From mboxrd@z Thu Jan 1 00:00:00 1970 From: Vlad Yasevich Subject: Re: [PATCH net 1/3] vlan: Fix tcp checksums offloads for Q-in-Q vlan. Date: Tue, 23 May 2017 08:59:45 -0400 Message-ID: <35dc2a59-7e6b-224d-49da-085d3f140fe7@redhat.com> References: <1495114265-23368-1-git-send-email-vyasevic@redhat.com> <1495114265-23368-2-git-send-email-vyasevic@redhat.com> <20170522.195921.1033609103219347751.davem@davemloft.net> Reply-To: vyasevic@redhat.com Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, makita.toshiaki@lab.ntt.co.jp To: David Miller , vyasevich@gmail.com Return-path: Received: from mx1.redhat.com ([209.132.183.28]:50890 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751308AbdEWM7x (ORCPT ); Tue, 23 May 2017 08:59:53 -0400 In-Reply-To: <20170522.195921.1033609103219347751.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-ID: On 05/22/2017 07:59 PM, David Miller wrote: > From: Vladislav Yasevich > Date: Thu, 18 May 2017 09:31:03 -0400 > >> It appears that since commit 8cb65d000, Q-in-Q vlans have been >> broken. The series that commit is part of enabled TSO and checksum >> offloading on Q-in-Q vlans. However, most HW we support can't handle >> it. To work around the issue, the above commit added a function that >> turns off offloads on Q-in-Q devices, but it left the checksum offload. >> That will cause issues with most older devices that supprort very basic >> checksum offload capabilities as well as some newer devices (we've >> reproduced te problem with both be2net and bnx). >> >> To solve this for everyone, turn off checksum offloading feature >> by default when sending Q-in-Q traffic. Devices that are proven to >> work can provided a corrected ndo_features_check implemetation. >> >> Fixes: 8cb65d000 ("net: Move check for multiple vlans to drivers") >> CC: Toshiaki Makita >> Signed-off-by: Vladislav Yasevich > > This is a tough one. I can certainly sympathize with your frustration > trying to track this down. > > Clearing NETIF_F_HW_CSUM completely is the most conservative change. > > However, for all the (perhaps many) cards upon which the checksumming > does work properly in Q-in-Q situations, this change could be > introducing non-trivial performance regressions. > > So I think Toshiaki's suggestion to drop IP_CSUM and IPV6_CSUM is, > on balance, the best way forward. > Thanks. I'll update and re-submit. -vlad > Thanks. >