From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bartlomiej Zolnierkiewicz Subject: Re: [GIT]: IDE Date: Mon, 9 Nov 2009 11:35:08 +0100 Message-ID: <200911091135.09051.bzolnier@gmail.com> References: <20091108.213055.54434626.davem@davemloft.net> Mime-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: Received: from mail-ew0-f207.google.com ([209.85.219.207]:41165 "EHLO mail-ew0-f207.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753024AbZKIKgN (ORCPT ); Mon, 9 Nov 2009 05:36:13 -0500 In-Reply-To: <20091108.213055.54434626.davem@davemloft.net> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: David Miller Cc: torvalds@linux-foundation.org, akpm@linux-foundation.org, linux-ide@vger.kernel.org, linux-kernel@vger.kernel.org On Monday 09 November 2009 06:30:55 David Miller wrote: > > The revert we thought we didn't need, we really do need still, even > with Bart Z.'s fix. > > The issue is that we are doing these PIO setting calls before we did > the device probe and therefore we don't yet know the media type. > > Yet the ->set_pio_mode() of some drivers key off of the media type to > determine whether to turn prefetching etc. on so it can't possibly > work properly. Thanks for quick reaction to the revert request. If somebody is interested in the proper way to fix it you just need to add ide_unknown media type and set it before probe instead of ide_disk one. Then audit ->set_pio_mode implementations for potential gotchas. [ However it is too late for -rc and nobody is currently working on IDE fixes anyway so I just requested the revert . ] -- Bartlomiej Zolnierkiewicz