Presence, RLS, and NAT Traversal - sr-users

25 Mar 2011


      Hello,
I am using Kamailio and nathelper to connect UAs behind a NAT.  However,
I want to use presence and RLS in this system as well.  The problem I
have is that the presence and RLS modules are not nathelper  aware and
the NOTIFY requests that they generate are targeted at the private
network IP address and port of the UA.
Has anyone any experience of this?  Can anyone help?
I had a cunning plan where, in route[NAT], I detected SUBSCRIBE requests
that originate from behind a NAT (and do not have myself as the last
hop).  When I see a request like this I use add_contact_alias(),
record_route() and forward it to myself.  I then added a
handle_ruri_alias() to the WITH_NAT section of route[RELAY].  The theory
being that an in-dialog NOTIFY from presence or RLS will use the
route-set established by the SUBSCRIBE transaction.
This almost works.  The presence.winfo NOTIFY from the presence module
and the initial presence NOTIFY from the RLS module make it back to the
client successfully.  However, when this forwarding of the SUBSCRIBE is
in place the RLS subscriptions don't happen.  I know it is the forward
of the SUBSCRIBE that is the trigger behind this problem because simply
commenting that out (and allowing the first SUBSCRIBE to get to RLS)
makes the RLS subscriptions happen (but in this case the NOTIFYs get
back to the client and are 404'd because the request URI is not
correct).
The only difference between the first SUBSCRIBE and the forwarded
SUBSCRIBE is an additional "Via:" header and a "Record-Route:" header.
I ran Kamailio with the -E -ddd options and observed that:
* modules_k/rls/subscribe.c:rls_get_service_list() is outputting
        the line "searching document for user sip:@"
      * This function is called by
        modules_k/rls/subscribe.c:rls_handle_subscribe() and the output
        suggests that GET_FROM_PURI(msg)->user and
        GET_FROM_PURI(msg)->host are empty strings
      * However, parse_from_header(msg) is called near the top of
        rls_handle_subscribe() and clearly succeeds to find the "From:"
        header in the message
      * As I mentioned above the forwarded SUBSCRIBE is almost identical
        to the first SUBSCRIBE, and so it does contain a valid "From:"
        header
All I can think of is that Kamailio is becoming confused by the two
near-identical SUBSCRIBE requests and that the "cleaning up" of the
first (statelessly forwarded) SUBSCRIBE is also "cleaning up" the
forwarded SUBSCRIBE that I am currently working with?
My route[NAT]:
route[NAT] {
                xlog("L_INFO", "$rm: route[NAT]\n");
        #!ifdef WITH_NAT
                force_rport();
                if (nat_uac_test("19") && $si!="MY_REAL_IP" && $si!
        ="127.0.0.1") {
                        xlog("L_INFO", "  Request from NAT'd UAC\n");
                        if (is_method("REGISTER")) {
                                fix_nated_register();
                        } else if (is_method("SUBSCRIBE")) {
                                add_contact_alias();
                               route(RECORD_ROUTE);
                               forward(127.0.0.1, MY_REAL_PORT);
                               exit;
                        } else {
                                fix_nated_contact();
                        }
                        setflag(FLT_NATS);
                }
        #!endif
                return;
        }
My route[RECORD_ROUTE]:
route[RECORD_ROUTE]
        {
                # remove preloaded route headers
                remove_hf("Route");
                if (is_method("INVITE|SUBSCRIBE"))
                        record_route();
return;
        }
The start of my route[RELAY]:
route[RELAY] {
                xlog("L_INFO", "$rm: route[RELAY]\n");
        #!ifdef WITH_NAT
                if (check_route_param("nat=yes")) {
                        setbflag(FLB_NATB);
                }
                if (isflagset(FLT_NATS) || isbflagset(FLB_NATB)) {
                        route(RTPPROXY);
                }
handle_ruri_alias();
        #!endif
Regards,
Peter
-- 
Peter Dunkley
Technical Director
Crocodile RCS Ltd