Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!haven.umd.edu!darwin.sura.net!newsserver.jvnc.net!yale.edu!nigel.msen.com!math.fu-berlin.de!unlisys!max.IN-Berlin.DE!not-for-mail
From: berry@max.IN-Berlin.DE (Stefan Behrens)
Newsgroups: comp.os.386bsd.bugs
Subject: Re: wd stops on update
Date: 18 May 1993 23:42:40 +0200
Organization: Private
Lines: 43
Message-ID: <1tbl8u$2al@max.IN-Berlin.DE>
References: <1tafgi$5s7@zaphod.axion.bt.co.uk>
NNTP-Posting-Host: max.in-berlin.de
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
In article <1tafgi$5s7@zaphod.axion.bt.co.uk> lessen@axion.bt.co.uk (Lee Essen) writes:
>This is a bit of a strange problem! At completely random times, luckily with
>a large MTBF, my wd light comes on and stays on and no more disk access is
>permitted, the processes just block!
>This was really annoying yesterday when I set my machine up at home for slip
>and dialed in from work...
This was annoying me for months, but it is fixed now.
It is caused by the wd.c driver and for the most part by the
com.c driver. It happens when you use the serial lines.
The problem is, that your controller is not able or does not
want to hold the interrupt-request-line for a long time. The
original 0.1 com.c driver and also cgd's driver locked out
interrupts for a very long time, up to 100ms I think.
My first fix for this was adding timeout handling to the
wd.c driver. I added it to the Barsoom wd-driver with the
extension for multi-sector-read/write requests.
The timeout code really fixed the problem, whenever the
controller gets locked and the driver lost the interrupt,
the timer recognized the situation and cleared the condition.
However the main reason for the problem was the com-driver.
So I tried the one from Bruce Evans. With success! No locked
wd-controller since then even without the timeout stuff. And
what is more I have *never* seen the ``silo overflow''
messages again. Even when using two modems and the serial
mouse at the same time.
For all uucp-links I use the e-protocol which does no error
correction or recognition at all, without any errors yet.
All on a three years old i38[67]@33Mhz with a V32bis/V42bis
modem at 38400cps and 57600cps, the UART is an AFN16550 one.
Get the barsoom-wd-driver and Bruce Evans com-stuff from your
favorite 386BSD ftp site in
.../386BSD/unofficial/from-ref/sys.extras/barsoom-wd-driver/*
.../386BSD/unofficial/from-ref/sys.extras/mod-barsoom-wd-driver/*
.../386BSD/unofficial/???maybe bde???/{com-0.2,intr-0.0,npx-0.5}.tar.Z
BTW
has someone hacked Julians st.c stuff so it can write to
Wangtek 5150ES drives???
--
Stefan (berry@max.IN-Berlin.DE)