*BSD News Article 21247


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!elroy.jpl.nasa.gov!swrinde!cs.utexas.edu!natinst.com!hrd769.brooks.af.mil!hrd769.brooks.af.mil!not-for-mail
From: burgess@hrd769.brooks.af.mil (Dave Burgess)
Newsgroups: comp.os.386bsd.bugs
Subject: Re: Problems with installation of NetBSD 0.9...
Date: 21 Sep 1993 08:26:13 -0500
Organization: Armstrong Laboratory, Brooks AFB, TX
Lines: 34
Distribution: world
Message-ID: <27mve3$plb@hrd769.brooks.af.mil>
References: <27ldlm$j27@acsc.com> <AARON.93Sep21184413@downstage.comp.vuw.ac.nz>
NNTP-Posting-Host: hrd769.brooks.af.mil

In article <AARON.93Sep21184413@downstage.comp.vuw.ac.nz> Aaron.Roydhouse@comp.vuw.ac.nz (Aaron Roydhouse) writes:
>==> "Frank" == Frank Mayhar <fmayhar@acsc.com> writes:
>
>Frank> before fsck runs. (I also see a message indicating an
>Frank> unexpected interrupt from wdc0 at this point, but I don't know
>Frank> whether that is related.) If it happens during boot, the system
>Frank> hangs at that point.
>
>I have NetBSD 0.9 on a 486dx2/33 and two identical IDE drives sharing
>the one controller. I began to experience just the behaviour you
[...]
>on boot-up; well I got the "ISA strayintr 7" anyway but it didn't lead
>to a hang. This isn't a show-stopper for me so long as I avoid the
>"fsck -p" during bootup and run it myself afterwards.
>

I think that we have a considerable quantity of anecdotal evidence that
there is at least one timing problem in the wd.c code.  Heavy disk usage
causing a hard drive hang (versus a coplete system hang) is a known bug,
and there is even some mention of it in the FAQ (which is available by
anonymous FTP in pub/FAQ:hrd769.brooks.af.mil).  I had a similar problem
while using an Ultrastore 24F in IDE emulation mode.  It is reasonable
to assume then that it is not a hardware problem, but a software
problem.  The description in the FAQ gives a pretty good explanation of
a cause and solution.  The solution is kind of 'brute force' but fixes
most of these hangs after the fact (except for mine, where the reset
semantics are just 'a little bit' different).


-- 
------
TSgt Dave Burgess
NCOIC AL/Management Information Systems Office
Brooks AFB, TX