected to it, and it shares IRQ line with the mo/bo USB controllers (auto. set to #10 in BIOS).

The problem:
At first, the system would get to the fsck part, spit out "Disabling IRQ #10", and die. With a 100% certainty. Using 'irqpoll' got me past this, but the system were utter crap. So I disabled the fsck from /etc/fstab to at least get it through to a workable system without 'irqpoll'. Now, the system might work for a while, but will eventually disable the IRQ yet again, and render the system unusable. By now I've tried so many options and things, I don't even remember. Different kernel options, lik 'noapic', 'acpi=off', blah, blah.. Changing the assigned IRQ # assignments.. Nothing helps!

Actually, I first witnessed problems with this machine some 6 months ago which might've been the same problem. The boot/root drive with the old Gentoo installation crashed (a system than had been running perfect for over a year). So I thought I'd go for an 'easier' distro this time, and installed Xubuntu 7.10 (Gutsy, kernel 2.6.22 ?).

Well, this installation had severe problems aswell, but(!), they all went away when I decided to try downgrade the kernel to an older Feisty kernel (2.6.15?)! This system ran happily for several months, 'til I accidently barred my only user from running sudo, heh. Now I'm trying out Annvix 3.0, which is sporting a -- dramatic pause -- a 2.6.22 kernel.

So, you see where I'm going with this.. I'm inclined to believe something has gone awry in the newer kernels -- with the cmd64x or the USB module, or something in between.

Does anyone around here, who might be a bit more involved in the developing of things, know anything about this? In searching for answers, I found this mailing thread, but it kind of runs out in the sand, with no answers. Maybe there's even a patch or two out there I could try...