From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [RFC] New Qlogic 10Gb Ethernet driver for 2.6.28 Date: Mon, 25 Aug 2008 14:05:22 -0700 (PDT) Message-ID: <20080825.140522.192264181.davem@davemloft.net> References: <20080822174626.GC2551@susedev.qlogic.org> <20080822.144343.231567307.davem@davemloft.net> <20080825170148.GA31059@susedev.qlogic.org> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: jeff@garzik.org, Linux-Driver@qlogic.com, netdev@vger.kernel.org To: ron.mercer@qlogic.com Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:33428 "EHLO sunset.davemloft.net" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1751526AbYHYVF0 (ORCPT ); Mon, 25 Aug 2008 17:05:26 -0400 In-Reply-To: <20080825170148.GA31059@susedev.qlogic.org> Sender: netdev-owner@vger.kernel.org List-ID: From: Ron Mercer Date: Mon, 25 Aug 2008 10:01:48 -0700 > I see your point. I will add the re-alignment but would like to jump > over it for platforms that don't need it. Is there a list somewhere of > the RISC platforms that take a trap on unaligned accesses? Or better > yet, a list of those that don't? If you're only doing this for the header portion it's not so much of a cost. For sizable packets you're using page references, and you don't have to copy those, such copies of the headers will also warm up the CPU cache as a nice side effect. Furthermore you stated that the alignment restriction only is necessary for early revisions of the hardware. Please, I do not want to give you a list of platform names to ifdef up your driver with, that's messy. Just make the copy when necessary to align things properly.