From mboxrd@z Thu Jan  1 00:00:00 1970
From: Eric Dumazet <eric.dumazet@gmail.com>
Subject: Re: [PATCH net-next] net: llc: remove init_net check
Date: Thu, 30 Nov 2017 12:12:46 -0800
Message-ID: <1512072766.19682.27.camel@gmail.com>
References: <20171130194526.8076-1-aring@mojatatu.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8bit
Cc: hadi@mojatatu.com, netdev@vger.kernel.org
To: Alexander Aring <aring@mojatatu.com>, davem@davemloft.net
Return-path: <netdev-owner@vger.kernel.org>
Received: from mail-pf0-f179.google.com ([209.85.192.179]:46387 "EHLO
        mail-pf0-f179.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1750773AbdK3UMs (ORCPT
        <rfc822;netdev@vger.kernel.org>); Thu, 30 Nov 2017 15:12:48 -0500
Received: by mail-pf0-f179.google.com with SMTP id c204so3580853pfc.13
        for <netdev@vger.kernel.org>; Thu, 30 Nov 2017 12:12:48 -0800 (PST)
In-Reply-To: <20171130194526.8076-1-aring@mojatatu.com>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On Thu, 2017-11-30 at 14:45 -0500, Alexander Aring wrote:
> Commit e730c15519d09 ("[NET]: Make packet reception network namespace
> safe") added several init_net compares to protocols with unknown
> net namespace support. The idea was as users started using the
> different
> protocols they would test and enable them.
> In our use-case we generate STP enabled Linux bridges in different
> namespaces. The bridges are connected via veth interfaces.
> This patch makes STP work in such a setup.
> 
> Signed-off-by: Alexander Aring <aring@mojatatu.com>
> ---
>  net/llc/llc_input.c | 4 ----
>  1 file changed, 4 deletions(-)
> 
> diff --git a/net/llc/llc_input.c b/net/llc/llc_input.c
> index 82cb93f66b9b..09b530d90610 100644
> --- a/net/llc/llc_input.c
> +++ b/net/llc/llc_input.c
> @@ -14,7 +14,6 @@
>  #include <linux/netdevice.h>
>  #include <linux/slab.h>
>  #include <linux/export.h>
> -#include <net/net_namespace.h>
>  #include <net/llc.h>
>  #include <net/llc_pdu.h>
>  #include <net/llc_sap.h>
> @@ -162,9 +161,6 @@ int llc_rcv(struct sk_buff *skb, struct
> net_device *dev,
>  	void (*sta_handler)(struct sk_buff *skb);
>  	void (*sap_handler)(struct llc_sap *sap, struct sk_buff
> *skb);
>  
> -	if (!net_eq(dev_net(dev), &init_net))
> -		goto drop;
> -
>  	/*
>  	 * When the interface is in promisc. mode, drop all the crap
> that it
>  	 * receives, do not try to analyse it.


Well, we use different netns for isolation.

You need more changes than simply removing this check, I guess.

__llc_sap_find() would need a per netns list, or proper netns checks.