If the fields you are adding are for a tunnel, and flow extraction will look beyond the tunnel header into the encapsulated packet, then I recommend adding your fields to struct flow_tnl--it will make the remaining changes easier. Otherwise, they probably should not be in flow_tnl.
On Wed, Nov 05, 2014 at 09:15:12AM +0000, Dani Camps wrote: > Hi Ben, > > Thanks. Looking at struct flow I see the structure flow_tnl. Should > matching fields of the GTP header be declared in here or directly as > members of the flow struct? > > Cheers > > Daniel > > On Tue, Nov 4, 2014 at 6:50 PM, Ben Pfaff <[email protected]> wrote: > > On Tue, Nov 04, 2014 at 07:42:36PM +0100, Dani Camps wrote: > >> I would appreciate if anyone can give me a high level orientation on > >> the best design approach to accomplish the above. For example should I > >> define the new header parameters in the ofp11_match structure and try > >> to replicate the code existing to match other parameters (e.g. in > >> lib/match.c), or is there an established framework in OVS to define > >> new match types? > > > > You should not modify ofp11_match. That would effectively change the > > OpenFlow 1.1 protocol. > > > > Start by adding a new member to struct flow in flow.h and a new field > > to the list in meta-flow.h. This will cause a large number of > > compiler warnings (I recommend configuring with --enable-Werror). The > > warnings point out all the places in the code that you need to update > > to add support. _______________________________________________ discuss mailing list [email protected] http://openvswitch.org/mailman/listinfo/discuss
