In message <20060511204030.B6AB12356D@thoreau.thistledown.com.au>, Simon Burge 
writes:
>Peter Seebach wrote:
>
>> I've been a bit distracted, but I just noticed this ongoing discussion.
>> 
>> FWIW, I have at least one ibm4xx system where the interrupt mapping in the
>> existing pci_machdep code is wrong.
>
>All of the interrupt mapping goop is machine-specific.  You're own
>board/port should be supplying this, not generic code.  With a very
>quick look, this seems to be what we're doing.  Which system are you
>talking about?

It's the TAMS 3011.

It may have been fixed, I have an old source tree.  In my source tree,
/sys/arch/powerpc/ibm4xx/pci/pci_machdep.c had the pci_intr_map code, because
it was the same for everyone.

Looks like it's been fixed.  I really need to get around to cleaning up
and submitting these patches!

-s