From mboxrd@z Thu Jan 1 00:00:00 1970 From: Robert Hancock Subject: Re: sata AHCI controller over non-PCI bus Date: Sun, 16 Aug 2009 10:55:54 -0600 Message-ID: <4A883A1A.9040009@gmail.com> References: <4A7FE245.8020002@ru.mvista.com> <4A88014D.9060309@garzik.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from an-out-0708.google.com ([209.85.132.250]:53326 "EHLO an-out-0708.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751967AbZHPRCx (ORCPT ); Sun, 16 Aug 2009 13:02:53 -0400 In-Reply-To: <4A88014D.9060309@garzik.org> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Jeff Garzik Cc: "Subbrathnam, Swaminathan" , Sergei Shtylyov , "linux-ide@vger.kernel.org" , LKML On 08/16/2009 06:53 AM, Jeff Garzik wrote: > Subbrathnam, Swaminathan wrote: >> Jeff, >> From the below link (from Sergei) it seems that you have already >> re-factored the AHCI implementation dependency on PCI. I would like to >> add support for the OMAPL138 SATA on top of your changes. That would >> be the ideal way forward for me. >> >> Have the ahci re-factoring changes queued for mainline merge already? >> I just joined the list and hence do not know the status. >> >> Sergei, >> Appreciate the response. > > I store the refactoring in git, on the "libahci" branch of > git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev.git > > Unfortunately, I am having second thoughts about an element of the > current design. With current Linux distributions, they do not appear to > deal well with the multi-module dependency libata -> libahci -> ahci. If > I had to guess, I would say that mkinitrd creation tools only look at > one tree level's worth of kernel module dependencies. > > Thus, ahci would wind up -not- in initrd, in a libahci solution. > > I am thinking that I will just add Marvell and ATP support to ahci.c, > and let someone else deal with libahci separation -- which is still needed. > > At this point, I would rather get Marvell/ATP support into users' hands, > rather than wait for distros to catch up to modern technology. If that's the case it seems clearly a mkinitrd bug. What distribution were you seeing this behavior on?