From mboxrd@z Thu Jan  1 00:00:00 1970
From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Subject: Re: [PATCH net] af_packet: fix raw sockets over 6in4 tunnel
Date: Tue, 15 Jan 2019 10:49:49 +0100
Message-ID: <3ebfc32c-6c2e-626b-493f-27e326ddb5b9@6wind.com>
References: <20190111144357.9700-1-nicolas.dichtel@6wind.com>
 <CA+FuTSc3N1FdXe+88xBCAUjHT5DEhzeBhCXedqFFPO-8coV1CA@mail.gmail.com>
 <CAF=yD-KBLi4vrnBjS1JBzEvBgED7hWt_mv0chynWZDrQ0ZR=vw@mail.gmail.com>
 <7baad0fc-6352-0350-5acd-76b9d780e89e@6wind.com>
 <CAF=yD-+jfzn8d8OgU=49_cw5_-Sqpc+MhCDwOWD+XxwcY=24Ug@mail.gmail.com>
 <CAF=yD-JSwz0FvB2RAZ34_8T8CjPLAW=ut0RyTKWJGnGQMjOpUg@mail.gmail.com>
Reply-To: nicolas.dichtel@6wind.com
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit
Cc: David Miller <davem@davemloft.net>,
        Network Development <netdev@vger.kernel.org>
To: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mail-wm1-f67.google.com ([209.85.128.67]:38908 "EHLO
        mail-wm1-f67.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1727560AbfAOJty (ORCPT
        <rfc822;netdev@vger.kernel.org>); Tue, 15 Jan 2019 04:49:54 -0500
Received: by mail-wm1-f67.google.com with SMTP id m22so2483921wml.3
        for <netdev@vger.kernel.org>; Tue, 15 Jan 2019 01:49:53 -0800 (PST)
In-Reply-To: <CAF=yD-JSwz0FvB2RAZ34_8T8CjPLAW=ut0RyTKWJGnGQMjOpUg@mail.gmail.com>
Content-Language: fr
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

Le 14/01/2019 à 16:38, Willem de Bruijn a écrit :
> On Mon, Jan 14, 2019 at 10:15 AM Willem de Bruijn
> <willemdebruijn.kernel@gmail.com> wrote:
[snip]
>> It is wrong because for other devices l2 header length is not zero, so
>> this incorrectly sets skb->network_header to the start of the link
>> layer header on all those devices.
Ok, thank you for the details.

>>
>> A one line variant of the above would be
>>
>> -               if (len < reserve)
>> +               if (len < reserve + sizeof(struct ipv6hdr))
> 
> and exclude the majority of devices with fixed hard header len. Those
> require len to be >= reserve, so this workaround does not apply to
> them:
> 
> if (len < reserve + sizeof(struct ipv6hdr) &&
>     dev->min_header_len != dev->hard_header_len)
> 
And what about:
- if (len < reserve)
+ if (dev->min_header_len != dev->hard_header_len)
 	skb_reset_network_header(skb);
?