From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: IPoIB: Fix multicast packet drops before join is complete Date: Thu, 11 Jun 2009 04:57:21 -0700 (PDT) Message-ID: <20090611.045721.05821546.davem@davemloft.net> References: <15ddcffd0906100318u2a61b7d8g991fe52439e07928@mail.gmail.com> <20090610.050122.261415364.davem@davemloft.net> <15ddcffd0906110445x25e2b2c5qa8445e0e1b344ef4@mail.gmail.com> Mime-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: rdreier@cisco.com, cl@linux-foundation.org, netdev@vger.kernel.org, yosefe@voltaire.com To: or.gerlitz@gmail.com Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:48058 "EHLO sunset.davemloft.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751893AbZFKL5T convert rfc822-to-8bit (ORCPT ); Thu, 11 Jun 2009 07:57:19 -0400 In-Reply-To: <15ddcffd0906110445x25e2b2c5qa8445e0e1b344ef4@mail.gmail.com> Sender: netdev-owner@vger.kernel.org List-ID: =46rom: Or Gerlitz Date: Thu, 11 Jun 2009 14:45:28 +0300 > On Wed, Jun 10, 2009 at 3:01 PM, David Miller wr= ote: >> From: Or Gerlitz >=20 >>> for --unicast-- the same problem [...] will come into play under bo= th Ethernet and IB, >>> so the user is limited by the socket buffer len and the neigh unres= _qlen systcl. >=20 >> But this generally does not cause problems via ARP, why might that >> be? Because resolution is 1) rare 2) quick. =A0Both of which do not >> apply to this IPoIB case. >=20 > To be precise, I don't see why ARP resolution is more rare then > joining to a multicast group on behalf of a sender in IPoIB, they > both happen on the first and if/when the kernel deletes the > neighbour. There are two issues: 1) Ethernet does not have to "resolve" anything for multicast sends. Therefore this issue is IPoIB specific. 2) ARP's tend to resolve on the DNS request for the site, or the initial connection attempt. Both of which are single packets, easily retransmitted, and the dropping of which does not result in losing significant application data. The multicast case over IPoIB does drop tons of actual application data.