Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 12 Jul 1998 20:54:12 -0700 (PDT)
From:      David Wolfskill <dhw@whistle.com>
To:        questions@FreeBSD.ORG
Subject:   Re: SCSI "device wiring" query
Message-ID:  <199807130354.UAA06193@pau-amma.whistle.com>
In-Reply-To: <Pine.BSF.3.96.980711215758.3277l-100000@resnet.uoregon.edu>

next in thread | previous in thread | raw e-mail | index | archive | help
>Date: Sat, 11 Jul 1998 21:59:19 -0700 (PDT)
>From: Doug White <dwhite@resnet.uoregon.edu>

>On Thu, 9 Jul 1998, David Wolfskill wrote:
>> OK; I think I understand how to "wire" the devices on a given SCSI bus
>> in the kernel config (running 2.2.6-RELEASE here, for this purpose).

>> However, how is it determined which SCSI host adapter is (say) ahc0 vs.
>> ahc1?  Does it have something to do with the physical slot in which the
>> host adapter is placed on the system board?

>You attach adapters to scbusses, then asociate devices with target IDs on
>those busses.  See LINT for details.

Well, LINT doesn't discuss the issue I was trying to address; that's
probably a failure in how I tried to express myself -- at least, I don't
see anything in LINT that tells how to determine which physical card
will have a given logical name.

If I properly understood (and am relating) the conversation I had with
Julian, it seems that there really isn't a simple way to determine this
ahead of time; there are apparently too many variables.

I ended up configuring & building a new kernel (with hard-wired devices
assuming that the controllers were going to be swapped and that I'd be
moving all of the narrow devices to one controller, while leaving only
wide devices on the other).  I then physically swapped the locations of
the SCSI host adapters, and on boot, the adapters were recognized with
the names I had hoped.

In this case, I have a couple of ahc devices: a 2940, and a 2940UW.  The
2940 is strictly a narrow adapter, and all of the devices connected to
it were external to the machine; each is also narrow.  The 2940UW had a
couple of narrow internal devices connected to it, as well as a couple
of wide internal devices.

One of the narrow internal devices happens to be the boot disk (a
Quantum Empire 1080S).

Unfortunately, I found that merely ensuring that a given disk is SCSI
target 0 on the first-detected SCSI host adapter (ahc0, in this case)
apparently has *no* bearing on what device the machine will use to try
to boot:  despite the hardware-hackery I did (above), the machine
insisted on trying to boot from a drive on ahc1.  (I had also changed
the termination settings for the controllers:  for the 2940, from
"enabled" to "disabled"; fo rthe 2940UW, from "high on/low off" to
"automatic".)

This was not useful.

After several attempts to try to find some clue as to what might be thus
perverting the boot order, I eventually saw that the 2940's "BIOS" was
"disabled" (whatever that means), while the 2940UW's "BIOS" was
"enabled".  In desperation, I tried switching the two -- enabling the
2940, and disabling the 2940UW.  (Naturally, every time one looks at a
different adapter's settings or changes them, a re-boot is forced.)

This caused the machine to issue a dire-sounding warning about some
geometry being different from that which was expected, and that data
corruption might occur.  Of course, at this point, I had no way of
knowing anything better to do, so I told it to go ahead... and things
*seem* to be OK....

Now, of course, if this server goes down, someone with access to the
server room will need to come on-site, ensure there's a keyboard &
monitor hooked up, and manually tell it that re-booting really is OK.  :-(

I'm trying again to get a full set of backups (previously, the machine
would fall over before the backups were completed, the failure apparently
involving the SCSI errors & timeouts)... and so far, there's been
absolutely no whining at all about any SCSI problems... so other than the
forced manual intervention to complete a boot sequence, this seems to have
been OK ... so far:  the backups aren't done yet....

david
-- 
David Wolfskill		UNIX System Administrator
dhw@whistle.com		voice: (650) 577-7158	pager: (650) 371-4621

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199807130354.UAA06193>