> Yeah, it probably doesn't... when I put in values for the AV machines, I
> was just following the lead of the other machines, where IOBase started
> with VIA1.

Yep. Looks like we might end up doing away with mapping in a big chunk
anyway.

My problem was that the video controller (and the interrupt stuff we've added)
resides squarely starting at 0x50036000 while NetBSD believes that
IOBase really starts at 0x50f00000. This is sort of true on other machines
but for most machines the 'n' in 0x50n00000 is a "don't care" value and
consequently the same stuff is just repeated over and over again. Not so
on the AV's.

I thought I was stuck because I can't access addresses such as
(IOBase - something) and I thought I couldn't map in the address.
It turns out that the device space addresses are (we think) physical
addresses and I can use them freely when doing bus_space_map stuff.

That was a major stumbling block for me. Now that I can map in the
bus space I can resume work on the AV color video driver.

> I agree that onboard video support isn't as good as it could be

Tell me about it. :) I'm working on it but it's taking me time because there
is just scant, spread out all over the universe, documentation. So I have
to do research, watch assembly, and of course, do the other things in my
life :)

> (color would be nice, for example, as well as support for changing bit
> depth/resolution within NetBSD, instead of having to use the Monitors
> control panel, detection of how much VRAM is present, a bunch of other
> stuff)

You'll be happy to know that I'm workin' on all of it.