Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1+some 2/3/84; site dual.UUCP
Path: utzoo!linus!decvax!decwrl!dual!fair
From: fair@dual.UUCP (Erik E. Fair)
Newsgroups: net.micro,net.micro.68k
Subject: Re: 68020 MMU Questions
Message-ID: <871@dual.UUCP>
Date: Wed, 21-Nov-84 00:07:30 EST
Article-I.D.: dual.871
Posted: Wed Nov 21 00:07:30 1984
Date-Received: Thu, 22-Nov-84 07:04:08 EST
References: <4667@fortune.UUCP>
Distribution: net
Organization: Dual Systems, Berkeley, CA
Lines: 35

DUAL Systems has never used a proprietary MMU scheme for any of the
68000 family of CPUs in any of our boards (unless you consider the
68451 from Motorola to be proprietary).

The very first DUAL UNIX system had no MMU because we waited for
Motorola to finish the one they were working on. In retrospect, that
was a mistake.  In the meantime, however, we just split memory into two
sections, one starting at location 0, and one starting at location
0x800000. The UNIX Kernel ran in the 0 section (or `kernel memory' as
we called it), and user programs ran out in the memory at 0x800000.
When a program in user mode attempted a reference to an address less
than 0x800000 (i.e. without address line #23 on), the system took a
protection trap, and we would blow the user away with SIGSEGV.

Needless to say, without an MMU, context switching was a major chore
(lotsa copying of user text, data, and Udots) and was *very* slow.
However, UNIX worked.

When the 68451 arrived, it was designed into the CPU board, and our
kernel was modified to take advantage of it. I don't think there are
any of the non-MMU DUAL UNIX systems left out there. (I certainly hope
not!).

	[begin disclaimer]

	This is soley my opinion, and does not necessarily reflect the
	opinion or policy of DUAL Systems Corporation.

	[end disclaimer]

	Erik E. Fair	ucbvax!fair	fair@ucb-arpa.ARPA

	dual!fair@BERKELEY.ARPA
	{ihnp4,ucbvax,hplabs,decwrl,cbosgd,sun,nsc,apple,pyramid}!dual!fair
	Dual Systems Corporation, Berkeley, California