From mboxrd@z Thu Jan 1 00:00:00 1970 From: Pravin Shelar Subject: Re: [PATCH net-next v4 1/7] vxlan: Add vxlan_handler. Date: Mon, 29 Jul 2013 13:17:42 -0700 Message-ID: References: <1374880488-2884-1-git-send-email-pshelar@nicira.com> <20130726165220.1106ac41@nehalam.linuxnetplumber.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Cc: netdev@vger.kernel.org To: Stephen Hemminger Return-path: Received: from na3sys009aog120.obsmtp.com ([74.125.149.140]:42617 "HELO na3sys009aog120.obsmtp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1754318Ab3G2URn (ORCPT ); Mon, 29 Jul 2013 16:17:43 -0400 Received: by mail-qc0-f171.google.com with SMTP id n1so3110747qcw.16 for ; Mon, 29 Jul 2013 13:17:42 -0700 (PDT) In-Reply-To: <20130726165220.1106ac41@nehalam.linuxnetplumber.net> Sender: netdev-owner@vger.kernel.org List-ID: On Fri, Jul 26, 2013 at 4:52 PM, Stephen Hemminger wrote: > On Fri, 26 Jul 2013 16:14:48 -0700 > Pravin B Shelar wrote: > >> vxlan and openvswitch modules needs to share port. Following >> patch breaks down vxlan per port abstraction in two component. >> So now vxlan_sock is broken into struct vxlan_sock handles socket >> and vxlan_handler which handler for user of vxlan protocol. >> >> Next commit will extend vxlan_handler for openvswitch. >> >> Signed-off-by: Pravin B Shelar > > Not sure why the added complexity here is needed. > > Since the are separate services, why not run them on separate > UDP ports. Otherwise unless the kernel and the Openvswitch > controller share state (which would be really hard given that > OVS controller is in user space), the chance of overlapping > configuration seems like a trap. > > There is already a lot of layering in VXLAN between the > device, forwarding table, VNI, and multiple UDP sockets. > Doing this needs more thought or persuasive use cases. I agree sharing state between kernel-vxlan and ovs-vxlan does not makes much sense. I will rework patches without vxlan-udp-port sharing which will eliminate that possibility entirely.