Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!yoyo.cc.monash.edu.au!mmcg
From: mmcg@yoyo.cc.monash.edu.au (Mike McGaughey)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: Strange install timeouts
Date: 21 Jan 97 10:38:29 GMT
Organization: Monash University
Lines: 51
Message-ID: <mmcg.853843109@yoyo.cc.monash.edu.au>
References: <5bqoee$km2@news1.iamerica.net> <5bt15q$8vt@uriah.heep.sax.de>
NNTP-Posting-Host: yoyo.cc.monash.edu.au
X-NNTP-Posting-User: mmcg
X-Newsreader: NN version 6.5.0 #4 (NOV)
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:34259
j@uriah.heep.sax.de (J Wunsch) writes:
>me@i.hate.spam.com (James S.) wrote:
>> wd1: interrupt timeout:
>> wd1: status 58<rdy, seekdone, drq> error 0
>> wd1: interrupt timeout:
>> wd1: status 58<rdy, seekdone, drq> error 1<no_dam>
>Your drives don't seem to talk nicely on the IDE bus. Maybe swapping
>the controller out for another one would help, but mind you, this kind
>of errors is why many people (including me) hate IDE. It's fine as
>long as it works, but terrible as soon as a problem pops up.
Joerg, so cynical :)
Your drive or motherboard is probably trying to be smart and spin down
the disk when it's been idle a while (that's an exclusive or). FreeBSD
times out while waiting for the drive to spin up again.
This happens on one of my Sirocco 2550AT IDE drives in a 3-yo pentium,
but not on a similar drive in a younger computer (the older computer
also has a scsi drive, which may or may not make a difference to this
behaviour; the younger computer has a cheaper but more recent motherboard.
One day I'll swap the hardware around and see what causes what).
The following (which I start from rc.local) solved it (warning - *not*
cut 'n pasted):
#!/bin/sh
while true
do
sleep 30
for i in `df | sed '1d' | grep '^/dev/w[sd]' | awk ' { print $(NF)}' `
touch $i/.keep_disk_spinning
done
done
Adjust timeout to taste. In any case, hearing the disks seek
every half minute on an otherwise quiescent machine is considerably less
annoying than hearing the bloody things spinning up after each rwho ping
comes in (especially when you're trying to grab a nap :)
BTW: Is it possible for the kernel to know when a disk has been
spun down or is being spun up?
Cheers,
Mike.
ps: my IDE drives are faster than my scsi, twice as big,
and cost half as much, so nyah :)