(Cc'ed to port-prep)

In <m14NIVA-000Fz5C@shanghai.realtime.bc.ca>
querbach@realtime.bc.ca wrote:

> > My IBM 7248 (prep) also has similar problem that
> > the firmware does not configure PCI interrrupt properly..
> 
> You might have a look at pci_attach_hook() in prep/pci/pci_machdep.c.  This
> function forces the interrupt line setting to a hard-coded value derived
> from a lookup keyed on the PCI device number, rather than using the
> interrupt line register in the device, as is the usual method.

Year, usually the firmware on the machines should setup PCI configuration
registers properly, but unfortunately some of them don't.
On PowerSeries 830, all PCI interrupts are wired to irq 15 at ICU.
On IBM 7248 (43P), it seems no PCI interrupt is wired to it.

> I'm not sure why this was done.  I didn't find similar code in the other
> ports.

There are some PCI FIXUP code on i386, arch/i386/pci/pci_*_fixup.[ch] etc.
We probably should write such code for each PCI-ISA bridge on prep.

Linuxppc patch for these machines
http://www.amberdata.demon.co.uk/carolina/
might be a good start for this.
---
Izumi Tsutsui
tsutsui@ceres.dti.ne.jp