[812] in linux-scsi channel archive
Re: SCSI Harddisk Utility/Linux ?
daemon@ATHENA.MIT.EDU (Juergen Sauer)
Thu Oct 17 20:59:55 1996
Date: Thu, 17 Oct 1996 19:56:23 +0200 (MET DST)
From: Juergen Sauer <jojo@sae.extern.fh-hannover.de>
To: Michael Taeschner <Michael.Taeschner@student.uni-magdeburg.de>
Cc: linux-scsi@vger.rutgers.edu
In-Reply-To: <Pine.HPP.3.91.961016105625.28662A-100000@dfnserv1.URZ.Uni-Magdeburg.DE>
On Wed, 16 Oct 1996, Michael Taeschner wrote:
> On Wed, 16 Oct 1996, Juergen Sauer wrote:
> > The badblocks command does it not very well. It works great on a
> > real dump device like IED drives but fails on inteligent SCSI devices.
> > A SCSI device hides any information of defective blocks until the spare
> > blocks are running out. Upto this point the badblocks will never find
> > a 'bad' block on the device.
>
> Is this feature on by default or must it be toggled by some tool
> like scsi-config / scsiinfo ?
This can be done by sending special commands to the device over a generic
interface to it. It's far away from standard operation.
> > shutdowned: not aceptable. And you in the hell wants to use a adaptec
> > controller, even the ncr810 serieses does the job better ?
>
> Why do you say this? I'm having big problems at the moment with my noname
> ncr810 (a passive terminated one) in an UM8810P AIO board (BIOS 3.2h, SDMS
> 3.07) running 2 scsi-II hd and a cdrom. I'm looking for a test to verify that
> cabeling, termination etc is 100% ok.
Which Chipset does the Mobo has ?
> After power on everything seems to be ok, after some time DOS-programs
> like scandisk (surface-test) or msbackup hang. This can only be overcome by
> a switch for handling timeouts (probably issuing an scsi-reset).
>
> Even on linux I got badblocks reported from sda, the same in all four
> write-cycles. No entries in grown defects list. Yesterday I even got (after
> some hours of operation) ext2fs Kernel Panics, not being able to read some
> bitmap block.
>
> I'm near to throw the whole thing out of the window... Does this sound
> like termination, cabeling or drive probs?
I encountered the same/simmular things things many times. Therewere
multiple causes:
- first prob: ASUS SP3 MoBo and pci plugged in controller like yours
solution: dumping the board into a novell server, which drivers
has got a work around for a PCI chipset bug. In linux this work
around was missing. (Unkown to Drew eckard, he couldn't repeat the
error himself) Linux 1.3.92 until now, not before)
- second prob: ASUS SP3G the same ...
- third one: I encounterd BUS errors onto a very full SCSI BUS
on a ASUS P55TP4. Solution was: bad cabeling, the cabel distance
between my CD Writer CDR100 and the next devices were too short.
Remember the SPECs: 15 cm minimum!
- fourth one: A Bus were running with a Adaptec 2940, but not with
Ncr810. Solution was: The Adaptec didn't run the bus synchronous
but the Ncr did. The wrong termination/Term Power injection points
did the rest of the crash.
- fifth one: Linux crashed during the Newsexpire. (Heavy I/O)
solution was: a device with passive termination on a chain of
SCSI-2 as termiating device. Changed to a active terminator,
and all was fine.
For your Problem you should:
Software:
Check, if the emm386 or qemm386 of your DOS intallation includes
the RAM BIOS address Range of your SCSI Bios. It has to be excluded.
sometimes The harddrives keep wrking, but accessing the Tape or CD
crashes!
But Remember, in DOS it may run perfectly! Because DOS/Win/Win95
do only use asynchronous 5MHz Bus clock! And DOS is not able to
do such a access burst like linux. DOS has a low speed I/O system,
so DOS may run fine!
Linux should be fine if there are no more hardwareproblems.
Hardware:
- check termination
your controller is a passive on. If the bus does not run
with straight passive (all term Pwr off), you should
switch to active termination:
Do it so:
enable one device's Term Pwr. (is be done if you terminate
with a IBM DORS32160).
disable Controller termination, terminate with an external
Terminator plug, SCSI-II, active.
- check cableling: 15 cm cabel between 2 devices is a must!
Good Luck!
mfG
Jojo
-Schneckenpost-SlowMail-+--------Telefon-----+---eMails-FastMail----------
Jürgen Sauer Tel. +49 4209-3021 jojo@sae.extern.fh-hannver.de
Neue Str. 11 +49 172-4254411 jojo@linux.rz.fh-hannover.de
28790 Schwanewede/Germany FAX +49 4209 5217 CI$: 100013,1564
