From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeff Garzik Subject: Re: Regression 3.2 -> 3.3-rc1 10 sec hang at boot and resume, COMRESET failed Date: Wed, 15 Feb 2012 13:29:29 -0500 Message-ID: <4F3BF989.8060607@garzik.org> References: <20120206004608.GA9413@gamma.logic.tuwien.ac.at> <1328492218.15079.38.camel@minggr> <20120206024052.GI9413@gamma.logic.tuwien.ac.at> <1328496591.15079.40.camel@minggr> <20120206031528.GA16093@gamma.logic.tuwien.ac.at> <1328503376.15079.55.camel@minggr> <20120206161943.GA3926@google.com> <20120213174445.GD12117@google.com> <20120215165706.GA24986@google.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from mail-qw0-f53.google.com ([209.85.216.53]:36830 "EHLO mail-qw0-f53.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752610Ab2BOS3d (ORCPT ); Wed, 15 Feb 2012 13:29:33 -0500 In-Reply-To: <20120215165706.GA24986@google.com> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Tejun Heo Cc: Brian Norris , Lin Ming , Norbert Preining , "Srivatsa S. Bhat" , linux-kernel@vger.kernel.org, "stern@rowland.harvard.edu >> Alan Stern" , linux-ide@vger.kernel.org, Linux PM mailing list , "Rafael J. Wysocki" , Jian Peng , Kevin Cernekee , valdis.kletnieks@vt.edu, Mark Lord On 02/15/2012 11:57 AM, Tejun Heo wrote: > Hello, > > On Tue, Feb 14, 2012 at 09:15:32PM -0800, Brian Norris wrote: >> Perhaps a module param (for ahci_platform) that sets a flag in >> ata_port_info? I'm not sure if/how I'm allowed to introduce new ATA >> flags... > > I think adding a module param directly to libahci.c should do. Just > add it after ignore_sss and apply it to all ahci's on the host. A module parameter is not necessarily the best/only option. ahci_platform already has infrastructure set up to deal with platform-specific quirks. An internal flag seems more appropriate to enable automatic handling of this on the specific platforms where it applies (plus the revert Tejun has already mentioned). Nothing wrong with debugging the regression further (Brian's option #3), but in the meantime we need to be actively using the best known working state, which means making the "fix" opt-in rather than unconditional or opt-out. Jeff