netdev - Re: [PATCH v2] ip6_tunnel: Correct tos value in collect

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20170615135432.78442af5@gmail.com>
Date:   Thu, 15 Jun 2017 13:54:32 +1000
From:   Peter Dawson <petedaws@...il.com>
To:     Haishuang Yan <yanhaishuang@...s.chinamobile.com>
Cc:     "David S. Miller" <davem@...emloft.net>,
        Alexey Kuznetsov <kuznet@....inr.ac.ru>,
        James Morris <jmorris@...ei.org>,
        Patrick McHardy <kaber@...sh.net>,
        Daniel Borkmann <daniel@...earbox.net>, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] ip6_tunnel: Correct tos value in collect_md mode

On Thu, 15 Jun 2017 10:30:29 +0800
Haishuang Yan <yanhaishuang@...s.chinamobile.com> wrote:

> Same as ip_gre, geneve and vxlan, use key->tos as tos value.
> 
> CC: Peter Dawson <petedaws@...il.com>
> Fixes: 0e9a709560db ("ip6_tunnel, ip6_gre: fix setting of DSCP on
> encapsulated packets”)
> Suggested-by: Daniel Borkmann <daniel@...earbox.net>
> Signed-off-by: Haishuang Yan <yanhaishuang@...s.chinamobile.com>
> 
> ---
> Changes since v2:
>   * Add fixes information
>   * mask key->tos with RT_TOS() suggested by Daniel
> ---
>  net/ipv6/ip6_tunnel.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
> index ef99d59..6400726 100644
> --- a/net/ipv6/ip6_tunnel.c
> +++ b/net/ipv6/ip6_tunnel.c
> @@ -1249,7 +1249,7 @@ int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
>  		fl6.flowi6_proto = IPPROTO_IPIP;
>  		fl6.daddr = key->u.ipv6.dst;
>  		fl6.flowlabel = key->label;
> -		dsfield = ip6_tclass(key->label);
> +		dsfield =  RT_TOS(key->tos);
>  	} else {
>  		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
>  			encap_limit = t->parms.encap_limit;
> @@ -1320,7 +1320,7 @@ int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
>  		fl6.flowi6_proto = IPPROTO_IPV6;
>  		fl6.daddr = key->u.ipv6.dst;
>  		fl6.flowlabel = key->label;
> -		dsfield = ip6_tclass(key->label);
> +		dsfield = RT_TOS(key->tos);
>  	} else {
>  		offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
>  		/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */

I don't think it is correct to apply RT_TOS

Here is my understanding based on the RFCs.

IPv4/6 Header:0 |0 1 2 3 |0 1 2 3 |0 1 2 3 |0 1 2 3 |
RFC2460(IPv6)   |Version | Traffic Class   |        |
RFC2474(IPv6)   |Version | DSCP        |ECN|        |
RFC2474(IPv4)   |Version |  IHL   |    DSCP     |ECN|
RFC1349(IPv4)   |Version |  IHL   | PREC |  TOS   |X|     
RFC791 (IPv4)   |Version |  IHL   |      TOS        |

u8 key->tos stores the full 8bits of Traffic class from an IPv6 header and;
u8 key->tos stores the full 8bits of TOS(RFC791) from an IPv4 header
u8 ip6_tclass will return the full 8bits of Traffic Class from an IPv6 flowlabel

RT_TOS will return the RFC1349 4bit TOS field.

Applying RT_TOS to a key->tos will result in lost information and the inclusion of 1 bit of ECN if the original field was a DSCP+ECN.

Based on this understanding of the RFCs (but not years of experience) and since RFC1349 has been obsoleted by RFC2474 I think the use of RT_TOS should be deprecated.

This being said, dsfield = ip6_tclass(key->label) = key->tos isn't fully correct either because the result will contain the ECN bits as well as the DSCP.

I agree that code should be consistent, but not where there is a potential issue.