On Sun, Sep 02, 2001 at 01:37:16AM -0000, eeh@netbsd.org wrote:
> There are several problems I have with this patch:
> 
> 1) What exactly is a `SUNW,Ultra-1-Engine'?  And what problem does it
> have with that machine that this patch is supposed to fix?
> 
An early ATX style Ultra 1 class machine. The patch is necessary because the
pci bus does not have a 'name' property in the prom, but DOES have
"compatible".  Looking at the linux code, they match either on 'name' or
on 'compatible', too.

> 2) The patch `seems to mostly work'.  What does not work?
> 
The onboard hme never sees interrupts.  I'm still puzzling through that
one.  At any rate, this patch actually allows me to get through autoconf
(instead of seeing 'pci not configured') and all devices are probed
correctly.

The register offset part of the patch is right (at least when compared
to the linux version of the same driver); as are the bus interrupt
registers being printed.  The ue registers aren't interesting when
a psycho_bus_{a|b} interrupt is triggered, but the pcictl_{afar|afsr}
registers are (I ran into this because of a bug in wdcprobe() in
OpenBSD generating pci-master-abort's).

> 3) The patch has not actually been tested on NetBSD.
> 
> Since you have access to that hardware and I do not, could you 
> at least test it to the point you are confident it fixes the problem
> you have on that haredware?
> 
I am still working on the interrupt problem, but the fixes I've sent
so far are the "obvious/easy" fixes.  It's much easier on me to see
this integrated into NetBSD and then send a further patch (or two)
when I've figured out the interrupt problem.

--Jason L. Wright