On an AMD dual athlon you need acpi=off pci=usepirqmask together. Surely acpi=off should stop trying to use acpi routing data ?? [Box works reliably in 2.4] It then later dies with a spew of oops data from a network operation of some kind.
Ok its crashing reliably on the ifconfig ifup of the network card (netgear). dev->change_flags (hence I guess ifup) dev->open alloc_skb cache_init_obs then an IRQ coming in and can't recover the top of the trace So it looks like a network driver with incorrect irq protection
Network oops turns out to be a side effect of the broke acard scsi module it seems (that or its just chance that acard scsi puts the corruption somewhere that matters). Other stuff still the same
how's things looking with the current kernel on this box ? do we need to fiddle stuff based on dmi strings ? Is it worth adding the acpi folks (acpi-bugzilla.net) to the cc: of this bug ? or is the DSDT too broken for them to work around?
Will investigate further
Works properly with FC3test2 so I guess the bug is fixed now