From owner-freebsd-current Wed Mar 15 16:09:04 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id QAA02769 for current-outgoing; Wed, 15 Mar 1995 16:09:04 -0800 Received: from cs.weber.edu (cs.weber.edu [137.190.16.16]) by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id QAA02763 for ; Wed, 15 Mar 1995 16:09:02 -0800 Received: by cs.weber.edu (4.1/SMI-4.1.1) id AA15691; Wed, 15 Mar 95 17:01:25 MST From: terry@cs.weber.edu (Terry Lambert) Message-Id: <9503160001.AA15691@cs.weber.edu> Subject: Re: newfs: sectors per cylinder (4096) disagrees with disk label (36) To: phk@ref.tfs.com (Poul-Henning Kamp) Date: Wed, 15 Mar 95 17:01:24 MST Cc: rgrimes@gndrsh.aac.dev.com, current@FreeBSD.org In-Reply-To: <199503152033.MAA08072@ref.tfs.com> from "Poul-Henning Kamp" at Mar 15, 95 12:33:21 pm X-Mailer: ELM [version 2.4dev PL52] Sender: current-owner@FreeBSD.org Precedence: bulk > > > > If no one objects (more specifically if Poul-Henning does not object) I > > > > am going to removed the newfs hacks for 1 track of 4096 cylinders. And > > > > fix src/etc/Makefile to not have the -t 0 -u 0. > > > > > > I do object. > > > > On what grounds??? [ ... ] > I would like to see something like the following rules added to newfs, > (by me or somebody else with better time) [ ... ] > Or something similar, suggestions welcome. The hacking is to not introduce rotdelay workaround when a head seek is expected when in fact head seeks occur at other location because the geometry is translated, thereby effectively doubling the condition that the optimization is supposed to help prevent, yes? I propose that part of the disk information available via ioctl is the translated geometry information. The SCSI device driver will lie about this information. Meanwhile, newfs (which should be mkfs, but that's another argument) needs to ask the disk its preferred format via ioctl. Then all this disktab/newfs options/etc. crap can go out the window into the garbage can where it belongs. Drivers which are not expected to have translated geometries (and which, it is pointed out, won't work anyway in -current because of a lack of bad144 support) will report their real geometry that they think they are using. Because of the disklabel crap, there really needs to be two sets of data returned, the geomtery that the drive reports when askes and the DOS apparent geometry. Kill the disklabel and this goes away too (but requires BIOS calls to set up the information in the first place, and so probably means using the information now passed in (?) from the boot blocks. Really, there needs to be an abstraction between the device and the geometry management, not only so that the bad144 slips in transparently for old drives, but so the partitioning is seperable from the block access -- a requirement for porting the crap to a box that doesn't use DOS disk layout. Terry Lambert terry@cs.weber.edu --- Any opinions in this posting are my own and not those of my present or previous employers.