From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751589AbXAHPcM (ORCPT ); Mon, 8 Jan 2007 10:32:12 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751590AbXAHPcM (ORCPT ); Mon, 8 Jan 2007 10:32:12 -0500 Received: from outpipe-village-512-1.bc.nu ([81.2.110.250]:57198 "EHLO lxorguk.ukuu.org.uk" rhost-flags-OK-FAIL-OK-FAIL) by vger.kernel.org with ESMTP id S1751588AbXAHPcL (ORCPT ); Mon, 8 Jan 2007 10:32:11 -0500 Date: Mon, 8 Jan 2007 15:42:49 +0000 From: Alan To: Jeff Garzik Cc: akpm@osdl.org, torvalds@osdl.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH] sata_via: PATA support, resubmit Message-ID: <20070108154249.6d8f5697@localhost.localdomain> In-Reply-To: <45A24159.7060001@pobox.com> References: <20070108122659.00c22754@localhost.localdomain> <45A24159.7060001@pobox.com> X-Mailer: Sylpheed-Claws 2.6.0 (GTK+ 2.10.4; x86_64-redhat-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org > The problem you need to fix or work around is ata_probe_ent, which > doesn't properly fill in ata_port info for this situation. Tejun has > posted patches that kill ata_probe_ent, which you were pointed to. And which are not yet in the main tree leaving many users unable to install Linux. This isn't the way to get stuff done. When you've got the new patches in the driver can use them if its worth it (which, see below, I question). > If you get the setup right, you don't bloat each hook with "is this port > PATA?" tests. At present, your sata_via patch introduces these needless > tests. Which I might note are actually smaller than the extra structs and on a code path usually executed twice per boot. So the needless tests are more efficient and not on a hot path and are shorter than the elegant not-present solution and let users actually install Linux on current VIA systems, which right now the generally cannot do. Out of boredom I'll also note that clock timings say that if the extra port info stuff causes a single extra L1 cache miss its also faster to do the tests. Alan