SUMMARY: low-level formatting of SCSI disks on SparcStation

From: Howard Schultens (hs@demeter.ukps.gwdg.de)
Date: Mon Aug 17 1992 - 11:20:25 CDT


SUMMARY: low-level formatting of SCSI disks on SparcStation
From: Howard Schultens hs@demeter.ukps.gwdg.de

I received 11 replies, and their diversity shows that there is a) general
interest in this problem and b) no consensus about the severity of the
errors I ran into. Sorry for the delay in getting this summary out.

Thanks very much to all who replied -- the list is a great help! Particularly
(in order of appearance):

Jeffrey Marans <jeff@erie.irc.nrc.ca>
Casper Dik <casper@fwi.uva.nl>
danny@ews7.dseg.ti.com
bb@math.ufl.edu (Brian Bartholomew)
cfblack@unmvax.cs.unm.edu (Forrest Black)
David Fetrow <fetrow@biostat.washington.edu>
dkruse.sd@xerox.com (Dave G. Kruse)
Christian Damsgaard Jensen <sigfried@diku.dk>

--------------------------------------------------------------------------------
        Resumee
--------------------------------------------------------------------------------

No complete explanation of how to get rid of the errors, particularly the last
fatal non-media error with an absurd block number (4432/3/22 when the device
has highest cylinder number 1545). This might indicate that the disk has a
serious defect and is unusable. The SunOS format command actually issues
a "format unit" command on the SCSI bus (ie. does a "low-level format").
To do more than that requires detailed knowledge of the tweakable values
that can be set by raw SCSI commands. I tried varying the parameters in
format.dat, but that did not fix the problem.

--------------------------------------------------------------------------------
        Details
--------------------------------------------------------------------------------

I'm sorry I was unclear about the term "formatting" As Casper Dik noted,
MS-DOS "formatting" is what is called "initializing" a disk in some other
operating systems -- it is like newfs in SunOS. It does not mean that sector
and track header/trailer records are written onto the disk. That is called
"primary formatting", and THAT is exactly what I did on the PC. Even then,
it seems that not everything is erased from an SCSI disk: the defect list and
the label are still there. So the SCSI controller is "intelligent", i.e. there
is apparently a layer between you and the physics of magnetizable media.

Forrest Black says the symptoms look like incorrect drive geometry, and that
is what I suspect. He says it can be a real nightmare. I think the primary
formatting on the PC did not destroy the geometry completely; since it sort
of works on the Sun but with many errors, it looks like the geometry is near
what the Sun needs, it just needs to be synchronized with the particular
SCSI controller on the Sun again. As I said in my request, what I want to
do is issue the command "format unit" on the SCSI bus. Casper Dik explained
that the format program does exactly that, among other things. After trying
a bunch of things, the only error that now occurs when I tell format to
>format< is

Block 1835008 (4432/3/22), Fatal non-media error (hardware error)
failed

However, a surface analysis reveals many un-repairable errors.

Forrest Black says:

>The 4432/3/22 here suggests to me that you may be trying to format with
>very wrong parameters -- what is the drive's type, and what format.dat
>entry or manual parameters are you using? What about defect list
>management?
>
>As far as using a PC, well, it's possible that the PC host adapter has
>given the drive a new geometry that chokes the MD21. Oddly enough, when
>I run into a drive that won't format correctly on the Sparc, I take it
>to a PC and repeatedly attempt to reformat it until it takes, then it
>usually will format fine on the Sun. I *DONT* suggest this in your
>case, unless you use another brand of host adapter.
>
>Hope something here helps, and feel free to send email to me on this.

Dave Fetrow suggested double-checking the hardware, which I did, and gave
me hope that the MS-DOS system didn't make some irrevocable changes in the
disk unit.

Dave Kruse was able to identify the exact kind of disk I am dealing with
from the little information I gave:
>
>Could it be that you are trying to format a CDC WREN IV 94171-344? The
>messages from format look familiar. I have not been successful using Sun's
>format.dat on these disks.
>
>I just define a new disk type to format, and specify 1522 cylinders, 2
>alternates, 1524 physical, 9 heads, 46 sectors/track. This works on a good
>disk. However, I have not been successful if format finds, and trys to
>"repair" bad sectors.
>
>It may be that the firmware in the -344's is not compatable with Sun's. I have
>had low success rates when connecting these to SparcStations.
>

This concurred with Christian D. Jensen's suggestion to try relabelling the
disk and experimenting with the disk type.

I tried this, but it didn't correct the errors.

