From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH] retain ABI definitions for obsolete multi-queue packet schedulers Date: Fri, 12 Sep 2008 16:01:54 -0700 (PDT) Message-ID: <20080912.160154.32092334.davem@davemloft.net> References: <20080912122548.48945e11@extreme> <20080912.153033.157394228.davem@davemloft.net> <20080912155830.5ff50c59@extreme> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: kaber@trash.net, netdev@vger.kernel.org To: shemminger@vyatta.com Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:48869 "EHLO sunset.davemloft.net" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1755254AbYILXCA (ORCPT ); Fri, 12 Sep 2008 19:02:00 -0400 In-Reply-To: <20080912155830.5ff50c59@extreme> Sender: netdev-owner@vger.kernel.org List-ID: From: Stephen Hemminger Date: Fri, 12 Sep 2008 15:58:30 -0700 > On Fri, 12 Sep 2008 15:30:33 -0700 (PDT) > David Miller wrote: > > > From: Stephen Hemminger > > Date: Fri, 12 Sep 2008 12:25:48 -0700 > > > > > On Fri, 12 Sep 2008 10:00:11 -0700 > > > Stephen Hemminger wrote: > > > > > > > Even though the sch_rr qdisc is now gone in 2.6.27, the kernel > > > > definitions for the attributes need to be maintained because applications > > > > like iproute need to be compatible with older kernels and use santized > > > > kernel headers. > > ... > > > This should be applied for 2.6.27 since it is a build regression. > > > > You use a local copy of this header file in iproute2, you don't even > > use the kernel header itself. It's the worst possible example and > > excuse for this change. > > > > Nobody references this header directly for these definitions. > > The one and only use is using a local copy. > > > > iproute2 is the only thing even remotely referencing this stuff. > > > > So there is no breakage. The only breakage is if you, Stephen, decide > > to copy the kernel header into iproute2 as-is, and that is your > > choice. :-) > > Every release I copy the sanitized kernel headers resulting > from 'make headers_install'. That is how they have been maintained and > up until now it worked. The sch_rr configuration portion was referencing > this. The point of the sanitized headers was to avoid this nonsense. But here is the flip side, we actually don't want new applications referencing those things, since no kernel from now and into the future will ever support those knobs again. They will always fail. Therefore, it makes sense to put the definitions into the one existing tool that used it. You can even put those defines directly into the sch_rr iproute2 code, and then your sanitized header bits will continue to work transparently.