Subject: More Patmos info
To: None <macbsd-general@NetBSD.ORG>
From: Aaron Rosenzweig <email@example.com>
Date: 07/10/1995 23:18:44
I understand that this list is for MacBSD but I think this is of
general interest. I don't plan on making too many more posts like these.
This comes straight from the author of Patmos. It tells where the
latest version can be found and answers some questions.
>Date: Mon, 10 Jul 95 06:10:21 PDT
>From: Jonathan.Kimmitt@camcon.co.uk (Jonathan Kimmitt)
>Subject: Re: MacMiNT II
>> True. Even if it has been done, who cares! We could learn from it, and
>> people would have another unix-like OS to run on their Mac. I like the
>> idea of it being a REAL OS like BSD, but I also like the idea of sharing
>> with MacOS. Maybe instead of taking over the whole machine like MacBSD,
>> MacMiNT would only take over part of it, so we'd have MiNT and MacOS
>> running together. Kind of like the way the 6600/60 DOS card works (hit a
>> key combination and you're in dos or windoze, hit it again and you're back
>> in MacOS).
>>Well, on this side of things, there is a program called "PATMOS",
>>for "Protected Address Translation Mode Operation System". It only
>>runs on 68040's and takes over the PMMU completely. It will run
>>concurrently with MacOS, though without inits like ramdoubler, or
>>virtual memory stuff. Patmos is binary compatible with 'MacBSD'
>>according to the docs. So this would be an option for running
>>NetBSD/mac68k apps on MacOS. The dependence on an 68040 hoses me
>>though. I guess what needs to really be hashed out, is what exactly
>>people want to see in UN*X type functionality. With a list of features,
>>it would be easier figure out what fits.
>Thankyou for mentioning Patmos on this list. There certainly are some
>to basing new stuff on the Patmos kernel:
>(A) You don't need to modify any source code from the BSD source tree to
>(B) You get a true version of fork()
>(C) You get MacTCP based networking, via the BSD AF_INET domain sockets
>(D) You can easily expand the support for special files to cope with multiple
>(E) You can use gdb without any special hacks to debug your programs
>(F) Bus errors (in most cases) don't crash the mac because the MMU physically
>(G) The kernel is only 80K and compiles with THIN C available on the nets
>(H) You don't have to worry about GNU source distribution because the
>the official version from prep.ai.mit.edu
>(I) You can use the BSD4.4 C-library with the database functions etc.
>(J) You don't have to specify in advance how big the stack will be for a
>program. It automatically grows to the right size. You can use the 'limit'
>command of csh to prevent stack runaway
>(K) You can get signals such as SIGALRM, SIGCHLD, SIGINT in your programs
>(L) You can use AF_UNIX domain sockets for system logging
>(M) You can easily add support in the kernel for sunos statically linked
>binaries and MacBSD executables
>(N) The kernel avoids the use of toolbox glue code so the source is freely
>distributable, it just needs header files to allow recompilation.
>On the downside there are a number of differences between the MMUs of
>motorola chips. The 68040 is the simplest and that's what I have so that's
>I wrote Patmos for. However you can write a system extension that programs the
>68030 MMU to act like a 68040. All that needs to be done is to deal with the
>different instructions for flushing the caches (logical v. physical) and other
>niceties. Obviously this task is not for the faint-hearted.
>The other downside is that MacBSD et al. often assume the presence of a true
>68040 with co-processor. The majority of new non-ppc macs have the 68LC040.
>Therefore you need to re-compile to use software floating point before you can
>achieve binary compatibility between Patmos, SunOS, and MacBSD
>The latest released version of Patmos is on
>suniams1.statistik.tu-muenchen.de in the incoming/Patmos directory
>I hope to have a much better version on the soon to be released Apprentice CD
>Such a shame I don't have more time to devote to this fascinating project.
The bearded Tubist, son of Ginger and Harry, Aaron Rosenzweig.