Very useful information was provided by SCSI guru Brian Bartholomew
(bb@math.ufl.edu), who told me what those strange messages with "PAGE
n: xxxxx" meant. That is actually text coming from the SCSI controller
itself. He dashed my hopes for an easy fix by saying that one might
have to fiddle with certain values stored in "mode pages" in the
controller, and that there is probably no tool for SunOS to do this.

                                :-) (-:

And lastly, please 'scuse that glaring spelling error: tUrpitude, not tErpitude.

                                :-) (-:

********************************************************************************
********************************************************************************
                        Original replies

I removed the "me, too" responses -- this summary is long enough. My original
request is included in these.

********************************************************************************
********************************************************************************

>From jeff@erie.irc.nrc.ca Mon Aug 10 16:37:50 1992
Return-Path: <jeff@erie.irc.nrc.ca>
Received: from erie.irc.nrc.ca (erie.irc.nrc.ca) by demeter.ukps.gwdg.de (4.1/SMI-4.1)
        id AA04078; Mon, 10 Aug 92 16:37:40 +0100
Message-Id: <9208101434.AA10069@ erie.irc.nrc.ca>
Date: Mon, 10 Aug 92 10:34:14 EDT
From: Jeffrey Marans <jeff@erie.irc.nrc.ca>
To: hs@demeter.ukps.gwdg.de
Subject: Re: Low-level formatting of SCSI disk
Status: R

I often do low level formats, but have to begin by labelling the disk.
I then have to extract a defect list and make that list current.

Jeff Marans.

********************************************************************************

>From casper@fwi.uva.nl Mon Aug 10 16:57:44 1992
Received: from mail.fwi.uva.nl by demeter.ukps.gwdg.de (4.1/SMI-4.1)
        id AA04092; Mon, 10 Aug 92 16:57:24 +0100
Received: by mail.fwi.uva.nl from adam.fwi.uva.nl with SMTP (5.65c(FWI)/2.0)
          id AA20698; Mon, 10 Aug 1992 16:55:44 +0200
Received: by adam.fwi.uva.nl from localhost.fwi.uva.nl with SMTP (5.65c(FWI)/2.0)
          id AA24430; Mon, 10 Aug 1992 16:55:42 +0200
Message-Id: <199208101455.AA24430@adam.fwi.uva.nl>
Return-Path: <casper@fwi.uva.nl>
X-Organisation: Faculty of Mathematics & Computer Science
                University of Amsterdam
                Kruislaan 403
                NL-1098 SJ Amsterdam
                The Netherlands
X-Phone: +31 20 525 7463
X-Telex: 10262 hef nl
X-Fax: +31 20 525 7490
To: hs@demeter.ukps.gwdg.de (Howard Schultens)
Subject: Re: Low-level formatting of SCSI disk
In-Reply-To: Your message of "Mon, 10 Aug 92 12:50:03 BST."
             <9208101150.AA03561@demeter.ukps.gwdg.de>
Date: Mon, 10 Aug 92 16:55:41 +0200
From: Casper Dik <casper@fwi.uva.nl>
Status: R

>Sun-managers:
>
>is there a way to perform a lowest-level format of an SCSI disk on the
>SparcStation 1? I made the mistake of formatting a defective disk on
>an MS-DOS PC for testing and now need to connect it back onto the SparcStation
>

There is no such thing as ``low level format''. This is a misconception
from the MS-DOS world. The MS-DOS program ``format'' writes out
directory structures and clears the FAT. (Comparable to newfs). Therefor
they had to invent a new phrase for what computer people call ``formatting''.

The SunOS format command does what in DOS-speak is called
``low level'' format. The SunOS format command does a number of
things (reading parameters from the service tracks) followed by a SCSI
format unit command. It seems to me that your disk is so defective
that formatting won't help.

Casper

********************************************************************************

>From danny@ews7.dseg.ti.com Mon Aug 10 18:38:18 1992
Return-Path: <danny@ews7.dseg.ti.com>
Received: from ti.com by demeter.ukps.gwdg.de (4.1/SMI-4.1)
        id AA04381; Mon, 10 Aug 92 18:38:12 +0100
Received: from ews7.dseg.ti.com ([128.247.212.189]) by ti.com with SMTP
        (5.59/LAI-3.2) id AA04646; Mon, 10 Aug 92 11:37:37 CDT
Received: from ews18.nbdc by ews7.dseg.ti.com (4.1/SMI-4.1)
        id AA00796; Mon, 10 Aug 92 11:36:30 CDT
Date: Mon, 10 Aug 92 11:36:30 CDT
From: danny@ews7.dseg.ti.com
Message-Id: <9208101636.AA00796@ews7.dseg.ti.com>
To: hs@demeter.ukps.gwdg.de
Subject: Re: Low-level formatting of SCSI disk
Status: R

never had this problem, but you may need to (re)select the drive
type. you can also use the "original" command in the defect
subsection to get the original (supposedly uncorrupted) defect
list, and then use it overwrite whatever defects are there.

