> Date: Fri, 10 May 1996 10:12:00 -0700 (PDT)
> From: Curt Sampson <curt@portal.ca>

> Gordon,
> 
> This is basically an FYI that I'm cc'ing to the list.
> 
> It seems that the problem on my 3/60s with -current kernels is
> definitely related to get cg4 autoconfig in some way. I tried out
> a -current kernel from a couple of days ago on my monochrome 3/60
> and it booted, although when it finally finished /etc/rc it just
> gave repeated messages on the console about getty repeating too
> fast on the keyboard and both serial ports. (There's the usual
> getty.core in the root directory, so I presume this is still the
> bug we know and love.)
> 
> So I removed all of the cgX stuff from a kernel config file and
> compiled a new kernel and, voila, it works on the colour 3/60!
> 
> I'm sure you've got plenty to work on with the VM bug at this point,
> and I can swap the one colour 3/60 I do have that uses a monitor
> for a monochrome one, so I think the best move is to just ignore
> this problem until I get around to actually doing some further work
> on it to track it down.

Thanks for the update.  If anyone can look into the cg4 problem
I'd appreciate it.  I have a 3/260 with a cg2 that works fine,
so this problem may be specific to the cg4.  Not sure...

Yes, I'm still hunting for the elusive VM (scsi?) bug.

I suspect there is some DMA hardware "feature" related to
interrupts that I don't understand.  I really wish I had a
copy of the "Sun3 SCSI Programmers Manual."  I've asked lots
of places, but never found one.

> I notice that the kgdb stuff for the zs chips is allegedly not
> tested yet. Have you used it at all? I tried to compile a system
> with `options KGDB', but it seems to be looking for
> sys/arch/sun3/sun3/kgdb_stub.c, which I assume needs all of those
> nifty functions that are in the sparc version of that file (but
> sun3 versions of them of course).

That has been a "back burner" project.  I got held up on it when
I realized the "debugging host" side gdb code needed work too.
It looks like the m68k remote debugging support has "code rot".

The sun3 kernel support for kgdb is almost done.  Note that I used
a different (I think better) approach than the hp300 for switching
to a debugger stack; locore.s does it in the trap 15 handler.  That
means we do not need any of the "kgdb_glue" stuff the hp300 uses.
Pehaps I can finish up the kernel side stuff, at least to the point
where it builds.  Someone will need to update host-side gdb stuff.

Gordon