From owner-freebsd-hackers Tue Apr 29 05:15:16 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.5/8.8.5) id FAA13081 for hackers-outgoing; Tue, 29 Apr 1997 05:15:16 -0700 (PDT) Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19]) by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id FAA13074 for ; Tue, 29 Apr 1997 05:15:11 -0700 (PDT) Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.5/8.6.9) id WAA02048; Tue, 29 Apr 1997 22:00:51 +1000 Date: Tue, 29 Apr 1997 22:00:51 +1000 From: Bruce Evans Message-Id: <199704291200.WAA02048@godzilla.zeta.org.au> To: freebsd-hackers@freebsd.org, syssgm@dtir.qld.gov.au Subject: Re: disklabel -- owner? Cc: davidg@root.com Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk >On a related topic, the disk device numbers are arrange in an ugly >way, presumably for backward compatibilty reasons. Are these reasons >still valid? Yes, there are a lot of /dev directories with "stable" device numbers out there. It's painful to boot with an old kernel and have it access the wrong partitions. >The current "type / unit2 / slice / major / unit1 / part" >could be rearranged to "major / unit / slice / part" (I don't know what >"type" is here. Grep didn't help me either.). "type" is supposed to be reserved for private use by drivers. AFAIK, it is only used by my version of the fd and ft drivers, which use the general definitions: #define FDUNIT(s) dkunit(s) #define FDTYPE(s) dktype(s) instead of special ones: #define FDUNIT(s) (((s)>>6)&03) #define FDTYPE(s) ((s)&077) This special cases for fd require lots of extracode in MAKEDEV and autoconf.c. >Oh, and the existing dkmakeminor() doesn't handle unit > 31, though >dkunit() does. I presume nobody has 32 or more disks. I haven't! Some people do. dkmakeminor() is little used. The bug certainly breaks devfs. In the sd driver, it seems to break sdsetunit(). Bruce