NetBSD Problem Report #10502
Received: (qmail 13433 invoked from network); 4 Jul 2000 08:10:59 -0000
Date: Tue, 4 Jul 2000 09:11:29 +0100 (BST)
Subject: CATS panic under heavy IDE usage
>Synopsis: CATS panic under heavy IDE usage
>Arrival-Date: Tue Jul 04 08:11:00 +0000 2000
>Last-Modified: Sat Nov 12 16:01:55 +0000 2005
>Originator: Stephen Borrill
>Release: 25th June 2000
Dr. Stephen Borrill, Director and Principal Engineer
Precedence Technologies Ltd Tel: +44 (0) 1223 562 500
86 Kings Hedges Road Fax: +44 (0) 1223 563 522
Cambridge, CB4 2PA, United Kingdom WWW: http://www.precedence.co.uk/
System: NetBSD black 1.4ZD NetBSD 1.4ZD (CATS192.1.4ZD) #0: root@cats:/usr/src/sys/arch/arm32/compile/CATS192.1.4ZD arm32
CATS system (64 or 128MB RAM). 2 x 20GB IDE discs (bus masters).
Still happens on 1.5ZC.
A CATS system can have up to 4 IDE devices attached. On heavy disc access, e.g. tarring data between
two discs, the system will completely freeze solid. This has been reported both on single drives (e.g. building
Jikes Java compiler), but is much more reproducible on two drive systems where any heavy access on two
drives at once will cause the machine to hang. This happens whether the two drives share a bus (i.e. master/slave)
or are on separate busses. The problem is so pervasive that after a crash, when fsck starts to check in parallel,
the machine will lock again (and thus to the user, the system appears to have completely died as it won't reboot).
This is a long-established problem. I've seen it under 1.3F, 1.4K, 1.4P and 1.4ZD kernels.
Have a CATS with two drives. If either of the following commands are run separately, the system will
work fine. If they are run together, the system will hang at some point.
dd if=/dev/wd0e of=/dev/null bs=10240k
dd if=/dev/wd1e of=/dev/null bs=10240k
mount /dev/wd0e /usr
mount /dev/wd1e /mnt
(cd /usr ; tar -cf - .) | (cd /mnt ; tar -xf -)
The latter seems to mainly fail when very large files are being copied (e.g. CD images).
State-Changed-When: Wed Oct 3 02:16:25 PDT 2001
I've looked into the PR the lockup occurs when the interrupts get confused, it seems that something goes wrong with the irq handling on the isa bus. Matt T made a comment that he's got a new schema for handling this kinda thing on the xscale board so once I've seen that I may well update cats to use that.
Responsible-Changed-When: Wed Oct 3 02:19:16 PDT 2001
This is a cats specific issue
From: "Chris Gilbert" <email@example.com>
Subject: Re: port-cats/10502
Date: Thu, 4 Apr 2002 21:34:47 +0100
Just an update on this PR.
The best test case is to boot to single user mode (or you risk trashing your
file systems!) and running:
dd if=/dev/wd0e of=/dev/null bs=10240k &
dd if=/dev/wd1e of=/dev/null bs=10240k &
give it a minute or 2 and it locks up.
Planning to investigate further but this occurs on 1.5ZC. And is likely to
be in 1.6 *sigh*
Subject: RE: port-cats/10502
Date: Mon, 20 May 2002 22:57:37 +0100 (BST)
It appears that this is some kind of UDMA issue, setting flags on wd to
0x0fac disables UDMA and all appears well.
I'll start investigating it as a UDMA issue.
$NetBSD: query-full-pr,v 1.39 2013/11/01 18:47:49 spz Exp $
$NetBSD: gnats_config.sh,v 1.8 2006/05/07 09:23:38 tsutsui Exp $
Copyright © 1994-2007
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.