In message <20030406145042.GA902@antioche.eu.org>, Manuel Bouyer writes:
>> FWIW, on this device, PCIBIOS_ADDR_FIXUP causes core dumps, because it gets
>> a 0 for the address for at least one device.  I'm not wholly optimistic.  ;)

>Well, no, I don't think it ca work this way. We have to try on a working BIOS
>for i386. For the kernel to do IRQ routing by itself there would be some
>amout of code to write I guess.

Hmm.  My understanding is, on this device, the IRQ *is* already routed - we
just don't seem to pick up on it correctly.  It's like my mom's laptop, where
you need ACPI to *find* the interrupts.  They're there, they're configured
correctly, but the BIOS doesn't report them correctly except in ACPI.  On this
system, there's no ACPI, but there's a manual that says "the cardbus bridge
uses IRQ 10".

-s