Subject: Re: CTIX architecture
To: David Maxwell <email@example.com>
From: John E. Clark <firstname.lastname@example.org>
Date: 12/08/1994 09:09:22
At 9:59 AM 12/8/94 -0500, David Maxwell wrote:
>On December 7, David Maxwell (Me,Myself, and I) wrote:
>> > M68000 CPU, ST506 and SCSI (and SMD) drives, multi-port serial cards, and a
>> > VME bus.
>Dave McGuire replied:
>> Seems to me that if it's a 68000 (not 68020) it would be a bit of a
>> problem...but if it's an '020 or better, it's probably just a matter
>> of device drivers.
>Oops! Let me restate that as m68k series CPUs. These all already run Unix,
>but an old,old,old version not longer being maintained or updated. All the
>ones I know of are '020 or better. I'd be surprised if they'd gotten Memory
>management running on any older chip ;-)
If the CPU is in fact a 68000/10 then the only type of 'swaping' is full
task swapping, ie no paging. I don't know if BSD was ever envisioned for
a purely swaping environment. However, SYSV-mumble-foo, and V6 un*x were.
If the processor has a 68020 + PMMU, or equivalent, then there may be
hope. I think even if it has the 'Sun/Stanford' MMU there could be hope...
The problem with 'paging' is that the 68000 couldn't restart instructions,
so if the task wasn't fully in memory then a bus error on a non-existent/paged
memory area would cause the task to fault out.
A trick was embedded into the compiler to test the bottom of the stack in the
subroutine which if not present, would cause the OS to allocate more stack.
'magic stack growth' could happen, but this was as special case.