Some time ago I was going to triumphantly report that my 500a works
perfectly with NetBSD, but after three weeks of trying to get it to run,
I've moved on to more pressing items.  My particular problem with a rev.
A04 board was not related to Jeff's problems, but rather once the
machine is booted (I've tried all iterations of INSTALL, GENERIC, and
custom kernels from the 12/4 and 12/31 snaps, and -current CVS sources),
I can't get it to stay up long without an imminent massive corruption of
the VGA display and an instant hard-reset.

It's fine for a while, until I use 'vi' and begin scrolling the display
(or do anything which touches virtual memory, it would seem).  Other
OS's don't choke.

I have to assume at this point that the damn pyxis bug is quite a doozy,
and is still haunting the code.

dhop

Jeff Roberson wrote:
> 
> Sorry to bother the list with so many mails, but I just noticed something
> interesting.  In cia_dma.c CIA_PYXIS_BUG_BASE was set to (1*128*1024).
> It's supposed to be at 128mb though.. So I changed it to (128*1024*1024),
> and the kernel stoped freezing.  I still can't send data over my nic
> though.
> 
> Thanks,
> Jeff
> 
> On Mon, 10 Jan 2000, Jeff Roberson wrote:
> 
> > Date: Mon, 10 Jan 2000 23:03:09 -0800 (PST)
> > From: Jeff Roberson <nomad@nop.aliensystems.com>
> > To: Ross Harvey <ross@ghs.com>
> > Cc: wilko@yedi.iaf.nl, mjacob@feral.com, port-alpha@netbsd.org
> > Subject: Re: More pyxis/500a info..
> >
> > It took me a bit to get back on track with this alpha.  Anyhow, I tried
> > the memory barrier to no avail.  I also tried commenting out the
> > workaround for the bad tlb flushing in cia_dma.c.  That didn't help
> > either.  If it's enabled the machine freezes, and if it isn't I can't use
> > my ethernet card.  I can ifconfig and everything, but I can't actually
> > write to the device.  I'm planning on going through freebsd kernel source
> > soon, but I'm stressed for time right now.  If anyone has any other
> > suggestions I'm open to them.  I purchased tru64 for use in the mean time,
> > but I'd still like to go back to netbsd.
> >
> > On a side topic, would it be possible to use digital's compiler to compile
> > the netbsd kernel?  I imagine their optimizations are far superior to
> > gcc's.  Could it run in OSF1 compat?  Or could you perhaps compile it on a
> > tru64 machine?
> >
> > Thanks!
> > Jeff
> >
> > On Tue, 4 Jan 2000, Jeff Roberson wrote:
> >
> > > Date: Tue, 4 Jan 2000 11:51:24 -0800 (PST)
> > > From: Jeff Roberson <nomad@nop.aliensystems.com>
> > > To: Ross Harvey <ross@ghs.com>
> > > Cc: wilko@yedi.iaf.nl, mjacob@feral.com, port-alpha@netbsd.org
> > > Subject: Re: More pyxis/500a info..
> > >
> > > I seem to remember adding a mb last night, which had no effect.  I will
> > > double check that tonight, along with your other suggestion.
> > >
> > > Thanks,
> > > Jeff
> > >
> > > On Tue, 4 Jan 2000, Ross Harvey wrote:
> > >
> > > > Date: Tue, 4 Jan 2000 11:43:40 -0800 (PST)
> > > > From: Ross Harvey <ross@ghs.com>
> > > > To: nomad@nop.aliensystems.com, wilko@yedi.iaf.nl
> > > > Cc: mjacob@feral.com, port-alpha@netbsd.org
> > > > Subject: Re: More pyxis/500a info..
> > > >
> > > > > [ Yet Another Miata Problem ]
> > > >
> > > > Here are a couple of things you can try: some untrustworthy folklore
> > > > suggests that adding an alpha_mb() call before the bogus line MIGHT
> > > > have an effect.
> > > >
> > > > Another thing to try would be s/if ((/if (0 && (/ at line 203, to see if
> > > > your Miata works better without the bug workaround.
> > > >
> > > >   ross
> > > >
> > >
> > >
> >
> >