From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hal Rosenstock Subject: Re: [PATCH 1/1] OpenSM: dfsssp - add support for multicast Date: Fri, 03 May 2013 13:45:25 -0400 Message-ID: <5183F7B5.3010809@dev.mellanox.co.il> References: <1367496926-6824-1-git-send-email-domke.j.aa@m.titech.ac.jp> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1367496926-6824-1-git-send-email-domke.j.aa@m.titech.ac.jp> Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Jens Domke Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Torsten Hoefler List-Id: linux-rdma@vger.kernel.org On 5/2/2013 8:15 AM, Jens Domke wrote: > Recent tests on a large system revealed a problem with loops in the multicast routing. > Using DFSSSP together with the default mcast routing algorithm of OpenSM can > produce loops in the fabric. > > This patch adds the mcast_build_stree function to the DFSSSP routing algorithm, > so that DFSSSP is able to calculate the correct mcast forwarding tables for the > subnet. > > It almost does the same steps as the default mcast routing, except that it > uses the Dijkstra algorithm to generate the spanning tree instead of using the > hop count information given by the unicast routing. > > General overview of the algorithm in pseudo-code: > 1) identify the ports, which are part of the multicast group > 2) find the 'best' switch (depending on the hop count) for the mcast group, > which can be used as a root of the spanning tree > 3) perform a dijkstra step with the root switch as starting point > to generate a spanning tree to all other switches in the subnet > 4) build the mcast forwarding tables for relevant switches: > 4.1) select a switch which has mcast member ports connected to it > 4.2) set the downstream ports for the mcast member ports in the mcft > 4.3) traverse towards the root of the spanning tree and set up-/downstream > ports on this path for all involved switches > 4.4) goto 4.1 until all switches have been processed > > The same mcast algorithm will be used for SSSP, because SSSP has the potential to > produce loops in the mcast forwarding table as well. > > Signed-off-by: Jens Domke Thanks. Applied. -- Hal -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html