[TUHS] Historical sources for 68010 + 68451 systems?

Arno Griffioen arno.griffioen at ieee.org
Wed Sep 16 15:21:49 AEST 2020


On Tue, Sep 15, 2020 at 04:28:36PM -0700, Chris Hanson wrote:
> I have an MVME121 that I’d like to run some stuff on. I’m planning what I’ll 
> need to do to port MINIX 1.5 but since this has a 68451 segmented MMU, I’d 
> like to actually make use of it.

Quite doable to adapt MINIX to this. Did a similar thing ages ago while 
still in school on a 68010 machine (not really VME, but similar card-cage
setup) that had a custom MMU built from some SRAM and some control logic on 
it. Used MINIX 1.1 at the time though..

Only had 1MB physical RAM and the MMU could provide a virtual address space 
of 4MB.

Initially ported MINIX to the machine 'as is' from the ST version and then
modified the MINIX kernel to allow it to use the 4MB virtual space and
such.

Because of the bare-bones MMU design it had some interesting quirks that made 
it a good learning experience at the time. (and some very odd serial chips
for the terminals that had their own 'interesting' behaviour but that's
a different story..)

Originally the machine did have a real SVR2 UNIX running on it, but sadly it 
was binary-only, so apart from some dis-assembly of bootstrap bits to learn
how to get MINIX to start on this machine it was not very useful.

> I figure at minimum I could have several segments set up to enforce 
> protections and a stable per-process address space, but it’d be good to 
> have an example.

Yup. You may run into some of the shortcuts in the MINIX kernel when you start
doing MMU work though, especially if you want to separate the kernel
'processes' as well. 

For performance the microkernel architecture of MINIX was violated in a few 
spots, mostly around FS and MM, where one kernel process would/could modify
another's memory without going through the message passing mechanism.

Introduce an MMU and that kinda breaks and needs some cleaning up ;)

As MINIX later on did get a 386 port and cleanups/fixes that may now be a
non-issue on 1.5.

Another option would be Linux/m68k and perhaps starting on the 'nommu'
version that runs on basic 68000's too and seeing how much functionality could 
be used from the 68451 to enhance it. 

Some of the Coldfire CPU's have similar limited MMU's that are supported for 
some functions of basic memory protection in the 'nommu' tree.

As far as real UNIX sources go.. SUN2's were 68010's too although AFAIK
with a custom SUN designed MMU logic? Perhaps some old sources available in 
that corner though.

							Bye, Arno.


More information about the TUHS mailing list