From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from gate.crashing.org (gate.crashing.org [63.228.1.57]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTP id B4F05DDF10 for ; Fri, 12 Jan 2007 08:55:17 +1100 (EST) Subject: Re: [PATCH 2/7] Powerpc MSI implementation From: Benjamin Herrenschmidt To: Greg KH In-Reply-To: <1168550403.22458.414.camel@localhost.localdomain> References: <1168514716.63474.857278133999.qpush@cradle> <20070111112503.0CC1BDDF13@ozlabs.org> <20070111194427.GA20450@kroah.com> <1168550403.22458.414.camel@localhost.localdomain> Content-Type: text/plain Date: Fri, 12 Jan 2007 08:54:57 +1100 Message-Id: <1168552498.22458.428.camel@localhost.localdomain> Mime-Version: 1.0 Cc: linuxppc-dev@ozlabs.org, Paul Mackerras , Olof Johannsson , linux-pci@atrey.karlin.mff.cuni.cz List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Fri, 2007-01-12 at 08:20 +1100, Benjamin Herrenschmidt wrote: > > So, no, I don't agree with this implementation and don't want to see it > > go through anyone's tree into mainline just yet. > > So why, when we posted it earlier, you said the exact opposite ? That > you actually liked it and wanted to see it as a replacement of the > current cruft ? > > That's been the logic from day one. We implement a generic MSI support > that supports multiple backends, we do it as a powerpc version at first > since that's what we can actively test and provide backends for, then as > a second step, we work with the intel folks to port their stuff over to > our implementation and replace the current crappy generic one. /me calms down a bit ... :-) Ok, so what about we do: - Move that new implementation to drivers/pci - Have a CONFIG_NEW_MSI or something like that to select between the old and the new implementation - Have the change to pci_dev be based on that config option Note: There should be no impact on drivers nor anything out of drivers/pci and associated arch support code, the API exposed to drivers should be identical. That way, it will be easier for Intel/Altix folks to port over to the new code as it will be accessible to all archs via a config option. Once Intel & Altix have ported over, the old code can be removed and CONFIG_NEW_MSI too. I know that other archs have actually been waiting for that new infrastructure (well, according to some feedback we got when Michael first posted it) so that's probably a better approach than having it in arch/powerpc indeed. Cheers, Ben.