From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tom Herbert Subject: Re: [net-next,1/2] net: Set sk_txhash from a random number Date: Tue, 8 Dec 2015 08:33:17 -0800 Message-ID: References: <1438124526-2129341-2-git-send-email-tom@herbertland.com> <566695D6.6060603@gmail.com> <1449580559.7632.10.camel@edumazet-glaptop2.roam.corp.google.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Cc: Alexander Drozdov , "David S. Miller" , Linux Kernel Network Developers , Kernel Team To: Eric Dumazet Return-path: Received: from mail-io0-f173.google.com ([209.85.223.173]:36534 "EHLO mail-io0-f173.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S965186AbbLHQdS (ORCPT ); Tue, 8 Dec 2015 11:33:18 -0500 Received: by iofh3 with SMTP id h3so29723984iof.3 for ; Tue, 08 Dec 2015 08:33:17 -0800 (PST) In-Reply-To: <1449580559.7632.10.camel@edumazet-glaptop2.roam.corp.google.com> Sender: netdev-owner@vger.kernel.org List-ID: On Tue, Dec 8, 2015 at 5:15 AM, Eric Dumazet wrote: > On Tue, 2015-12-08 at 11:33 +0300, Alexander Drozdov wrote: >> 29.07.2015 02:02, Tom Herbert wrote: >> > This patch creates sk_set_txhash and eliminates protocol specific >> > inet_set_txhash and ip6_set_txhash. sk_set_txhash simply sets a >> > random number instead of performing flow dissection. sk_set_txash >> > is also allowed to be called multiple times for the same socket, >> > we'll need this when redoing the hash for negative routing advice. >> It seems that this patch and some previous txhash-related >> ones break af_packet hash features for outgoing packets: >> - PACKET_FANOUT_HASH >> - TP_FT_REQ_FILL_RXHASH >> >> af_packet now thinks that hashes for for incoming and outgoing >> packets of the same TCP stream differ. That is true for TCP >> sessions initiated by the host. > > There never has been such guarantee. Even rx hashes for a single TCP > flow can differ, if packets are received on two different NIC with > different RSSS keys. > +1, it is a salient property that hashes can differ in each direction for a flow and that the hash for a flow can change over time. > >