From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH net-next] gro: relax ID check in inet_gro_receive() Date: Thu, 21 Mar 2013 17:39:56 -0400 (EDT) Message-ID: <20130321.173956.1308474437226033128.davem@davemloft.net> References: <20130321.170511.612698812448620739.davem@davemloft.net> Mime-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-2 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: maze@google.com, dkravkov@gmail.com, eric.dumazet@gmail.com, netdev@vger.kernel.org, dmitry@broadcom.com, eilong@broadcom.com, pshelar@nicira.com, hkchu@google.com To: jesse@nicira.com Return-path: Received: from shards.monkeyblade.net ([149.20.54.216]:47398 "EHLO shards.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751372Ab3CUVj7 convert rfc822-to-8bit (ORCPT ); Thu, 21 Mar 2013 17:39:59 -0400 In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: =46rom: Jesse Gross Date: Thu, 21 Mar 2013 14:37:38 -0700 > On Thu, Mar 21, 2013 at 2:05 PM, David Miller w= rote: >> From: Maciej =AFenczykowski >> Date: Thu, 21 Mar 2013 13:47:07 -0700 >> >>> They might create non incrementing IP ID packet sequences on the wi= re, >>> and this may prevent GRO from kicking in, but it does not make TSO >>> non-reversible >>> (if anything it makes it easy to detect that it happened). >> >> This is not an acceptable outcome for any form GSO, it should be >> completely transparent. The packets we would have output individual= ly >> and those which result from the segmented frame must be >> indistinguishable. >=20 > We actually don't set the outer IP ID for GRE packets with the DF bit > set in the non-GSO/TSO case either (and this is not new behavior). > Obviously TCP/IP header compression doesn't apply to the GRE packet > itself. >=20 > Therefore, I think if the driver is switched to increment the inner > ID, which it sounds like it can be, then everything should be > transparent. That's my impression as well.