From owner-freebsd-current  Mon Mar 25 09:20:35 1996
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id JAA13747
          for current-outgoing; Mon, 25 Mar 1996 09:20:35 -0800 (PST)
Received: from GndRsh.aac.dev.com (GndRsh.aac.dev.com [198.145.92.241])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id JAA13737
          Mon, 25 Mar 1996 09:20:28 -0800 (PST)
Received: (from rgrimes@localhost) by GndRsh.aac.dev.com (8.6.12/8.6.12) id JAA04958; Mon, 25 Mar 1996 09:18:52 -0800
From: "Rodney W. Grimes" <rgrimes@GndRsh.aac.dev.com>
Message-Id: <199603251718.JAA04958@GndRsh.aac.dev.com>
Subject: Re: Adaptec 1542B and 1522 on current kernels
To: toor@dyson.iquest.net (John S. Dyson)
Date: Mon, 25 Mar 1996 09:18:52 -0800 (PST)
Cc: syssgm@devetir.qld.gov.au, gibbs@freefall.freebsd.org,
        freebsd-current@freebsd.org
In-Reply-To: <199603251315.IAA25614@dyson.iquest.net> from "John S. Dyson" at "Mar 25, 96 08:15:59 am"
X-Mailer: ELM [version 2.4ME+ PL11 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-current@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

> > 
> > My problem turned up after an upgrade (amongst other things) of my processor
> > from a 486DX33 to a 486DX4/100.  I run FreeBSD 2.0.5 on it.  I got lots of
> > sd0: timed out and MBO was 01 should be 00.  My solution was to attempt to
> > avoid immediate reuse of mailboxes, based on a theory of how my BT545S could
> > be postponing mailbox updates when it was really busy.  The MBO problems
> > disappeared, but the timeouts remained.
> > 
> Certainly not to dispute problems that some people have been having, but
> using the bt driver with the BT545S, I had *extremely* stable SCSI subsystem
> operation.  My machine was a 486/66 non-write back (early CPU.)

> I wonder if some MBs don't support cache invalidation properly?

I can confirm for you that there are broken cache designs out there that
do not do proper invalidation.  Most often it is the low cost boards with
L2 write back caches, these can often be made to work in write through
mode, but some are so broken they don't even get that right.

-- 
Rod Grimes                                      rgrimes@gndrsh.aac.dev.com
Accurate Automation Company                 Reliable computers for FreeBSD