>> 	/pci106b,1@B
>> 	/pci1011,14@D
>> 	/pci106b,7@E
>> 	/chips65550@11
>> 	/ti1130@13
>> 	/ti1130@13,1
>
>It's one of those, and it isn't the chips65550, 'cause that's your video.
>
>Like I said, to my knowledge, it works with MkLinux, whose tulip driver
>was ripped straight out of NetBSD apart from some tiny one-line patches
>and I think massive changes to the probe/attach routines, so if it doesn't
>work under NetBSD, it's probably some trivial issue with probing/attaching
>and/or Open Firmware not enabling the device's I/O and config space or
>similar.

Some of these machines have the ethernet chip somewhat bridged behind
an "OHareII" chip that acts like a cascaded interrupt controller, and
thus must be handled a bit like the Heathrow/Gatwick combo of the
wallstreet. The ethernet interrupts is routed to that second ohare,
which itself is cascaded on an interrupt line of the main controller.
You'll also have to do some fixup of interrupt numbers of the OF tree.

I'd suggest you have a look at linux/ppc arch/ppc/kernel/pmac_pic.c
where all the code to handle that is lying ;) Note that the cache coherency
of these isn't very good and you may run into trouble if you don't use
the tulip chip "feature" of modifying the ring descriptor alignement to
align each descriptor on a cache line boundary.

Regards,
Ben.