I have committed my driver for SGI NG1 ("newport") graphics controller.
While it hasn't been too widely tested (only tested on my Indy with rev
1 8bit newport), it *should* work on most (hopefully all) newport
boards, both in Indys and Indigo2s. (I'm very interested in hearing
whether it works on 24bit newports...)

The driver currently trusts the ARCS to setup the newport to some extent
(VC2 timings and GIO configuration), which means that it *may* fail on
dualhead boards for the second head (I don't have access to dualhead
Indy so I haven't been able to confirm whether or not ARCS does setup
for the 2nd head too).

I also committed the new GIO probe code (it seems quite robust except
for the harmless "product id 0x04 revision 0x00 not configured"s on
Indigo2s, which are a byproduct of really stupidly broken hardware) and
code for attaching pckbc & newport as the console device (this means you
don't need to use serial console if you have monitor and keyboard
around).

PS. Now we need X for this thing... XF86 already has a (horrible, yuck)
newport driver (in our tree, too), which could be coaxed to work pretty
easily by making the kernel driver able to mmap() the device registers
to userland and making the X driver use that instead of the linuxy hacks
they do now...

PPS. Would someone *please* clean up consinit(), it's horribly messy
already and getting worse all the time as we add more devices. ;)

-- 
Ilpo Ruotsalainen - <lonewolf@iki.fi> - http://www.iki.fi/lonewolf/