Subject: Re: Which Sun3 architectures?
To: Holger Veit <Holger.Veit@gmd.de>
From: Adam Glass <glass@sun-lamp.cs.berkeley.edu>
List: port-sun3
Date: 02/03/1994 14:23:26
> [...]
> > the port has been developed on the sun3 desktop models i.e the 3/50
> > and 3/60.  It should also work on any of the machines that use the same
> > motherboard as the 3/60 (Theo, fill in the details?).
> > 
> > To get the port working on any of the other machines you need to solve
> > the following problems:
> > 	until we have a scsi driver, you need nfs diskless mount your
> > 		root.  This means that we need a driver for the intel
> > 		ethernet found in the non-desktop machines.
> Probably the ie-driver could be borrowed +-some address changes from the
> i386 arch. Regarding scsi, I currently have the SunOS on the SCSI disk,
> perhaps the xylogics controller is simpler to access, must have a look
> at the doc I found for that (no idea whether it is complete). What I don't
> really like is to prepare a diskless boot, because I would have to
> port tftp and rarpd to OS/2 TCPIP.
> 

I agree that this is a bad scene.  However, doing the scsi driver
would not be that difficult.  NetBSD already has a scsi infrastructure
(disks, tape,cd) etc, and we have a ncr5380 driver for it.  It would
still require some work to get this to work within the sun3 port, but
it is a great start.

There should be some fresh/recently debugged ie-type code coming down
the pipe soon for the i386 port.  We should be able to leverage this...

Where there are holes in commonly available documentation, we will
need to pool resources/sources.    There is definitely code in mach
and Sprite that while we perhaps can't  lift intact, from which we can
learn much.

> > 	no vme support.  It would be easy to do, and the autoconfig
> > 		scheme can be made to understand it w/o much work.
> > 	any wierd post-3/60 architectural enhancements like virtually
> > 		addressed caches.  I think i've got code in there to
> > 		deal with this, but it has never been tested.
> I am not sure whether the 260 already has this, at least it runs a
> generic sun3 kernel. The 3/75 I could have got alternatively wanted a
> "sun3x" one, and I've been told this is "totally different", likely due
> to different MMU and other incompatible stuff.
> Anyway, it looks like a challenge...
> 
> Holger
> 

the 260 does have a VAC, so think of it as an opportunity to debug my
code :)

the 3/75 is not a sun3x architecture.  only the 3/80 and the other 030
based machine.

to clarify something:

You only need a nfs diskless boot setup to get past mountroot() right
now.  You can still work on the device-driver problems, as
autoconfiguration is earlier.  However, this does not help Holger and
his OS/2 machine....

later,
Adam Glass

------------------------------------------------------------------------------