From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?ISO-8859-1?Q?Nicolas_de_Peslo=FCan?= Subject: Re: [PATCH] bonding: reset queue mapping prior to transmission to physical device Date: Thu, 02 Jun 2011 22:13:57 +0200 Message-ID: <4DE7EF05.8030505@gmail.com> References: <20110602185659.GA2749@hmsreliant.think-freely.org> <1307041789.2812.27.camel@bwh-desktop> <20110602194621.GB2749@hmsreliant.think-freely.org> <20110602.130448.917724318321380583.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: nhorman@tuxdriver.com, bhutchings@solarflare.com, netdev@vger.kernel.org, fubar@us.ibm.com, andy@greyhouse.net To: David Miller Return-path: Received: from mail-wy0-f174.google.com ([74.125.82.174]:33973 "EHLO mail-wy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753354Ab1FBUOB (ORCPT ); Thu, 2 Jun 2011 16:14:01 -0400 Received: by wya21 with SMTP id 21so914943wya.19 for ; Thu, 02 Jun 2011 13:14:00 -0700 (PDT) In-Reply-To: <20110602.130448.917724318321380583.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-ID: Le 02/06/2011 22:04, David Miller a =E9crit : > From: Neil Horman > Date: Thu, 2 Jun 2011 15:46:21 -0400 > >> Potentially, yes. I only fixed this because I was looking at bondin= g and its >> queue_mapping behavior, and saw that this needed fixing. Bridging a= nd IP >> forwarding should also likely clear the queue mapping in the forward= ing path >> somewhere to avoid selecting an output tx queue that is a function o= f whatever >> queue and device it arrived on during ingress. I've not yet looked = to see if >> thats already being done. > > No we do not do this, intentionally. > > That way the input classification and queue selection is mirrored > on the transmit side, which we absolutely want to happen. Can you confirm that this is the expected behavior for IP forwarding an= d bridge but not for bonding? To be more precise, due to the way bonding use queue mapping for slave = selection, it is desirable to=20 clear the mapping before sending to the slave, because the meaning of t= he mapping for the slave=20 interface might be really different from the meaning for the bonding in= terface. Arguably, this is=20 the mapping usage in bonding which is "different" from other usages, bu= t... Nicolas.