From owner-freebsd-current  Thu Sep 17 18:09:13 1998
Return-Path: <owner-freebsd-current@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id SAA23976
          for freebsd-current-outgoing; Thu, 17 Sep 1998 18:09:13 -0700 (PDT)
          (envelope-from owner-freebsd-current@FreeBSD.ORG)
Received: from morgan.angis.su.OZ.AU (morgan.angis.su.OZ.AU [129.78.25.38])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id SAA23796
          for <current@FreeBSD.org>; Thu, 17 Sep 1998 18:08:12 -0700 (PDT)
          (envelope-from tonym@angis.usyd.edu.au)
Received: (from tonym@localhost) by morgan.angis.su.OZ.AU (8.8.7/8.6.6) id LAA25488; Fri, 18 Sep 1998 11:07:40 +1000 (EST)
Date: Fri, 18 Sep 1998 11:07:40 +1000 (EST)
From: Tony Maher <tonym@angis.usyd.edu.au>
Message-Id: <199809180107.LAA25488@morgan.angis.su.OZ.AU>
To: gibbs@narnia.plutotech.com
Subject: Re: CAM this and that
Cc: current@FreeBSD.ORG
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

> I shouldn't lock up.  What kind of tape device do you have?  Were there
> any console messages leading up to the hang?

I'll try again over weekend to see if I can reproduce and send details then.

>>    change 
>>    Queue Algorithm Modifier:  0 -> 1
> 
> Don't even think about doing this.  If your other disks have this
> turned on, you are begging for problems.  Setting the Q.A.M to
> 1 allows the drive to re-order transactions without regard to any
> write integrity (i.e. write data to LBA X, read data to LBA X, read
> is serviced before write and you get stale data).

Ok thanks!
I only modified the disk which doesn't report Tagged queueing 
to match what I saw on the other two.

One of these disks was with the system when bought new, the other came off
solaris box at work (it was flaky). I have never modified the scsi info and
all the disks were installed using sysinstall.

Is this how disks are configured from factory?
Should I (everyone) check all disks installed?


>>    DQue:  1 -> 0
> 
> Having DQue as 1 will prevent us from attempting tagged queuing.  You
> would have to save this change and reboot for the system to notice this

It was saved and rebooted but still didn't want to do tagged queuing.
Will try again.

>>    (da2:ahc0:0:2:0): no longer in timeout, status = 34b
>>    ahc0: Issued Channel A Bus Reset. 17 SCBs aborted
> 
> That really looks like a bogus cable or termination.  The fact that
> the problem seems to get better after boot (i.e the device warms up)
> also points to a setup/connector problem.

Termination is correct but the cable could be suspect. Will try a replacement.

(This is old disk which was removed from work machines due to a couple of
glitches.  I have been using it at home without problem for about 12 months
but under little load)

Thanks again.

tonym

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message