From mboxrd@z Thu Jan 1 00:00:00 1970 From: Neil Horman Subject: Re: [PATCH] IPv6: DAD from bonding iface is treated as dup address from others Date: Fri, 7 Oct 2011 13:29:02 -0400 Message-ID: <20111007172902.GD11408@hmsreliant.think-freely.org> References: <1317873550-1677-1-git-send-email-Yinglin.Sun@emc.com> <20111006110047.GA22462@hmsreliant.think-freely.org> <27199.1317927933@death> <7122.1317949193@death> <20111007061326.GL23845@angus.ind.WPI.EDU> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: netdev@vger.kernel.org To: Yinglin Sun Return-path: Received: from charlotte.tuxdriver.com ([70.61.120.58]:50526 "EHLO smtp.tuxdriver.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751487Ab1JGR3K (ORCPT ); Fri, 7 Oct 2011 13:29:10 -0400 Content-Disposition: inline In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Fri, Oct 07, 2011 at 09:59:06AM -0700, Yinglin Sun wrote: > On Thu, Oct 6, 2011 at 11:13 PM, Chuck Anderson wrote: > > > > On Thu, Oct 06, 2011 at 06:24:36PM -0700, Yinglin Sun wrote: > > > On Thu, Oct 6, 2011 at 5:59 PM, Jay Vosburgh w= rote: > > > > =A0 =A0 =A0 =A0Why are you setting up the port channel after co= nfiguring the > > > > bond? > > > > > > > > =A0 =A0 =A0 =A0As a possible workaround, if you have control ov= er the setup > > > > process (perhaps it's some sort of manual process), adding one = slave to > > > > the bond, leaving the other soon-to-be slaves down, then settin= g up the > > > > switch, and finally adding the remaining slaves should work aro= und the > > > > issue, since if the bond has only one slave it won't see any lo= oped > > > > packets. > > > > > > > > =A0 =A0 =A0 =A0Or you could bring the bond up as active-backup,= then change the > > > > mode to balance-xor once the switch is configured. > > > > > > > > =A0 =A0 =A0 =A0Ultimately, though, the problem stems from the s= ettings mismatch > > > > between the switch and the bonding system; balance-xor is meant= to > > > > interoperate with etherchannel, and when the switch is not conf= igured > > > > properly, correct behavior is difficult to guarantee. > > > > > > > > > > Jay, > > > > > > Thanks a lot for the suggestion. > > > > > > It's mainly about usability. We would like to provide customers w= ith > > > consistent IPv6 configuration procedures as IPv4. =A0Such workaro= unds > > > could be confusing and generate customer calls. > > > > You've created/encouraged your customers to create a broken network > > configuration by connecting two bonded links to a non-bonded, > > non-etherchannel switch port pair. =A0This type of misconfiguration= , > > when applied to inter-switch trunks, can cause major network issues= , > > like looping and broadcast storms, taking down the entire network > > unless something like Spanning Tree is enabled to protect against s= uch > > accidental loops. =A0It should be avoided at all costs. =A0Luckily,= if the > > Linux host in this case is not being used as a switch/bridge, the > > impact of this might not be so bad--perhaps limited to the IPv6 DAD > > issue you report. > > > > If you want better usability and plug-n-play bonding, then require > > LACP/802.3ad to be used. =A0Please don't encourage your customers t= o > > connect misconfigured devices to the network, thanks. >=20 > You are right. LACP is the good choice. It should be able to solve > this issue, since all LACP bonding slaves are down before port channe= l > is set up on switch. Thanks. >=20 > I'm not sure this is kind of broken network configuration. If > customers happen to add some IPv6 addresses on bonding interface > before setting up port channel on switch, they have to reconfigure al= l Bringing up an interface prior to having it, and the peer interfaces co= nfigured to use an agreed upon mode, is rather by definition broken :)