From owner-freebsd-questions  Tue Feb 27 07:13:46 1996
Return-Path: owner-questions
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id HAA25203
          for questions-outgoing; Tue, 27 Feb 1996 07:13:46 -0800 (PST)
Received: from localhost.cdrom.com (localhost.cdrom.com [127.0.0.1])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id HAA25190
          Tue, 27 Feb 1996 07:13:40 -0800 (PST)
Message-Id: <199602271513.HAA25190@freefall.freebsd.org>
X-Authentication-Warning: freefall.freebsd.org: Host localhost.cdrom.com [127.0.0.1] didn't use HELO protocol
To: se@zpr.uni-koeln.de (Stefan Esser)
cc: Dave Andersen <angio@aros.net>, questions@freebsd.org
Subject: Re: Ncr scsi: overlapped commands attempted 
In-reply-to: Your message of "Tue, 27 Feb 1996 13:45:07 +0100."
             <199602271245.AA29271@Sysiphos> 
Date: Tue, 27 Feb 1996 07:13:39 -0800
From: "Justin T. Gibbs" <gibbs@freefall.freebsd.org>
Sender: owner-questions@freebsd.org
Precedence: bulk

>} The only thing that comes to mind is that an untagged request
>} sense command is pending when another tagged command is queued
>} to the target.  The aic7xxx driver never tagges request sense commands
>} since the spec dictates that they must be untagged (and some drives
>} choke on this) and it also gurantees that no other commands are posted
>} to the target until the request sense completes successfully.
>
>Yes, I know about this, and teh NCR driver makes
>sure that no tags are used with REQUEST SENSE 
>commands.

Okay.

>The HP drive problems are most likely caused by
>several tagged START STOP UNIT commands being sent
>to the drive. Most drives do the right thing (i.e.
>ignore a start unit if the motor is running :), 
>but a few seem to perform a soft reset and don't
>tolerate several of these commands being active 
>at a time.

Hmmm.  This problem again.

>The GENERIC SCSI code calls scsi_start_unit() 
>whenever a raw disk partition is opened, and two
>programs do several opens within a millisecond: 
>FSCK and DUMP.
>
>I've been thinking about sending only teh first
>start unit to each LUN, but for now I think the 
>generic SCSI code should NOT issue them in the 
>first place for EACH open.

I think we should only send a start unit during
bus probe and in certain cases during error
handling (unit not ready?).  It sounds too dangerous
to do it any more often.

>Regards, STefan
>-- 
> Stefan Esser, Zentrum fuer Paralleles Rechnen		Tel:	+49 221 4706021
> Universitaet zu Koeln, Weyertal 80, 50931 Koeln	FAX:	+49 221 4705160
> =============================================================================
>=
> http://www.zpr.uni-koeln.de/~se			  <se@ZPR.Uni-Koeln.DE>

--
Justin T. Gibbs
===========================================
  FreeBSD: Turning PCs into workstations
===========================================