From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bartlomiej Zolnierkiewicz Subject: Re: Re+: SATA hard drive no longer comes up as /dev/sda on 2.6.22 with Intel ICH6M Date: Thu, 26 Jul 2007 21:11:35 +0200 Message-ID: <200707262111.35691.bzolnier@gmail.com> References: <469E2F81.8080002@alcatel-lucent.com> <46A661C9.4010900@alcatel-lucent.com> <46A8393E.4040603@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: Received: from ug-out-1314.google.com ([66.249.92.170]:63978 "EHLO ug-out-1314.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933010AbXGZT7m (ORCPT ); Thu, 26 Jul 2007 15:59:42 -0400 Received: by ug-out-1314.google.com with SMTP id j3so591345ugf for ; Thu, 26 Jul 2007 12:59:41 -0700 (PDT) In-Reply-To: <46A8393E.4040603@gmail.com> Content-Disposition: inline Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Tejun Heo Cc: "P.C.Chan" , linux-ide@vger.kernel.org Hi, On Thursday 26 July 2007, Tejun Heo wrote: > Hello, > > P.C.Chan wrote: > > The last straw which breaks the SATA support is > > "CONFIG_IDE_GENERIC=y". Please see enclosed config files. > > > > The description being: > > > > *generic/default IDE chipset support (IDE_GENERIC) > > > > If unsure, say Y.* > > > > The linux distributions would have a hard time in preparing a kernel for > > general use, when CONFIG_IDE_GENERIC and libata are somewhat mutally > > exclusive for Intel chipset. > > Distros usually know better and don't build in the generic driver. It's > built as a module and used only as the last resort. > > > I guess it should at least say: > > > > *If unsure, say N.* > > > > It did no harm prior to 2.6.22 though. > > That didn't to any harm probably because ATA quirk code reserved the > area for libata during PCI initialization, so the generic driver > couldn't grab the port before libata. Now that the quirk is gone, the > generic driver grabs the first IDE channel before libata and you don't > have dma for your drive. > > cc'ing Bartlomiej. Bartlomiej, I think it would be better not to > default to Y for CONFIG_IDE_GENERIC. For most cases, it just causes > misconfiguration and confusion these days. Fully agreed, care to submit a patch? Thanks, Bart