********************************************************************************

>From bb@math.ufl.edu Mon Aug 10 19:16:15 1992
Return-Path: <bb@math.ufl.edu>
Received: from llull.math.ufl.edu by demeter.ukps.gwdg.de (4.1/SMI-4.1)
        id AA04410; Mon, 10 Aug 92 19:16:05 +0100
From: bb@math.ufl.edu
Received: by llull.math.ufl.edu (4.1/4.03)
        id AA19618; Mon, 10 Aug 92 13:14:37 EDT
Message-Id: <9208101714.AA19618@llull.math.ufl.edu>
To: hs@demeter.ukps.gwdg.de (Howard Schultens)
Subject: Re: Low-level formatting of SCSI disk
In-Reply-To: Your message of Mon, 10 Aug 92 12:50:03 +0100.
             <9208101150.AA03561@demeter.ukps.gwdg.de>
Date: Mon, 10 Aug 92 13:14:36 EDT
Status: R

> There must be an SCSI tool to access the host adapter "directly" and
> give the thing a "format unit" command, or some such. However, when
> I asked various technically knowledgeable vendors how to do this I
> often received a curious answer: this is not so much a technical
> problem as an issue of moral terpitude -- the disk was supplied
> pre-formatted and this is immutable; it was a degenerate thing to do
> to do a low-level format on a PC, the disk is forever damned to
> MS-DOS, and I am too, since the act was performed on my volition.

A symptom of ignorant vendors. You shouldn't be able to do anything
to a SCSI drive that you can't recover given sufficient understanding.
I suppose if the internal errorlist got corrupted that would be bad,
but that would take a hardware glitch or a bug in the drive firmware.

> PAGE 1: retries= 16 (0)
> Warning: Using default error recovery params.
>
> PAGE 2: inactivity limit= 40 (10)
> PAGE 4: cylinders= 1549 (1549) heads= 9 (9)
> PAGE 38: cache mode= 0x11 (0x11)
> min. prefetch multiplier= 0 max. prefetch multiplier= 0
> threshold= 255 (255) min. prefetch= 0 (0) max. prefetch= 53 (53)
> PAGE 3: trk skew= 5 (5) cyl skew= 16 (16) sects/trk= 46 (45)
> trks/zone= 9 (1) alt trks= 9 (18) alt sects/zone= 3 (1)

SCSI drives have large quantities of tweakable values, stored in "mode
pages". If your DOS formatter set the tweakables in a way that the
SunOS format(8) can't understand, it would explain this error.

Recovery? You need a program to send raw SCSI commands to the drive.
I have such a beast for the NeXT, but I haven't seen one for the Sun.
You also need to reset the tweakables to reasonable values, because
the behavior of "format unit" depends on the values. If you get the
drive manual you might find a command to set all the tweaks back to
the factory settings, in which case format(8) would probably work
again.

Good luck!

"Come work at math, where they pay their consultants a sysadmin's wage!"

Brian Bartholomew - bb@math.ufl.edu - Univ. of Florida Dept. of Mathematics

********************************************************************************

>From cfblack@unmvax.cs.unm.edu Mon Aug 10 20:45:03 1992
Return-Path: <cfblack@unmvax.cs.unm.edu>
Received: from unmvax.cs.unm.edu by demeter.ukps.gwdg.de (4.1/SMI-4.1)
        id AA04458; Mon, 10 Aug 92 20:44:43 +0100
Received: from coral.cs.unm.edu by unmvax.cs.unm.edu (5.61/3.3) with SMTP
        id <AA02548@unmvax.cs.unm.edu>; Mon, 10 Aug 92 12:43:05 -0600
From: cfblack@unmvax.cs.unm.edu (Forrest Black)
Message-Id: <9208101843.AA02548@unmvax.cs.unm.edu>
Subject: Re: Low-level formatting of SCSI disk
To: hs@demeter.ukps.gwdg.de
Date: Mon, 10 Aug 92 12:43:04 MDT
In-Reply-To: <9208101150.AA03561@demeter.ukps.gwdg.de>; from "Howard Schultens" at Aug 10, 92 12:50 pm
X-Mailer: ELM [version 2.3 PL8]
Status: R

