From owner-freebsd-hackers  Sat Apr 22 04:24:53 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id EAA20193
          for hackers-outgoing; Sat, 22 Apr 1995 04:24:53 -0700
Received: from linux4nn.iaf.nl (root@linux4nn.iaf.nl [193.67.144.34])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id EAA20119
          for <hackers@freefall.cdrom.com>; Sat, 22 Apr 1995 04:24:49 -0700
Received: from uni4nn.iaf.nl (root@uni4nn.iaf.nl [193.67.144.33]) by linux4nn.iaf.nl (8.6.9/8.6.9) with SMTP id NAA10321; Sat, 22 Apr 1995 13:37:20 +0200
Received: by uni4nn.iaf.nl with UUCP id AA19024
  (5.67b/IDA-1.5); Sat, 22 Apr 1995 13:25:18 +0100
Received: by iafnl.iaf.nl with UUCP id AA07043
  (5.65c/IDA-1.4.4); Sat, 22 Apr 1995 10:46:14 +0200
Received: (from wilko@localhost) by yedi.iaf.nl (8.6.8/8.6.6) id SAA01248; Fri, 21 Apr 1995 18:33:31 +0200
From: Wilko Bulte <wilko@yedi.iaf.nl>
Message-Id: <199504211633.SAA01248@yedi.iaf.nl>
Subject: Re: [DEVFS] your opinions sought!
To: dufault@hda.com (Peter Dufault)
Date: Fri, 21 Apr 1995 18:33:30 +1596657 (MET DST)
Cc: terry@cs.weber.edu, pritc003@maroon.tc.umn.edu, rgrimes@gndrsh.aac.dev.com,
        hackers@freefall.cdrom.com
In-Reply-To: <199504211356.JAA16639@hda.com> from "Peter Dufault" at Apr 21, 95 09:56:52 am
X-Mailer: ELM [version 2.4 PL22]
Content-Type: text
Content-Length: 1745      
Sender: hackers-owner@FreeBSD.org
Precedence: bulk

> > Yuck! Powering up/down for SCSI devices on the fly is generally a
> > sure way to hang the buses, screw the disks etc. It *should* work,
> > but my experience shows that it has something like a 20% chance to
> > bomb.
> 
> This is true, though with practice you only have to restore your
> disks once a year.  We still have a nasty bug someplace that results in
> unending disk timeouts and eventual system death.  Usually, though
> our disk drivers

I've also seen 'adapter frozen' or something similar on AH154x cards.
Is this what you mean?

> will redo a transfer that was in progress when a device powered up on
> the bus, assuming the disk drive properly detects the power up.

That's the way it should work yes. Don't count on all hostadapters
to keep their act together though ;-)

> A utility that suspends all activity on the SCSI bus will make powering
> up a device safe, though, and I've been thinking of doing it (suspend
> activity, beep when all activity is suspended, delay for 10 seconds, and
> permit activity to resume).

Should work fine. The raidsystems I use at work (==DEC) use a similar
approach to 'hotswap' drives.

While we are at it: the 'scsi' tool reputedly should allow a re-probe
of all devices. What is the exact syntax for this? The man page
is less than obvious to me (this is _still_ 1.1.5, I need stability 
more than features (...))

> Peter Dufault               Real Time Machine Control and Simulation

_     __________________________________________________________________________
 |   / o / /  _   Wilko Bulte             email: wilko@yedi.iaf.nl
 |/|/ / / /( (_)  Private FreeBSD site  - Arnhem - The Netherlands
--------------------------------------------------------------------------------