Path: utzoo!attcan!uunet!lll-winken!lll-tis!helios.ee.lbl.gov!nosc!ucsd!brian
From: brian@ucsd.EDU (Brian Kantor)
Newsgroups: comp.os.xinu
Subject: New Port - advice requested
Message-ID: <970@ucsd.EDU>
Date: 9 Jun 88 19:04:52 GMT
Reply-To: brian@ucsd.edu (Brian Kantor)
Organization: The Avant-Garde of the Now, Ltd.
Lines: 40

I'm in the process of porting Xinu V7 to an 80186-based network
controller - the SANDPAC PS-186.  This is a fast little board designed
to be a dedicated network controller for the Amateur Packet Radio
Network.  It has four serial channels capable of running async or HDLC
at up to 1 Mbit/S each simultaneously, 256K of static battery-backed
RAM, and some other nifty things related to its intended purpose as an
automated unattended radio-system controller (probably living in some
weatherproof box on a snow-covered mountaintop).

The architecture doesn't look much like a PC or AT - the interrupt
controller and DMA are QUITE different - and there's no console nor
keyboard.  The timers, time-of-day clock, and parallel ports are
different too.  We couldn't have kept the hardware the same as a PC or
AT and still gotten the kind of performance out of it that we need.
Oh, it does have an NCR SCSI controller, so it IS possible to attach a
hard drive controller to it (a DTC or Xebec are available here), but the
final system will probably be loaded from EPROM - or uploaded over the
radio network link by a simple loader in EPROM, since ordinary disks
won't take the environment.

I've started by using as much of the V7 Xinu as seems to port directly,
and grabbing large pieces of the V6 8086 Intel-board port.

The intent is to have a box that is a very efficient IP switch that
knows how to do ARP, speak AX.25 over HDLC (that's the legally-required
radio networking protocol) and to provide some services like Telnet.
As you might guess, I've never ported something quite this large before,
and I'd like to reduce the amount of error in the inevitable
trial-and-error part.  It would help if it weren't quite such a trial,
too.

I'd appreciate any suggestions on methods of attack.  Clearly the first
thing to do is get some of the very low-level kernel stuff working -
I've already got kputc and kprintf working.  I suspect that interrupt
handling is next.  I'm told we may have an 80186 in-circuit emulator
available in a few months, but that's not certain yet.

I'd also like to know what the restrictions would be on giving away
copies of the software when it's all working.

	Brian Kantor	WB6CYT	UC San Diego   brian@ucsd.edu