I think there's some plan9.ini foo which says "no matter how many PCI devices this machine claims to have, it has only N", which if you're lucky might clip out the second #y.
The only thing I could find in the driver was "pcmcia0=disabled" (I forget the exact value) that disables all PCMCIA configuration attempts. There's nothing to override the number of card slots.
I'm not sure changing the PCI probe limits is going to help here. I think for now I'm going to start adding prints to the 589 driver to see exactly where the push of the received data up to userland is falling over. I still don't know if this is related to the IRQ5 confusion or what. I also need to examine the FreeBSD driver for the cardbus bridge to see if they're doing anything special.
--lyndon
