[ovs-discuss] GTPv1-U in OVS

[Ahmed Talha Khan]

From: "Kyle Mestery (kmestery)" <kmestery at cisco.com>
To: Maryam Hanfi <maryam.hanfi at xflowresearch.com>
Cc: "<discuss at openvswitch.org>" <discuss at openvswitch.org>
Date: Fri, 14 Dec 2012 14:59:36 +0000
Subject: Re: [ovs-discuss] GTPv1-U in OVS
On Dec 13, 2012, at 11:35 PM, Maryam Hanfi <maryam.hanfi at xflowresearch.com>
wrote:
> >
> >Hi All,
> > I want to implement GTPv1-U in Open vSwitch. OVS expects inner MAC
header after tunnel header. But GTPv1-U expects IP header after GTPv1-U
header. Is there >
> >any way to modify the OVS tunneling so inner MAC not required anymore?
 Does tunneling work without inner MAC header?
> >
> >
>There is a thread on the dev list recently where we are discussing this
very thing. See here:
>
>http://openvswitch.org/pipermail/dev/2012-December/023711.html
>
>We are looking at implementing LISP which has the same  >requirement as
GTPv1-U. From what Jesse has indicated, this is a >large amount of work at
this point. We >would be happy to ?>collaborate with you on these changes
however. Unicast me if this >is something you'd like to pursue.

Hey Kyle,

I have a suggestion after looking at this patch. In this patch the
pre_tunnel work is being tied to the port_ops which makes sense but in most
of such cases the pre-tunnel work would depend on the flow itself rather
then being associated with the port. Which means that a better hook would
be, to allow an interface of this pre_tunneling code to the previous layer
of action execution/flow match.

e.g a use-case can be, that the pre_tunnel work depends on the nw_dst of
the packets.


What are you thoughts on this?
-- 
Regards,
-Ahmed Talha Khan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20121215/9dd3a4bb/attachment-0001.html>