From owner-freebsd-questions Thu Dec 25 15:33:39 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id PAA07706 for questions-outgoing; Thu, 25 Dec 1997 15:33:39 -0800 (PST) (envelope-from owner-freebsd-questions) Received: from emu.sourcee.com (emu.sourcee.com [199.201.159.173]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id PAA07701 for ; Thu, 25 Dec 1997 15:33:33 -0800 (PST) (envelope-from nrice@emu.sourcee.com) Received: (from nrice@localhost) by emu.sourcee.com (8.8.8/8.8.3) id SAA08238; Thu, 25 Dec 1997 18:33:10 -0500 (EST) Message-ID: <19971225183309.07319@emu.sourcee.com> Date: Thu, 25 Dec 1997 18:33:09 -0500 From: Norman C Rice To: Julian Elischer Cc: freebsd-questions@freebsd.org Subject: Re: How to add slice to existing FS References: <19971225172526.55319@emu.sourcee.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Description: email message X-Mailer: Mutt 0.88 In-Reply-To: ; from Julian Elischer on Thu, Dec 25, 1997 at 02:50:36PM -0800 Sender: owner-freebsd-questions@freebsd.org X-Loop: FreeBSD.org Precedence: bulk On Thu, Dec 25, 1997 at 02:50:36PM -0800, Julian Elischer wrote: > The 'c' partition is treated specially in the kernel. it happens to work, > but it is defined to be the same as the whole slice on which the disklabel > is placed. it is sometimes synthesized. The correct name fo rthat > 'partition' would be the same name, with just the s1 (or maybe without > that if you don't even have slices..) > > Theoretically you should be able to make a filesystem on, > /dev/sd0 [whole disk is one filesystem] > /dev/sd0s1 [whole of first slice is one filesystem] > /dev/sd0s1a [a sub-partition of the first slice is a filesystem] > /dev/sd0a [a sub-partition of a dangerously dedicated disk is a > filesystem] > > The first two may not work right now but the same effect can be achieved > by defining the 'a' partition to be the whole area. using the 'c' > partition relies on something that 'while currently true' is not > necessarily logically correct. > > If you were a unix newby.. > and you wanted to makw a filesystem on the whole disk > which would be the device you would expect to use? > > sd0 > sd0s1c > sd0s1a > > the answer is: > "If you wanted to do the whole disk, why bother partitionning it at all?" > sd0 > "If you were going to partition it then why use 'c'? what's so > magic about 'c'?" The usual correct answer is to use 'a' > and keep 'c' hanging around for compatibilty. > If you actually use 'c' now, it's no real work to define an 'a' > that covers the same area. > > > So the long answer to your question is: > 'c' was defined to be special back when disklabels were first > introduced because once htey subdeviced a disk, they had no other > way of specifying the whole disk. This is not true in FreeBSD > so it is relying on the continuance of this 'compatibilty feature' > to expect 'c' to always do what you expect. If we ever 'recover' > teh 'c' partition in teh same way that we recoveredd the 'd' partition > (which used to be special too) then 'c' will become > just another partition. > In my new code, 'c' is not needed and since I don't allow overlapping > partitions to be shown in /dev, 'c' is not shown by default. > > I have just realised that 'c' is not overlapping for you, > so it probably will show up on your system... > hmmm > > let's just say that using 'c' as a partition instead of defining > a new one has always been 'ok' but not done by people who want to > avoid trouble in the future. (i.e. the paranoid). [snip] I will consider myself forewarned. I guess the habit of using the 'c' partition (from my Sun3/Sun4 days) needs to be broken. I hope that the use of 'c' partitions on existing systems won't end up biting me -- time will tell. If this does end up biting me, I will take your advice and "convert" to using the 'a' partition. Thanks for the explanation. -- Regards, Norman C. Rice, Jr.