Xref: utzoo comp.sys.amiga:67483 comp.os.minix:12601 Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uunet!aplcen!uakari.primate.wisc.edu!zaphod.mps.ohio-state.edu!mips!wyse!vsi1!zorch!xanthian From: xanthian@zorch.SF-Bay.ORG (Kent Paul Dolan) Newsgroups: comp.sys.amiga,comp.os.minix Subject: Re: MINIX on the Amiga... Message-ID: <1990Oct1.000617.3676@zorch.SF-Bay.ORG> Date: 1 Oct 90 00:06:17 GMT References: <1990Sep24.114627.14857@watserv1.waterloo.edu> <1990Sep24.103505@wjg.enet.dec.com> <7726@star.cs.vu.nl> Organization: SF-Bay Public-Access Unix Lines: 53 sreiz@cs.vu.nl (Reiz Steven) writes: >guineau@wjg.enet.dec.com (W. John Guineau) writes: > >>I've been talking with Andy Tanenbaum about Minix. >>He says he doubts that it will run on a 68030 based machine >>(ie A3000, A2500/30). Minix does not set up the MMU. [...] >>Has anyone gotten Minix running on one of these machines? > >Not having tried this myself I can only guess at the exact nature of the >problems encountered when running Amiga Minix 1.5 on 680x0s where x>1: > >- The highest 8 bits of some addresses are used for a 'dirty trick' by the > minix kernel. This means that you must be able to instruct your 68020/ > 68030 board to use 24 bit addressing, like the 68000 and 68010. Whew! We already got bitten by this one with AmigaBASIC. I hope someone guts this with a huge set of "#ifdef SANE_DESIGN"s pretty fast and leaves the full 32-bit memory addressing accessible. This is pretty gross for code written for the 1990's. >- Minix can't access memory outside the 16 MB range. This is caused by > the way addresses are encoded in the memory manager. I don't know if > any 68020/68030 boards really have memory outside the 16 MB range, but this > doesn't seem a very serious problem to me anyhow. It does to me! I'm running a vanilla Amiga 2000 with 9.5 meg of RAM, and I could easily put ten times as much to use on a regular basis. I can promise you, it is going to be a very short time, we may be there already, when the trade press is going to be talking about ways to get around the "Poorly Planned 32 Bit Barrier" for real memory addresses. A 24 bit inherent limitation is a joke in bad taste in today's world, where that much memory costs less than a second, used car, easily within the reach of the dedicated hobbiest, not to mention commercial sites. Noted in another article is that MINIX is a learning tool. One of the first lessons to learn looks like: "Don't assume there is _any_ resource for which the user can't find use for more; leave room for growth beyond your wildest imaginings". That wisdom should be a fundamental part of the design of _any_ OS today, and designs as described above should get failing grades in college "Principles of Operating Systems" classes, much less quasi-commercial products. There is little if any excuse for turning address longwords into records to save a byte here or there in a memory management scheme's overhead when memory is dirt cheap and widely available. Kent, the man from xanth. -- Fanning the flames a bit, my hands were cold.