Re: [PATCH net-next v2 10/12] gtp: add IPv6 support

From: Harald Welte <laforge@netfilter.org>
To: Jonas Bonn <jonas@norrbonn.se>
Cc: netdev@vger.kernel.org, pablo@netfilter.org
Subject: Re: [PATCH net-next v2 10/12] gtp: add IPv6 support
Date: Sat, 12 Dec 2020 12:22:53 +0100	[thread overview]
Message-ID: <X9SoDToVmUdhgP0D@nataraja> (raw)
In-Reply-To: <20201211122612.869225-11-jonas@norrbonn.se>

Hi Jonas,

thanks again for your patches, they are very much appreciated.

However, I don't think that it is "that easy".

PDP contexts (at least) in GPRS/EDGE and UMTS come in three flavors:
* IPv4 only
* IPv6 only
* IPv4v6 (i.e. both an IPv4 and an IPv6 address within the same tunnel)

See for example osmo-ggsn at https://git.osmocom.org/osmo-ggsn
for an userspace implementation that covers all three cases,
as well as a related automatic test suite containing cases
for all three flavors at
https://git.osmocom.org/osmo-ttcn3-hacks/tree/ggsn_tests

If I read your patch correctly

On Fri, Dec 11, 2020 at 01:26:10PM +0100, Jonas Bonn wrote:
> -	struct in_addr		ms_addr_ip4;
> -	struct in_addr		peer_addr_ip4;
> +	struct in6_addr		ms_addr;
> +	struct in6_addr		peer_addr;

this simply replaces the (inner) IPv4 "MS addr" with an IPv6 "MS addr".

Sure, it is an improvement over v4-only.  But IMHO any follow-up
change to introduce v4v6 PDP contexts would require significant changes,
basically re-introducing the ms_add_ip4 member which you are removing here.

Therefore, I argue very much in favor of intrducing proper IPv6 support
(including v4v6) in one go.

Regards,
	Harald

-- 
- Harald Welte <laforge@gnumonks.org>           http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
                                                  (ETSI EN 300 175-7 Ch. A6)