Re: [PATCH] seg6: using DSCP of inner IPv4 packets

[Ahmed Abdelsalam]

Hi David,

SRv6 as defined in [1][2] does not mandate that the hop_limit of the 
outer IPv6 header has to be copied from the inner packet.

The only thing that is mandatory is that the hop_limit of the inner 
packet has to be decremented [3]. This complies with the specification 
defined in the Generic Packet Tunneling in IPv6 [4]. This part is 
actually missing in the kernel.

For the hop_limit of the outer IPv6 header, the other SRv6 
implementations [5][6] by default uses the default ipv6 hop_limit. But 
they allow also to use a configurable hop_limit for the outer header.

In conclusion the hop limit behavior in this patch is intentional and in 
my opnion correct.

If you agree I can send two patches to:
- decrement hop_limit of inner packet
- allow a configurable hop limit of outer IPv6 header


[1] https://tools.ietf.org/html/rfc8754
[2] 
https://tools.ietf.org/html/draft-ietf-spring-srv6-network-programming-16
[3] 
https://tools.ietf.org/html/draft-ietf-spring-srv6-network-programming-16#section-5
[4] https://tools.ietf.org/html/rfc2473#section-3.1
[5]https://github.com/FDio/vpp/blob/8bf80a3ddae7733925a757cb1710e25776eea01c/src/vnet/srv6/sr_policy_rewrite.c#L110
[6] 
https://www.cisco.com/c/en/us/td/docs/routers/asr9000/software/asr9k-r6-6/segment-routing/configuration/guide/b-segment-routing-cg-asr9000-66x/b-segment-routing-cg-asr9000-66x_chapter_011.html#id_94209


On 06/08/2020 02:40, David Miller wrote:
> From: Ahmed Abdelsalam <ahabdels@xxxxxxxxx>
> Date: Tue,  4 Aug 2020 07:40:30 +0000
>
> > This patch allows copying the DSCP from inner IPv4 header to the
> > outer IPv6 header, when doing SRv6 Encapsulation.
> >
> > This allows forwarding packet across the SRv6 fabric based on their
> > original traffic class.
> >
> > Signed-off-by: Ahmed Abdelsalam <ahabdels@xxxxxxxxx>
>
> You have changed the hop limit behavior here and that neither seems
> intentional nor correct.
>
> When encapsulating ipv6 inside of ipv6 the inner hop limit should be
> inherited.  You should only use the DST hop limit when encapsulating
> ipv4.
>
> And that's what the existing code did.