From owner-freebsd-hackers Wed Jan 4 03:03:51 1995 Return-Path: hackers-owner Received: (from root@localhost) by freefall.cdrom.com (8.6.9/8.6.6) id DAA21546 for hackers-outgoing; Wed, 4 Jan 1995 03:03:51 -0800 Received: from hda.com (hda.com [199.232.40.182]) by freefall.cdrom.com (8.6.9/8.6.6) with ESMTP id DAA21537 for ; Wed, 4 Jan 1995 03:03:48 -0800 Received: (dufault@localhost) by hda.com (8.6.9/8.3) id GAA01683; Wed, 4 Jan 1995 06:04:33 -0500 From: Peter Dufault Message-Id: <199501041104.GAA01683@hda.com> Subject: Re: Bad sectors on SCSI drive! To: tinguely@plains.nodak.edu (Mark Tinguely) Date: Wed, 4 Jan 1995 06:04:33 -0500 (EST) Cc: hackers@freebsd.org, mtaylor@gateway.cybernet.com In-Reply-To: <199501040412.AA08177@plains.NoDak.edu> from "Mark Tinguely" at Jan 3, 95 10:12:08 pm X-Mailer: ELM [version 2.4 PL24] Content-Type: text Content-Length: 1644 Sender: hackers-owner@freebsd.org Precedence: bulk Mark Tinguely writes: > > one time I had a drive that somehow had an error below scsi driver > but not a physical error that reported as a medium error. I backup > what I could and low-level the drive and that cleaned up the error. > I would give it try before sending the drive back. IGNORE MY LAST MAIL - I was wrong and Jordan was wrong. See below. First, for Mark's mail: if it is not a medium error you should get a "non-media hardware failure" as the freefall owners will tell you. The reformat should work for real media failures because the defects should be on the grown defect list. But if the automatic reallocation is off it will come back again. Why I was Wrong: Look at section 8.3.3.6 of the SCSI-II spec. Automatic reallocation on read will only happen if the disk can fully recover the data. If the disk can't recover the data it isn't allowed to try to slip the sector (makes sense; maybe you can recover part of the data "manually"). Over the life of a drive it is reasonable that it will get an error on a block that has to be read. So we need some utility or something in the driver to handle that case. If you don't mind nuking the data (and there probably isn't much else you can do) I bet a write to that logical block will induce the drive to slip the block. "Just" do a read block / ignore and log errors / write block pass to the entire unmounted raw disk... Peter -- Peter Dufault Real Time Machine Control and Simulation HD Associates, Inc. Voice: 508 433 6936 dufault@hda.com Fax: 508 433 5267 -- Formerly hd@world.std.com. E-mail problems? Tell hdslip@iii.net