From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH net-next-2.6 3/3] bonding,ipv4,ipv6,vlan: Handle NETDEV_BONDING_FAILOVER like NETDEV_NOTIFY_PEERS Date: Thu, 21 Apr 2011 21:13:49 -0700 (PDT) Message-ID: <20110421.211349.193726731.davem@davemloft.net> References: <22334.1302913805@death> <1303153792.2857.32.camel@bwh-desktop> <26398.1303240321@death> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: bhutchings@solarflare.com, andy@greyhouse.net, kaber@trash.net, netdev@vger.kernel.org, brian.haley@hp.com To: fubar@us.ibm.com Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:53804 "EHLO sunset.davemloft.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750775Ab1DVEOX (ORCPT ); Fri, 22 Apr 2011 00:14:23 -0400 In-Reply-To: <26398.1303240321@death> Sender: netdev-owner@vger.kernel.org List-ID: From: Jay Vosburgh Date: Tue, 19 Apr 2011 12:12:01 -0700 > Ben Hutchings wrote: > >>Why would we activate a slave without link up? Perhaps if the previous >>active slave is removed? > > It's special sauce for Infiniband; I don't recall the details > except that the submitter said that without it the initial gratuitous > ARP could be lost. I didn't (and still don't) have IB hardware to test > this on. I vaguely remember this IB has too, but also forget the details. If someone would get to the bottom of this and add a nice big comment to the code, to prevent such difficulties in remembering exactly why we do this in the future, I would very much appreciate it.