From owner-freebsd-current Thu May 25 13:45:51 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id NAA10661 for current-outgoing; Thu, 25 May 1995 13:45:51 -0700 Received: from sbstark.cs.sunysb.edu (sbstark.cs.sunysb.edu [130.245.1.47]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id NAA10651 for ; Thu, 25 May 1995 13:45:47 -0700 Received: (from root@localhost) by sbstark.cs.sunysb.edu (8.6.12/8.6.9) with UUCP id QAA08024; Thu, 25 May 1995 16:45:01 -0400 Received: (from gene@localhost) by starkhome.cs.sunysb.edu (8.6.11/8.6.9) id OAA11883; Thu, 25 May 1995 14:18:34 -0400 Date: Thu, 25 May 1995 14:18:34 -0400 From: Gene Stark Message-Id: <199505251818.OAA11883@starkhome.cs.sunysb.edu> To: cs.weber.edu!terry@sbstark.cs.sunysb.edu CC: blaise.ibp.fr!roberto@sbstark.cs.sunysb.edu, phk@ref.tfs.com, FreeBSD.org!current@sbstark.cs.sunysb.edu In-reply-to: Terry Lambert's message of Thu, 25 May 95 11:32:51 MDT <9505251732.AA20705@cs.weber.edu> Subject: newfs weirdness... Sender: current-owner@FreeBSD.org Precedence: bulk >It is just as pessimal to precompensate for a "seek" on an IDE drive >based on a fictional (WRONG) geometry as it is to do so on a ZBR'ed SCSI >device Agreed. >The FS *really*, *really* wants to use the real geometry for the device >to let it do its optimization. Agreed. >To make both the BIOS and the UFS access happy, the geometry needs to >be fictitious for the BIOS and real for UFS. And the slice drivers >need to insure track boundries are observed. And because there is Agreed, though it is not always the case that the BIOS needs a fictitious geometry. The sticking point for me is that it is just as wrong to newfs a UFS filesystem with a fictitious BIOS geometry as it is to newfs it with a 4096/1 geometry. Using the fictitious BIOS geometry at least has the advantage that an integral number of cylinder groups fit into the partition, so you are not wasting a few thousand sectors at the end of the partition. Also, the "real" geometries supplied by the drives generally have a nice relationship to the fictitious BIOS geometries, so that again you get an integral number of cylinder groups in the partition. In the absence of information to the contrary, for an IDE drive it seems to me that the best available information is that which is supplied by the drive. Barring that, what is in the disklabel is next best. I can't possibly see why a 4096/1 geometry is going to be uniformly better than these other choices. Plus you have to go to extra trouble to avoid wasting sectors and getting the warning message. - Gene