From mboxrd@z Thu Jan 1 00:00:00 1970 From: roopa Subject: Re: [patch net-next] switchdev: call bridge setlink/dellink ndos recursively Date: Mon, 16 Mar 2015 10:07:02 -0700 Message-ID: <55070DB6.7020504@cumulusnetworks.com> References: <1426515774-21038-1-git-send-email-jiri@resnulli.us> <5506F4DF.30700@cumulusnetworks.com> <20150316163638.GI2058@nanopsycho.orion> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, davem@davemloft.net, sfeldma@gmail.com To: Jiri Pirko Return-path: Received: from mail-pa0-f53.google.com ([209.85.220.53]:36123 "EHLO mail-pa0-f53.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932499AbbCPRHE (ORCPT ); Mon, 16 Mar 2015 13:07:04 -0400 Received: by padcy3 with SMTP id cy3so70163344pad.3 for ; Mon, 16 Mar 2015 10:07:04 -0700 (PDT) In-Reply-To: <20150316163638.GI2058@nanopsycho.orion> Sender: netdev-owner@vger.kernel.org List-ID: On 3/16/15, 9:36 AM, Jiri Pirko wrote: > Mon, Mar 16, 2015 at 04:21:03PM CET, roopa@cumulusnetworks.com wrote: >> On 3/16/15, 7:22 AM, Jiri Pirko wrote: >>> There has been a discussion about if it's better to let masters to >>> propagate call down themself or if its better just blindly go down and >>> try to call ndo on every lower netdev. Turned out that more people (me >>> not included) like the second option better. >>> >>> This patch changes bridge setlink/dellink in that direction. >>> Sorry Roopa for forcing you to do it the way I liked initially. >>> >>> Signed-off-by: Jiri Pirko >> no worries. thanks for submitting the patch Jiri. >> >> One thing though (Which i also mentioned in one of the threads on this), >> the below command will not work with layered devices with the below patch. >> Because 'self' commands will directly try to find the switch port driver from >> rtnetlink.c and they dont use the switch dev api. >> >> bridge link set dev bond0 learning off self >> >> >> The code that currently exists in the tree with bond and team supporting the >> op >> will actually work. > Hmm, interesting. > > DaveM, this might be a good argument for call propagation. What do you > think? > For the stp api, it is not...because stp is run in the bridge driver and always involves the switchdev api. lets hold on to in-tree getlink/setlink before we find a better way. my 2c