From owner-freebsd-current  Fri Aug  8 17:18:43 1997
Return-Path: <owner-freebsd-current>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id RAA25045
          for current-outgoing; Fri, 8 Aug 1997 17:18:43 -0700 (PDT)
Received: from shell.uniserve.com (tom@shell.uniserve.com [204.244.210.252])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id RAA25040
          for <freebsd-current@FreeBSD.ORG>; Fri, 8 Aug 1997 17:18:38 -0700 (PDT)
Received: from localhost (tom@localhost)
	by shell.uniserve.com (8.8.5/8.8.5) with SMTP id RAA28233;
	Fri, 8 Aug 1997 17:15:05 -0700 (PDT)
X-Authentication-Warning: shell.uniserve.com: tom owned process doing -bs
Date: Fri, 8 Aug 1997 17:15:05 -0700 (PDT)
From: Tom <tom@uniserve.com>
To: Alex <garbanzo@hooked.net>
cc: "John W. DeBoskey" <jwd@unx.sas.com>, freebsd-current@FreeBSD.ORG
Subject: Re: scsi time-out & lockup under smp
In-Reply-To: <Pine.BSF.3.96.970808095710.27428C-100000@zippy.dyn.ml.org>
Message-ID: <Pine.BSF.3.96.970808171158.28159A-100000@shell.uniserve.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-current@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk


On Fri, 8 Aug 1997, Alex wrote:

> On Fri, 8 Aug 1997, John W. DeBoskey wrote:
> 
> > sd0: SCB 0x1 - timed out in command pahse, SCSISIGI == 0x86
> > SEQADDR = 0x8c SCSISEQ = 0x12 SSTAT0 = 0x7 SSTAT1 = 0x3
> > sd0: abort message in message buffer
> > sd0: SCB 1 - Abort Completed.
> > sd0: no longer in timeout
> > sd0: SCB 0x1 - timed our while idle, LASTPHASE == 0x1, SCSISIGI = 0x0
> > SEQADDR = 0xb SCSISEQ = 0x12 SSTAT0 = 0x5 SSTAT1 = 0x2
> > sd0: Queueing an Abort SCB
> > 
> > 
> >    It only seems to occur when I start to initiate heavy disk io. It
> > does not happen in the uni-proccesor situation. The complete output
> > from dmesg is appended to this mail. If anyone can help me track this
> > down, I'd really appreciate it.
> > 
> > Thanks,
> > John
> 
> This is a fairly well known problem with the 2940U/UW, and probably
> anything else based on the 7880.  All I can say is that for heavy disk io,
> don't use that controller, two controllers might be causing some of those
> problems too.  However I'd also suggest that you take the number of busses
> down a notch, unless you have devices on bus 255, and that you cvsup the
> most recent FreeBSD sources,as I think some improvments have been made to 
> this driver, and many improvments have been made to the smp code.
> 
> - alex

  Well known problem?  I've never seen it, even with heavy io on a 3940UW.
I use AHC_TAGENABLE, but NOT AHC_SCBPAGING.

  I don't believe that any changes have been made to the ahc driver
lately.

  I think this problem is device dependant.  Certain devices will seize
the bus, and driver is unable to get it back.

Tom