From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH 1/2 v2] net: Add function to get SW rxhash Date: Tue, 24 Sep 2013 17:17:37 -0700 Message-ID: <1380068257.3165.130.camel@edumazet-glaptop> References: <1380058203.3165.121.camel@edumazet-glaptop> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: David Miller , Linux Netdev List To: Tom Herbert Return-path: Received: from mail-pd0-f176.google.com ([209.85.192.176]:53740 "EHLO mail-pd0-f176.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755288Ab3IYARi (ORCPT ); Tue, 24 Sep 2013 20:17:38 -0400 Received: by mail-pd0-f176.google.com with SMTP id q10so5295351pdj.35 for ; Tue, 24 Sep 2013 17:17:38 -0700 (PDT) In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Tue, 2013-09-24 at 15:16 -0700, Tom Herbert wrote: > If it's a big win for conntrack to process packets on same CPUs for > both directions, we should program the RFS table to accomplish that. conntrack uses no userland code, no user thread, so what are you going to put in RFS table ? Right now, RPS can be used for proper steering, you only have to "ethtool -K eth{x} rxhash off" [ Of course, this is becoming less a concern with multiqueue nics ]