From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [net-next-2.6 PATCH 2/5] ixgbe: drop support for UDP in RSS hash generation Date: Tue, 20 Jul 2010 18:39:37 +0200 Message-ID: <1279643977.2498.89.camel@edumazet-laptop> References: <20100719235831.14112.14175.stgit@localhost.localdomain> <20100719235925.14112.65890.stgit@localhost.localdomain> <1279607980.2458.82.camel@edumazet-laptop> <4C45CAC6.1050006@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: "Kirsher, Jeffrey T" , "davem@davemloft.net" , "netdev@vger.kernel.org" , "gospo@redhat.com" , "bphilips@novell.com" , "Skidmore, Donald C" To: Alexander Duyck Return-path: Received: from mail-ww0-f42.google.com ([74.125.82.42]:43910 "EHLO mail-ww0-f42.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1761312Ab0GTQjl (ORCPT ); Tue, 20 Jul 2010 12:39:41 -0400 Received: by wwf26 with SMTP id 26so1751933wwf.1 for ; Tue, 20 Jul 2010 09:39:40 -0700 (PDT) In-Reply-To: <4C45CAC6.1050006@intel.com> Sender: netdev-owner@vger.kernel.org List-ID: Le mardi 20 juillet 2010 =C3=A0 09:11 -0700, Alexander Duyck a =C3=A9cr= it : > The packets will still be hashed on source and destination IPv4/IPv6=20 > addresses. The change just drops reading the UDP source/destination=20 > ports since in the case of fragmented packets they are not available = and=20 > as such were being parsed as IPv4/IPv6 packets. By making this chang= e=20 > the queue selection is consistent between all packets in the UDP stre= am. >=20 Excellent, this is perfect IMHO. > The only regression I would expect to see would be in testing between= =20 > two fixed systems since the IP addresses of the two systems would be=20 > fixed and so running multiple flows between the two would yield the s= ame=20 > RSS hash for multiple UDP streams. As long as multiple ip addresse= s=20 > are used you should see multiple RSS hashes generated and as such th= e=20 > load should still be distributed. >=20 Ack. Fortunately, one can still use RPS to spread load onto multiple cpus in this case. This until ixgpe fills skb->rxhash with a non null value. If it happens one day, we shall remind _not_ filling it for UDP packets= =2E BTW, this reminds me a netdev discussion we had for bnx2x http://www.kerneltrap.com/mailarchive/linux-netdev/2010/4/23/6275415/th= read And now, I understand why Toepliz hash doesnt use src port/dst port, since this is not available on fragments, obviously... Thanks !