From owner-freebsd-current Tue Apr 4 19:58:04 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id TAA00574 for current-outgoing; Tue, 4 Apr 1995 19:58:04 -0700 Received: from obiwan.pmr.com (obiwan.pmr.com [199.98.84.130]) by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id TAA00568 for ; Tue, 4 Apr 1995 19:58:00 -0700 Received: by obiwan.pmr.com (Smail3.1.29.1 #4) id m0rwLHu-000300C; Tue, 4 Apr 95 21:57 CDT Message-Id: From: bob@obiwan.pmr.com (Bob Willcox) Subject: Re: Proper procedure to partition/label disk now? To: bde@zeta.org.au (Bruce Evans) Date: Tue, 4 Apr 1995 21:57:46 -0500 (CDT) Cc: bde@zeta.org.au, freebsd-current@freefall.cdrom.com In-Reply-To: <199504041445.AAA24580@godzilla.zeta.org.au> from "Bruce Evans" at Apr 5, 95 00:45:57 am X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 2641 Sender: current-owner@FreeBSD.org Precedence: bulk Bruce Evans wrote: > > >> For using the whole disk for FreeBSD, just run disklabel. The man > >> page may actually be complete now that the 'd' partition isn't > >> special. This assumes that the disk really is clean, without > >> misleading junk in the partition table. > > >Ok, I've done that. When I mount the filesystem (after having > >newfs'd it) I get the following messages: > > >Apr 4 08:14:27 luke /kernel: sd2: invalid primary partition table: no magic > >Apr 4 08:14:27 luke /kernel: sd2: raw partition size != slice size > >Apr 4 08:14:27 luke /kernel: sd2: start 0, end 782599, size 782600 > >Apr 4 08:14:27 luke /kernel: sd2c: start 0, end 782335, size 782336 > > These messages are only warnings. Should they be removed? The first one I fear that they are going to be a concern to users as they convert from 1.1.5 or 2.0 to 2.1. Lots of folks will think something is broken when they see these messages. > says that there is apparently no partition table. `disklabel -B' would > write a suitable partition table. The message should never be printed > for bootable disks. The last 3 say that you made the raw partition > size smaller than the disk size. This is a strange thing to do because > the raw partition is supposed to cover the whole disk. There is no > reason to round it to a cylinder boundary. I did the ``disklabel -B'' and got rid of the first message. The disklabel was created in the pre-2.1 style using the ficticious geometry popular during that era. What would really be nice would be a facility for labeling disks similar to what sysinstall provides. The manual calculations are a real pain. > > >Here is the disklabel for the drive: > > ># /dev/rsd2c: > >... > >sectors/unit: 782600 > >... > >4 partitions: > ># size offset fstype [fsize bsize bps/cpg] > > a: 782336 0 4.2BSD 1024 8192 16 # (Cyl. 0 - 381) > > c: 782336 0 unused 0 0 # (Cyl. 0 - 381) > > d: 782336 0 unused 0 0 # (Cyl. 0 - 381) > > The sectors/unit value isn't rounded. Good. Perhaps it should be checked > instead of the 'c' partition size. No, I just remembered why not. Old > labels have to be converted, and the sectors/unit value is guaranteed > to be wrong (being for the whole disk ond not for the BSD slice) except > when there is only one slice, so the 'c' partition size has to be trusted. > > The 'd' partition shouldn't be necessary. Yeah, this is an artifact of the disktab entry I used to create the label. -- Bob Willcox bob@obiwan.pmr.com (or obiwan%bob@uunet.uu.net) Austin, TX