From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262865AbTDVDkB (ORCPT ); Mon, 21 Apr 2003 23:40:01 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262872AbTDVDkB (ORCPT ); Mon, 21 Apr 2003 23:40:01 -0400 Received: from [12.47.58.203] ([12.47.58.203]:22947 "EHLO pao-ex01.pao.digeo.com") by vger.kernel.org with ESMTP id S262865AbTDVDkA (ORCPT ); Mon, 21 Apr 2003 23:40:00 -0400 Date: Mon, 21 Apr 2003 20:52:31 -0700 From: Andrew Morton To: "Martin J. Bligh" Cc: linux-kernel@vger.kernel.org Subject: Re: 67-mjb2 vs 68-mjb1 (sdet degredation) Message-Id: <20030421205231.4392aaa7.akpm@digeo.com> In-Reply-To: <147950000.1050981750@[10.10.2.4]> References: <1425480000.1050959528@flay> <20030421144631.4987db46.akpm@digeo.com> <147950000.1050981750@[10.10.2.4]> X-Mailer: Sylpheed version 0.8.11 (GTK+ 1.2.10; i586-pc-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 22 Apr 2003 03:51:57.0835 (UTC) FILETIME=[85B295B0:01C30882] Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org "Martin J. Bligh" wrote: > > > Here's a backout patch. Does it fix it up? > > Yeah, that fixes it. Ho hum ... I wonder if we can find something that > works well for both cases? I guess the options would be: > > 1. Some way to make the rwlock mechanism itself faster. > 2. Try to fix the contention itself somehow for this instance. You're using PIII's, which don't mind the additional buslocked operation. P4's hate it. We'd need to retest on big P4 and other architectures to decide. I suspect the spinlock is better that the rwlock overall - having multiple CPUs looking up things in the same address_space is relatively uncommon. Having one CPU looking things up in one address space is very common, and that's the thing which the s/rwlock/spinlock/ speeds up.