From owner-cvs-sys Tue Jan 9 03:23:18 1996 Return-Path: owner-cvs-sys Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id DAA02290 for cvs-sys-outgoing; Tue, 9 Jan 1996 03:23:18 -0800 (PST) Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id DAA02280 Tue, 9 Jan 1996 03:22:52 -0800 (PST) Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id WAA17097; Tue, 9 Jan 1996 22:08:17 +1100 Date: Tue, 9 Jan 1996 22:08:17 +1100 From: Bruce Evans Message-Id: <199601091108.WAA17097@godzilla.zeta.org.au> To: graichen@omega.physik.fu-berlin.de, peter@freefall.freebsd.org Subject: Re: cvs commit: src/sys/i386/conf Makefile.i386 Cc: CVS-committers@freefall.freebsd.org, cvs-sys@freefall.freebsd.org Sender: owner-cvs-sys@FreeBSD.ORG Precedence: bulk >> Modified: sys/i386/conf Makefile.i386 >> Log: >> Also move /var/db/kvm_kernel.db to /var/db/kvm_kernel.old.db IF we change >> sysctl kern.bootfile - since libkvm expects it there. >> ... >just one related question: why is dev.db in /var/run and not in /var/db ? Because it is a temporary file associated with the current kernel. kvm_kernel.db was also in /var/run in FreeBSD-1.1, but the 4.4lite implementation of kvm_mkdb was so slow that preserving kvm_kernel.db across boots was worthwhile. kvm_mkdb was speeded up by a factor of about 5 (from 20+ seconds to 4+ seconds on my 486DX33 IDE system) on 1995/01/10 by using mmap. The speed can be improved by an additional factor or about 2.5 without changing the application by removing ufs's write-through for full data blocks. Write-through is very bad for random writes if the same block is often rewritten. Bruce