From owner-freebsd-current Mon Mar 27 05:29:39 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id FAA09337 for current-outgoing; Mon, 27 Mar 1995 05:29:39 -0800 Received: from irbs.com ([199.182.75.129]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id FAA09330 for ; Mon, 27 Mar 1995 05:29:35 -0800 Received: (from jc@localhost) by irbs.com (8.6.11/8.6.6) id IAA06005; Mon, 27 Mar 1995 08:27:57 -0500 From: John Capo Message-Id: <199503271327.IAA06005@irbs.com> Subject: Re: Slice errors To: bde@zeta.org.au (Bruce Evans) Date: Mon, 27 Mar 1995 08:27:56 -0500 (EST) Cc: freebsd-current@freefall.cdrom.com (freebsd-current) In-Reply-To: <199503270525.PAA08376@godzilla.zeta.org.au> from "Bruce Evans" at Mar 27, 95 03:25:40 pm X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 2717 Sender: current-owner@FreeBSD.org Precedence: bulk Bruce Evans writes: > > >I have a -current system with 3 Conner CP30200 SCSI drives. I have > >been using a C/H/S geometry of XXX/8/256 since Rodney mentioned > >this as the optimum geometry for SCSI disks and ufs. Prior to that > >I used XXX/64/32 as reported by the Adaptec controller. > > >With the new slice scheme I get this error for all drives: > > > sd0: raw partition size != slice size > > sd0: start 0, end 416107, size 416108 > > sd0c: start 0, end 413695, size 413696 > > >The capacity reported for these drives is 416108 blocks. It appears > >to be complaining that the slice is 416108 blocks but partition c > >is only 413696 blocks. > > 416108 - 413696 = 2412 = 2048(one cylinder) + 364. It used to be hard > avoid the size being rounded to a cylinder boundary (editing the label > with `disklabel -e' always rounded it). The off by one error for the > number of cylinders may have been from believing the BIOS geometry > instead of the geometry that was printed by the the driver. The BIOS > sometimes subtracts one from the number of of cylinders. > > >Ok, I'll make the c partition == 416108, > >who cares. Disklabel says, partition c extends past end of unit > >and does not write the label. There are no DOS partitions on these > >Conner drives, 100% FreeBSD. > > I can't explain this. Is the fdisk output that you quoted (with 50000 > sectors) for this drive. This was with the 50000 block disklabel. I tried to label it with in unmounted in multi-user and in single user. The old label was still there. I have seen the warning before that a partition extended past the end of the unit but it always wrote the label anyway. > > >All of my partitions end on cylinder boundaries due to bad experiences > >in the past. > > I don't know of any problems with partitions not ending on cylinder > boundaries except in the user interface. I had serious problems with not ending on cylinder boundaries at one point. I have not been willing do try it any other way since. I don't mind wasting a bit of disk space. > > >Do I have to use the bogus geometry reorted by the > >drive and rework all of my labels for this message to go away? > > No. > > >I have another system with a 1Gig SCSI disk that has a DOS partiton > >and a FreeBSD partition. The physical disk is larger than my c > >partition but I do not get any error messages. The capacity of this > >disk is 2031705 blocks. > > > sd0s1: start 32, end = 409599, size 409568: OK > > sd0s2: start 409600, end = 2031615, size 1622016: OK > > The c partition should have the same size as the slice (smaller than > the physical disk of you have more than one (nonempty) slice). > > Bruce > -- John Capo