From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932191AbWBXOpl (ORCPT ); Fri, 24 Feb 2006 09:45:41 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932200AbWBXOpl (ORCPT ); Fri, 24 Feb 2006 09:45:41 -0500 Received: from kanga.kvack.org ([66.96.29.28]:28320 "EHLO kanga.kvack.org") by vger.kernel.org with ESMTP id S932191AbWBXOpk (ORCPT ); Fri, 24 Feb 2006 09:45:40 -0500 Date: Fri, 24 Feb 2006 09:40:28 -0500 From: Benjamin LaHaise To: Alan Stern Cc: Andrew Morton , sekharan@us.ibm.com, Kernel development list Subject: Re: [PATCH] Avoid calling down_read and down_write during startup Message-ID: <20060224144028.GB7101@kvack.org> References: <20060223161631.6f8fa41d.akpm@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.1i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Feb 23, 2006 at 10:18:18PM -0500, Alan Stern wrote: > Ben, earlier you expressed concern about the extra overhead due to > cache-line contention (on SMP) in the down_read() call added to > blocking_notifier_call_chain. I don't remember which notifier chain in > particular you were worried about; something to do with networking. > > Does this still bother you? I can see a couple of ways around it. Yes it's a problem. Any read lock is going to act as a memory barrier, and we need fewer of those in hot paths, not more to slow things down. -ben -- "Ladies and gentlemen, I'm sorry to interrupt, but the police are here and they've asked us to stop the party." Don't Email: .