From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: small RPS cache for fragments? Date: Tue, 17 May 2011 17:10:00 -0400 (EDT) Message-ID: <20110517.171000.1166144155994185790.davem@davemloft.net> References: <20110517.164929.1737248436066795381.davem@davemloft.net> <1305666050.2691.4.camel@edumazet-laptop> Mime-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: therbert@google.com, netdev@vger.kernel.org To: eric.dumazet@gmail.com Return-path: Received: from shards.monkeyblade.net ([198.137.202.13]:51236 "EHLO shards.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932296Ab1EQVKE convert rfc822-to-8bit (ORCPT ); Tue, 17 May 2011 17:10:04 -0400 In-Reply-To: <1305666050.2691.4.camel@edumazet-laptop> Sender: netdev-owner@vger.kernel.org List-ID: =46rom: Eric Dumazet Date: Tue, 17 May 2011 23:00:50 +0200 > Le mardi 17 mai 2011 =E0 16:49 -0400, David Miller a =E9crit : >> From: Tom Herbert >> Date: Tue, 17 May 2011 13:02:25 -0700 >>=20 >> > I like it! And this sounds like the sort of algorithm that NICs m= ight >> > be able to implement to solve the UDP/RSS unpleasantness, so even >> > better. >>=20 >> Actually, I think it won't work. Even Linux emits fragments last to >> first, so we won't see the UDP header until the last packet where it= 's >> no longer useful. >>=20 >> Back to the drawing board. :-/ >=20 > Well, we could just use the iph->id in the rxhash computation for fra= gs. >=20 > At least all frags of a given datagram should be reassembled on same > cpu, so we get RPS (but not RFS) That's true, but one could also argue that in the existing code at leas= t one of the packets (the one with the UDP header) would make it to the proper flow cpu. That could be as much as half of the packets. So I don't yet see it as a clear win.