*BSD News Article 37876


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!uwm.edu!lll-winken.llnl.gov!enews.sgi.com!decwrl!vixie!nnrp!vixie
From: vixie@gw.home.vix.com (Paul A Vixie)
Newsgroups: comp.unix.bsd
Subject: Re: HELP: installing netbsd+64mb RAM+adaptec
Date: 14 Nov 94 10:10:11
Organization: Vixie Enterprises
Lines: 26
Message-ID: <VIXIE.94Nov14101011@gw.home.vix.com>
References: <39vogs$78p@dagny.galt.com> <MICHAELV.94Nov11143004@mindbender.headcandy.com>
	<3a1f98$pil@pdq.coe.montana.edu> <Cz9J3I.EyC@cogsci.ed.ac.uk>
NNTP-Posting-Host: gw.home.vix.com
In-reply-to: richard@cogsci.ed.ac.uk's message of Mon, 14 Nov 1994 15:02:05 GMT

99% or so of the 386/486 machines in the world have an ISA bus.
80% of these have no other bus.
NetBSD _needs_ bounce buffers.
FreeBSD and BSD/OS already have them.

Nobody cares whether it is or isn't a hack.  If we were going to shy away
from obsolete designs, we wouldn't be using CPU's descended from sewing
machines on computers designed to have 64KB of memory and cassette tape
drives.

What matters is whether somebody can grab the bits, dump 'em in, and get
action.  If NetBSD won't run on somebody's machine because they have more
than 16MB of memory and only an ISA bus, they don't care whether or not
bounce buffers are a hack, they'll just say "thank you for playing" and
move on to Linux or whatever.  Not a good way to prove your point.

Now that we're in violent agreement, I suggest that you do what I did in
the TNIC driver: grab a bunch of MCL's during initialization and use them
as copy buffers if you ever get a live MCL during operation that is above
the 16MB point.  I see no performance problems since most MCL's are in fact
from the <16MB point, this is just a corner case anyway.
--
Paul Vixie
La Honda, CA
<paul@vix.com>
decwrl!vixie!paul