Subject: Re: Danger: Hardware Recommendation Request
To: John S. Dyson <toor@dyson.iquest.net>
From: Chris G. Demetriou <cgd@cs.cmu.edu>
List: port-i386
Date: 04/16/1997 13:19:00
> > 
> > It seem to vary with the drive type, as well as the revision of
> > the driver. I've generally had good luck with Quantum drives, and
> > poor luck with HP drives. Often just pulling in a newer verison of
> > the driver will help, though.
> > 
> I have had problems with HP drives and Buslogic also.

So, since we're adding personal experience:

I have an OEM'd hp3325.

under the ncr driver, it simply would not work.  Not one single I/O
would ever actually succeed.  I seem to recall hearing speculation
that it was because of some tagged-queueing-handling bogosity in the
ncr driver (combined with weird, but not necessarily disallowed,
behaviour on the part of the drive).

On a BT-542B (several years old, but with a recent firmware upgrade so
that i'd work properly with my Jaz drive), it works great.  (It worked
fine before the firmware upgrade, but the majority of the testing --
use -- has been afterwards.


I've also seen the ncr driver fall over badly with '825 chips with no
disks on the (properly-terminated) SCSI bus.  it would spew error
after error after error (for pages) during autoconfiguration, but
eventually complete and not disturb the system further.



My general advice: If you can get an aic7xxx-based board and thereby
avoid the 'ncr' driver, definitedly do so.  The later revs of the
'ahc' driver seem to be quite stable.



cgd