Date: Tue, 24 Jan 1995 21:09:36 +0000 (GMT) From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> To: phk@ref.tfs.com (Poul-Henning Kamp) Cc: terry@cs.weber.edu, bde@zeta.org.au, freebsd-hackers@freefall.cdrom.com, kuku@gilberto.physik.rwth-aachen.de Subject: Re: disklabel (1.1.5.1), partitions Message-ID: <199501242109.VAA10844@gndrsh.aac.dev.com> In-Reply-To: <199501241823.KAA21387@ref.tfs.com> from "Poul-Henning Kamp" at Jan 24, 95 10:22:58 am
next in thread | previous in thread | raw e-mail | index | archive | help
> > > > > Not all SCSI devices are thus translated? > > > mostly they are, unless the "> 1GB" option is set. > > > > Sorry, this isn't really a valid reason for lying about the geometry > Hey terry, you lost the reason: A scsi-disk >HAS< no geometry we should > care about. The keywords there are ``should care about''. SCSI disk only report 3 items, total number of sectors, physical number of heads and physical number of cylinders. The sectors/track is what we ``fake''. > > > > Hold on! How about because the kernel should know the real BIOS apparent > > > > geometry for instead of making things up like that? > > > true, except we cannot trust this one anymore :-( Some IDE and SCSI > > > drivers get their "geometry" in CONFIG.SYS these days. > > > > Actually, most EIDE controllers get them from a modified system boot block > > installed by the OEM for the machine. Blowing the system boot block in > > that case really screws you up. > No, 540Mb and above seems to get it from various "disk-managers" loaded > in the config.sys with an increasingly annoying frequency. There are many ways this problem is solved: 1) NEW BIOS'es that are LBA aware can handle drives into the GBytes 2) Special boot code that munges with the drive paramter table 3) Disk managers hidden on track 0 4) Device drivers loaded by config.sys > > I don't understand... what's wrong with: > > > > Geometry CCCC/HH/SS (Physical CCCC/HH/SS) > The Physical isn't true, It's two years since I saw anything which added > (multiplied actually) up to the number of sectors available. > The numbers are fiction, make belive, has no use, value or connection > with our use of the drive. I think it is time to change the sd probe message to ONLY output the total sectors, since that is the relevant number. And/Or clearly label the geometry as fictitious. > > > Has nothing to do with it. If you pass some of the weird geometries to > > > newfs it will become very confused... 32/64 works sensibly most of the > > > time. > > > > This is a bug in newfs, like the one where it doesn't create a lost+found, > > not a reason to bogify other parts of the system to hack around it. > Not true. Newfs and UFS was optimized to know about disk-geometry. Since > most (95%) drives these days are zoned (variable sectors/track), this isn't > of any particular use. The existense of caches on the drives doesn't improve > it either. > All the geometry is really used for is to size the "cylinder"-groups. Probably time to rewrite newfs and parts of ufs to deal with reality :-). -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Custom computers for FreeBSD
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199501242109.VAA10844>