>
> Sun-managers:
>
> is there a way to perform a lowest-level format of an SCSI disk on the
> SparcStation 1? I made the mistake of formatting a defective disk on
> an MS-DOS PC for testing and now need to connect it back onto the SparcStation
> again. When I hook it up and try >format< I get error messages:
>
> format> format
> Ready to format. Formatting cannot be interrupted
> and takes 16 minutes (estimated). Continue? y
> Beginning format. The current time is Mon Aug 10 11:00:42 1992
>
> PAGE 1: retries= 16 (0)
> Warning: Using default error recovery params.
>
> PAGE 2: inactivity limit= 40 (10)
> PAGE 4: cylinders= 1549 (1549) heads= 9 (9)
> PAGE 38: cache mode= 0x11 (0x11)
> min. prefetch multiplier= 0 max. prefetch multiplier= 0
> threshold= 255 (255) min. prefetch= 0 (0) max. prefetch= 53 (53)
> PAGE 3: trk skew= 5 (5) cyl skew= 16 (16) sects/trk= 46 (45)
> trks/zone= 9 (1) alt trks= 9 (18) alt sects/zone= 3 (1)
>
> Block 1835008 (4432/3/22), Fatal non-media error (hardware error)
> during test unit ready
> failed
>
> When I try a surface analysis "compare", the operation is performed,
> but there are many bad blocks that cannot be repaired. There must be
> an SCSI tool to access the host adapter "directly" and give the thing a
> "format unit" command, or some such. However, when I asked various
> technically knowledgeable vendors how to do this I often received a
> curious answer: this is not so much a technical problem as an issue of
> moral terpitude -- the disk was supplied pre-formatted and this is
> immutable; it was a degenerate thing to do to do a low-level format on
> a PC, the disk is forever damned to MS-DOS, and I am too, since the act was
> performed on my volition.
>
> Does this mean that the host adapter on the Sparc Station (Emulex MD21??)
> is not capable of formatting the disk from scratch, i.e. simply as disks
> of magnetizable medium? Or does it need labels, defect lists, and so forth
> to be able to do this? Does the "non-media error" message mean that the
> controller on the disk is busted?
>
> Anyone have experience with this? Thanks muchly in advance,
>
> --------------------------------------------------------------------------------
> \ ..\ /../ Howard Schultens Tel: ++49 551 39 5914
> \.o.\ /../ Zentrum Physiologie FAX: ++49 551 39 5923
> \o.o> /: / Abteilung Neuro- und
> \o/ / \ Sinnesphysiologie E-Mail: hs@demeter.ukps.gwdg.de
> Y R .oo \ Humboldtallee 23
> / o o.o \ D-3400 Goettingen
> <_*o*_*.*_> Germany
> --===r=R=r=Rr===----------------------------------------------------------------
>

It sounds like the drive's geometry is now incorrect. I've been through
this before, and it can be a real nightmare.

By 'defective', I assume that the drive was returning read errors in
some places rather than data. You should use the 'repair' command in
the SunOS format program rather than re-formatting the disk, unless the
number of defects was very large.

> Block 1835008 (4432/3/22), Fatal non-media error (hardware error)

The 4432/3/22 here suggests to me that you may be trying to format with
very wrong parameters -- what is the drive's type, and what format.dat
entry or manual parameters are you using? What about defect list
management?

As far as using a PC, well, it's possible that the PC host adapter has
given the drive a new geometry that chokes the MD21. Oddly enough, when
I run into a drive that won't format correctly on the Sparc, I take it
to a PC and repeatedly attempt to reformat it until it takes, then it
usually will format fine on the Sun. I *DONT* suggest this in your
case, unless you use another brand of host adapter.

Hope something here helps, and feel free to send email to me on this.

Forrest Black
cfblack@unmvax.cs.unm.edu

********************************************************************************

>From fetrow@biostat.washington.edu Tue Aug 11 01:21:23 1992
Return-Path: <fetrow@biostat.washington.edu>
Received: from orac.biostat.washington.edu by demeter.ukps.gwdg.de (4.1/SMI-4.1)
        id AA04713; Tue, 11 Aug 92 01:21:20 +0100
Received: by orac.biostat.washington.edu
        (5.64/UW-NDC Revision: 2.21 ) id AA04903; Mon, 10 Aug 92 16:18:12 -0700
From: David Fetrow <fetrow@biostat.washington.edu>
Message-Id: <9208102318.AA04903@orac.biostat.washington.edu>
Subject: Re: Low-level formatting of SCSI disk
To: hs@demeter.ukps.gwdg.de
Date: Mon, 10 Aug 92 16:18:11 PDT
In-Reply-To: <9208101150.AA03561@demeter.ukps.gwdg.de>; from "Howard Schultens" at Aug 10, 92 12:50 pm
X-Mailer: ELM [version 2.3 PL11]
Status: R

 I've certainly done low-level formats before on uninitialized disks.
I can't picture even an MS-DOS SCSI controller screweing things up so
badly that there is no hope but am not a SCSI-knlwedgeable guy.

