From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH] bnx2x: add support for receive hashing Date: Tue, 27 Apr 2010 16:30:59 +0200 Message-ID: <1272378659.2295.193.camel@edumazet-laptop> References: <1272302434.19143.76.camel@edumazet-laptop> <20100426.110432.104061817.davem@davemloft.net> <4BD6E887.3000804@athenacr.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: David Miller , therbert@google.com, netdev@vger.kernel.org, rick.jones2@hp.com To: Brian Bloniarz Return-path: Received: from mail-bw0-f219.google.com ([209.85.218.219]:51358 "EHLO mail-bw0-f219.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752497Ab0D0ObI (ORCPT ); Tue, 27 Apr 2010 10:31:08 -0400 Received: by bwz19 with SMTP id 19so131862bwz.21 for ; Tue, 27 Apr 2010 07:31:05 -0700 (PDT) In-Reply-To: <4BD6E887.3000804@athenacr.com> Sender: netdev-owner@vger.kernel.org List-ID: Le mardi 27 avril 2010 =C3=A0 09:37 -0400, Brian Bloniarz a =C3=A9crit = : > David Miller wrote: > > How damn hard is it to add two 16-bit ports to the hash regardless = of > > protocol? > > =20 > Come to think of it, for UDP the hash must ignore > the srcport and srcaddr, because a single bound > socket is going to wildcard both those fields. >=20 =46or your application maybe ;) Here, I have thousand of RTP flows to big mediagateways, so the (srcaddr, dstaddr) is shared by all these flows. Tom Herbert also wants a threaded DNS server. =46or UDP, we could have a bitmap (system level ?) to say if a particul= ar destination port wants multi-cpu (RPS) spreading or not, even if the NI= C decided to use a single queue to submit frames to the host (ie mask the src_addr and src_port). In this case, RFS on non conected UDP sockets could be activated as well. Or just a global sysctl to be able to mask the src_addr and/or src_port in our software rxhash. > So the best we can hope for is for the hash to > include destport and destaddr? From looking at > my BCM5709's multiq behavior, I think broadcom's > hash includes the destaddr but not the destport. Toepliz hash for UDP is a hash(src_addr, dst_addr) Both src port and dst port are ignored.