More can't get this $@%! SCSI to work

[Carl Lowenstein]

> Date: Sun, 07 Jul 2002 16:53:35 -0700
> From: "Paul G. Allen" <pgallen at randomlogic.com>
> Organization: Random Logic
> To: kplug-list at kernel-panic.org
> Subject: Re: More can't get this $@%! SCSI to work
>
Paul Allen writes:

> I had some older and a newer drive (coincidentally, all in the same HDD
> chassis I spoke of earlier). I knew all the old drives worked, and I
> knew the chassis and cable worked. It would not all work together
> though. Since I got everything from Western Scientific, was short on
> time, cables, controllers, and terminators (making it hard to test
> different combinations of stuff), I had Western look into it for me.
>
> It turned out to be a combination of factors, all of which by themselves
> caused different problems, and all of which combined caused even more
> problems.

Prob. 1 & 2 snipped.

>
> Problem 3:
>
> The new drive was U/W differential and the old drives were Wide single
> ended. The controller was setting itself to work at the U/W speed, and
> not dropping back to the slower speeds for the older drives. In
> addition, the required termination for the two drive types was
> different.

I am sure you mean LV differential here (low voltage) because that is
the only thing that can co-exist with single-ended drives.  Of course
the way that this happens is that all the LV diffential units must
drop back to the older and slower SE standard.

> The solutions here were several. First, all devices were set to the
> slowest common speed in the SCSI BIOS. Second, the new drive was set to
> Single Ended (indicated on IBM drives by a jumper labeled SE or
> similar). Third, the Host Adapter termination was enabled (DO NOT use
> the Automatic setting) for Single Ended termination. (NOTE: If you are
> using the internal and external connectors on a single channel card, the
> the host adapter termination must be DISABLED as it is in the middle of
> the bus, and the termination belongs on either end.) Fourth, a new
> active terminator was put on the end of the cable (actually, it's
> plugged into the back of the HDD chassis). Fifth, one and only one
> device is providing termination power to the bus (currently, I do not
> know which device that is without looking at jumper settings and the
> SCSI BIOS - sorry Lan, I'm just not going to tear things apart to figure
> it out :).

Conventionally it is the host adapter that provides termination power.

> Today I have 5 different systems with U/W SCSI in them. I have cables
> and terminators and can f' with SCSI issues 'til my heart is content (or
> explodes as the case may be ;). Fortunately, I learned from the previous
> nightmare and generally get it right the first time, aside from the
> occasional SCSI ID conflict when I don't pay close attention to which
> jumper is put where (I hate that because it always amounts to removing
> the drives to check the jumper settings. What a PITA!! :)

One could put stick-on labels on internal SCSI drives to show what ID
they are set to.  I don't think I have ever done that, but it seems like
a good idea.  Of course when you change the jumpers you have to change
the label too.  Sort of like changing the comments to fit the code.

Note in passing -- I think that Lan's system is all narrow SCSI so
the whole set of problems relating to LV differential vs UltraWide
go away.

    carl