From mboxrd@z Thu Jan 1 00:00:00 1970 From: Nicolas Dichtel Subject: Re: [PATCH v2 net] net: Always untag vlan-tagged traffic on input. Date: Mon, 29 Sep 2014 15:58:15 +0200 Message-ID: <54296577.6050401@6wind.com> References: <1407523333-31455-1-git-send-email-vyasevic@redhat.com> <54295449.7030203@6wind.com> <20140929130543.GB1839@nanopsycho.orion> <542962CF.6000602@6wind.com> <542964B2.3090303@redhat.com> Reply-To: nicolas.dichtel@6wind.com Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: netdev@vger.kernel.org, Patrick McHardy , Nithin Nayak Sujir , Michael Chan To: vyasevic@redhat.com, Jiri Pirko Return-path: Received: from mail-we0-f171.google.com ([74.125.82.171]:46846 "EHLO mail-we0-f171.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751921AbaI2N6U (ORCPT ); Mon, 29 Sep 2014 09:58:20 -0400 Received: by mail-we0-f171.google.com with SMTP id w61so327796wes.2 for ; Mon, 29 Sep 2014 06:58:19 -0700 (PDT) In-Reply-To: <542964B2.3090303@redhat.com> Sender: netdev-owner@vger.kernel.org List-ID: Le 29/09/2014 15:54, Vlad Yasevich a =E9crit : > On 09/29/2014 09:46 AM, Nicolas Dichtel wrote: >> Le 29/09/2014 15:05, Jiri Pirko a =E9crit : >>> Mon, Sep 29, 2014 at 02:44:57PM CEST, nicolas.dichtel@6wind.com wro= te: >>>> Le 08/08/2014 20:42, Vladislav Yasevich a =E9crit : >>>> [snip] >>>>> >>>>> The patch attempt to fix this another way. It moves the vlan hea= der >>>>> stipping code out of the vlan module and always builds it into th= e >>>>> kernel network core. This way, even if vlan is not supported on >>>>> a virtualizatoin host, the virtual machines running on top of suc= h >>>>> host will still work with VLANs enabled. >>>> After this patch (0d5501c1c828 ("net: Always untag vlan-tagged tra= ffic on >>>> input.")), tcpdump will not be able to display the VLAN header on = input path. >>>> Is this really intended? >>> >>> I assume you are talking about the case when CONFIG_VLAN_8021Q is "= n" >>> right? Because if it is "m" of "y" I see no change in the code. >> Right. >> >>> >>> In case of CONFIG_VLAN_8021Q is "n" the only change is that for >>> non-accelerated path, skb->vlan_tci is set and vlan header is strip= ped. >>> This makes the path same to accelerated path. >>> >>> tcpdump should be able to cope with skb->vlan_tci >> My fault, I've tested with a quite old tcpdump version, with a recen= t one it >> works perfectly! > > Thanks. BTW, did that old tcpdump version work correctly when CONF= IG_VLAN_8021Q > was y or m? My guess is that it would not. Your guess is right ;-) Nicolas