From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750900AbWGKOid (ORCPT ); Tue, 11 Jul 2006 10:38:33 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750885AbWGKOic (ORCPT ); Tue, 11 Jul 2006 10:38:32 -0400 Received: from gw.exalead.com ([193.47.80.25]:58531 "EHLO exalead.com") by vger.kernel.org with ESMTP id S1750759AbWGKOic (ORCPT ); Tue, 11 Jul 2006 10:38:32 -0400 Message-ID: <44B3B7E6.7020902@exalead.com> Date: Tue, 11 Jul 2006 16:38:30 +0200 From: Xavier Roche Organization: Exalead User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; fr-FR; rv:1.8.0.2) Gecko/20060404 SeaMonkey/1.0.1 Mnenhy/0.7.3.0 MIME-Version: 1.0 To: Erik Mouw CC: linux-kernel@vger.kernel.org Subject: Re: Huge performance issue with cciss driver on HP DL385 servers (2.6.13 -> 2.6.17) References: <44B3A178.2060908@exalead.com> <20060711132231.GG9790@harddisk-recovery.com> In-Reply-To: <20060711132231.GG9790@harddisk-recovery.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Erik Mouw a écrit : > AFAIK buffers for direct IO need to be *page* aligned. Use something > like: > I guess you got away with it cause your 512 byte alignment happened to > align on a page, but you shouldn't count on that. However... This example was actually just a minimal example - the running test is correctly aligned. > You should check the return value of pwrite(). Also checked in the "real" stress program. Note that this problem is hard to reproduce, and might be related to very strange I/O (and/or DMA access) operation ordering. The fact that the program must be run just after a reboot (or the problem is not easily reproductible) is something really fishy.