From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Graf Subject: Re: [ovs-dev] OVS + BPF, make sense? Date: Mon, 22 Dec 2014 10:53:21 +0100 Message-ID: <20141222095321.GA21348@pox.localdomain> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: "dev@openvswitch.com" , "netdev@vger.kernel.org" To: Andy Zhou Return-path: Received: from mail-wg0-f53.google.com ([74.125.82.53]:35527 "EHLO mail-wg0-f53.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754289AbaLVJxZ (ORCPT ); Mon, 22 Dec 2014 04:53:25 -0500 Received: by mail-wg0-f53.google.com with SMTP id l18so6250745wgh.12 for ; Mon, 22 Dec 2014 01:53:24 -0800 (PST) Content-Disposition: inline In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: Thanks a lot for sharing these minutes. On 12/19/14 at 06:49pm, Andy Zhou wrote: > Possible use cases of BPF in OVS Linux kernel datapath > =========================================== > > [...] > > 4. Using BPF to implement overall OVS kernel module functionality > > Alexei likes this approach the most. The potential benefits are: This would be my favourite as well long term assuming that the performance benefits we hope for can be proven. A logical evolutionary process might be 2, 3 and then go for the full coverage. A small but: We can't just remove the existing Netlink based action data path as non-OVS users exist which rely on it. So this would need to exist in parallel unless we can get all users on board to transition over to this new architecture.