*BSD News Article 31294


Return to BSD News archive

Xref: sserve comp.unix.bsd:14106 comp.os.mach:3924
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!agate!howland.reston.ans.net!swrinde!pipex!sunic!news.funet.fi!nntp.hut.fi!nntp!jvh
From: jvh@cs.hut.fi (Johannes Helander)
Newsgroups: comp.unix.bsd,comp.os.mach
Subject: BSD vm (Re: Anyone using MachTen on a Mac?)
Date: 05 Jun 1994 23:21:51 GMT
Organization: Helsinki University of Technology, CS Lab.
Lines: 30
Distribution: inet
Message-ID: <JVH.94Jun6022152@hutcs.cs.hut.fi>
References: <1994Jun3.210215.20699@midway.uchicago.edu> <2sp809$hnq@xochi.tezcat.com>
	<1994Jun4.091322.13685@midway.uchicago.edu>
	<2sruh6$hnh@news.u.washington.edu> <2ss0lm$54s@xochi.tezcat.com>
	<VIXIE.94Jun5145032@office.home.vix.com>
NNTP-Posting-Host: hutcs.cs.hut.fi
In-reply-to: vixie@vix.com's message of 5 Jun 94 14:50:32

In article <VIXIE.94Jun5145032@office.home.vix.com> vixie@vix.com (Paul A Vixie) writes:

>   "Based heavily" might be too strong a wording.  BNR2 included the Mach 2.5
>   "pmap" code, with lots of local modifications to get it to fit into the BSD
>   kernel "style" and also increase performance (which wasn't really a concern
>   for the Mach folks).  The mmap() stuff was also borrowed somewhat from Mach,

The vm_objects, vm_maps, vm_fault, etc. are a much more significant
part. The pmap layer is just the lowest machine dependent part that
manages the actual page tables and physical page allocations. Mapping,
copy on write, inheritance, sharing, lookup, etc are managed by the
machine independent layers.

>   but modified even more radically.

Not being familiar with all the details I'd say it was basically
unixified. This includes changing the application interface and gluing
the code together with vnodes, the buffer cache etc.

The performance improvements are indeed valuable but the application
interface changes I wouldn't call improvement.

>   Pmap is an ugly hairy nasty mess.

At least it isolates the machine dependent mess from the rest of the
system. As for the machine independent code it might be hairy but if
you call it ugly you could as well call 99%+ of all software ugly (not
necessarily untrue).

	Johannes