From mboxrd@z Thu Jan 1 00:00:00 1970 From: John Fastabend Subject: [RFC PATCH v1 0/3] Expose switching attributes via PF_BRIDGE Date: Tue, 29 May 2012 20:06:54 -0700 Message-ID: <20120530030531.7443.72024.stgit@jf-dev1-dcblab> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Cc: sri@us.ibm.com, gregory.v.rose@intel.com, netdev@vger.kernel.org, bhutchings@solarflare.com, jeffrey.t.kirsher@intel.com, eric.w.multanen@intel.com To: krkumar2@in.ibm.com, hadi@cyberus.ca, shemminger@vyatta.com, mst@redhat.com, buytenh@wantstofly.org, eilong@broadcom.com Return-path: Received: from mga03.intel.com ([143.182.124.21]:29217 "EHLO mga03.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932261Ab2E3DVH (ORCPT ); Tue, 29 May 2012 23:21:07 -0400 Sender: netdev-owner@vger.kernel.org List-ID: This series decouples the remaining netlink PF_BRIDGE messages from the bridging module and moves them into rtnetlink proper. By doing this we can use these netlink messages to handle any type of bridge and extend the attributes as needed. I hope this resolves some of the concerns with the DSA patch below and expect the attached series can be extended to support the DSA infrastructure as needed: http://patchwork.ozlabs.org/patch/16578/ Also this should resolve a patch here that tried to expose the switching modes but did so using a device specific hook: http://lists.openwall.net/netdev/2012/04/16/10 I've used a hacked version of the 'bridge' tool Stephen Hemminger submitted as an RFC some months back to test this the output looks like this: [root@jf-dev1-dcblab iproute2]# ./br/br bridge show eth2: bridge mode: VEB embedded eth3: bridge mode: VEB embedded [root@jf-dev1-dcblab iproute2]# ./br/br bridge mode dev eth2 mode vepa [root@jf-dev1-dcblab iproute2]# ./br/br bridge show eth2: bridge mode: VEPA embedded eth3: bridge mode: VEB embedded I could have just added a ndo op and IFLA_XXX message to set the switching mode but IMHO this is not going to scale as more bridging functionality becomes offloaded. The DSA example is a case where we already have a fully offloaded switch. Any solution we come up with should support both embedded switches and SW switches. Any comments would be appreciated Thanks! --- John Fastabend (3): ixgbe: add setlink, getlink support to ixgbe and ixgbevf net: add VEPA, VEB bridge mode net: create generic bridge ops drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 100 +++++++++++++++ drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c | 3 drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c | 10 ++ include/linux/if_link.h | 16 ++ include/linux/netdevice.h | 10 ++ net/bridge/br_device.c | 2 net/bridge/br_netlink.c | 75 ++--------- net/bridge/br_private.h | 3 net/core/rtnetlink.c | 137 +++++++++++++++++++++ 9 files changed, 291 insertions(+), 65 deletions(-) -- Signature