kiyohara@kk.iij4u.or.jp wrote:

> I think that the thing that two or more interrupt can be registered is
> necessary for cobalt.  There is certainly a problem that Firmware doesn't
> set interrupts.  I want to examine where of ICU the interrupt line of
> cobalt PCI is connected.

I've changed the panic to a printf so attaching the device
will fail properly rather than panicking.
(I'll also send a pullup request later)

Once we investigate which ICU (i8259) irq pins are used
for each PCI interrupt PINs of the PCI slot, we can handle
it in sys/arch/cobalt/pci/pci_machdep.c:pci_conf_interrupt().
---
Izumi Tsutsui