> the "multiple botch" happens only when it tries isa tbdf and can probably > ignored? > > i wonder why the first mpintrenablex() with the pci tbdf would fail. the > only case where it would fail that way without leaving some other kernel > print behind is when there is no matching pin (aintr->irq) entry in the > bus->aintr list no? > > that list is build from mp table. maybe should figure out the pin of these > pci devices?
i'd dump the irq list next. there's code in 9atom to present /dev/mpirq. - erik