Subject: Re: r5k O2 support
To: None <port-sgimips@netbsd.org>
From: None <aliver@xexil.com>
List: port-sgimips
Date: 07/23/2003 01:06:33
On Wed, 23 Jul 2003, Christopher SEKIYA wrote:
> Very old.  I think 5200s work now, and there is no serial speed
> restriction.

Excellent.

> I'm about halfway done with the ethernet driver.

	Out of curiousity, how did you get the documentation? Was it a
chip that's also used elsewhere so you could sort of "clone" the docs, or
did you get some information directly from SGI? I'm a C coder myself who
would love to get involved with NetBSD, but I've never written a device
driver. I've only been coding for about 2 years now, but I'm certainly at
the "intermediate" level where I've written some 15,000 line applications
etc.. I have respect for anyone who can hack a device driver, and does so
for free in NetBSD.

> Every PCI card that I've thrown at the box has worked.

	That's great! I heard someone else say adding a USB card to the
machine made it usable with USB audio, keyboards, etc.. That's a major
bonus with the O2; it has a PCI slot. I used to have one, and it was
fairly tricked out except for the fact that it as an R5k 180. Does NetBSD
support the R10k and R12k O2's yet?
	Also, off the top of your head are there any "combo" cards which
could be used to leverage the one PCI slot that it has? I know Adaptec has
a PCI card which has IEEE1394 and USB 2.0 on it.  That might be helpful
for an O2 owner. What about PCI video cards? Do you think it'd ever be
reasonable to put an ATI 9500 in an O2 and expect it to work? I wonder if
anyone is ever going to make a USB 2.0 framebuffer. I'm sure the
performance would be terrible, but think of all the tricky things you
could do with something like that. Especially in an environment like a
server farm with 1U rack machines (ie.. plug in a little USB LCD monitor
etc..). Of course there is nothing wrong with serial terminals, either ;-)


> L2 cache works.

I'm sure that without the L2 cache the box would probably crawl. So this
is another great development.

> The parallel port works. Onboard SCSI works in synchronous mode.  The
> little box is very stable, actually.

So is asychronus access still left to-do, or was it done already?

> Unfortunately, most of the code isn't in-tree yet.  I've made a patch
> against -current available at
> http://www.rezrov.net/~wileyc/ip32-20030723 for those so inclined.

Maybe I should trawl Ebay for another O2... I dunno it's getting crowded
in here. I'm typing this on an Indy running Irix 6.5. I have a Challenge S
I plan to use with NetBSD real-soon-now. Other than the mezzazine
SCSI/Ethernet it's the perfect machine to try NetBSD on right now since
it's a serial-only, headless box anyway. The day I can load NetBSD on my
Indy, use the keyboard, mouse, XL24 graphics, sound, and VINO port is the
day I will dance a jig of joy. I don't want to even think about an O2 with
all the hardware supported; that'd just be too damn cool.

aliver

PS: too bad there are no CPU accelerators for older SGI machines.