 My only guesses would be make sure you have the right format.dat for
the drive and everything is hooked up nicely and all the jumpers are set
for Unixland rather than DOSland.

-- 
 - dave fetrow-					fetrow@biostat.washington.edu

********************************************************************************

>From Dave_G._Kruse.sd@xerox.com Tue Aug 11 04:28:39 1992 Return-Path: <Dave_G._Kruse.sd@xerox.com> Received: from alpha.xerox.com by demeter.ukps.gwdg.de (4.1/SMI-4.1) id AA04750; Tue, 11 Aug 92 04:28:34 +0100 Received: from Apollo.sd.xerox.xns by alpha.xerox.com via XNS id <11709>; Mon, 10 Aug 1992 19:27:01 PDT X-Ns-Transport-Id: 08003700AD1BBDDC2E15 Date: Mon, 10 Aug 1992 19:26:34 PDT Sender: Dave_G._Kruse.sd@xerox.com From: dkruse.sd@xerox.com Subject: Re: Low-level formatting of SCSI disk In-Reply-To: "sun-managers-relay%ra.mcs.anl:gov's message of 10 Aug 92 04:50:03 PDT (Monday)" To: hs@demeter.ukps.gwdg.de Cc: dkruse.sd@xerox.com Message-Id: <"10-Aug-92 19:26:34 PDT".*.Dave_G._Kruse.sd@Xerox.com> Status: R

Hello,

Could it be that you are trying to format a CDC WREN IV 94171-344? The messages from format look familiar. I have not been successful using Sun's format.dat on these disks.

I just define a new disk type to format, and specify 1522 cylinders, 2 alternates, 1524 physical, 9 heads, 46 sectors/track. This works on a good disk. However, I have not been successful if format finds, and trys to "repair" bad sectors.

It may be that the firmware in the -344's is not compatable with Sun's. I have had low success rates when connecting these to SparcStations.

Hope this helps, Dave

********************************************************************************

>From sigfried@skinfaxe.diku.dk Tue Aug 11 11:03:42 1992 Return-Path: <sigfried@skinfaxe.diku.dk> Received: from odin.diku.dk by demeter.ukps.gwdg.de (4.1/SMI-4.1) id AA05377; Tue, 11 Aug 92 11:03:00 +0100 Received: from skinfaxe.diku.dk by odin.diku.dk with SMTP id AA09768 (5.65+/IDA-1.3.5 for hs@demeter.ukps.gwdg.de); Tue, 11 Aug 92 11:01:05 +0200 Received: by skinfaxe.diku.dk id AA21348 (5.65+/IDA-1.3.5 for hs@demeter.ukps.gwdg.de); Tue, 11 Aug 92 11:00:44 +0200 From: Christian Damsgaard Jensen <sigfried@diku.dk> Message-Id: <9208110900.AA21348@skinfaxe.diku.dk> Subject: Re: Low-level formatting of SCSI disk To: hs@demeter.ukps.gwdg.de Date: Tue, 11 Aug 92 11:00:42 MET DST In-Reply-To: <9208101150.AA03561@demeter.ukps.gwdg.de>; from "Howard Schultens" at Aug 10, 92 12:50 pm X-Mailer: ELM [version 2.3 PL11] Status: R

Howard Schultens writes: > > Sun-managers: > > is there a way to perform a lowest-level format of an SCSI disk on the > SparcStation 1? I made the mistake of formatting a defective disk on > an MS-DOS PC for testing and now need to connect it back onto the > SparcStation again.

My guess is that the disklabel is corrupt, and the defectlist is gone too. PC's don't use disklabels, and it has probably been overwritten when it was formatted on the PC. Try playing with the format program, start with defining the disk type (the "type" command), then partition the disk and recreate the manufacturers defect list on the disk. After the defect list has been recreated, you will have to format the disk again, but the result will hopefully be better.

regards

Christian D. Jensen. -- Christian Damsgaard Jensen (System Administrator) Email: sigfried@diku.dk DIKU (Dept. Comp. Sci. Univ. Copenhagen) Fax: +45 31 39 02 21 Universitetsparken 1 DK-2100 Copenhagen, Denmark

********************************************************************************



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:47 CDT