Hello Benny, As I don't have much time to be checking the PJSIP code to make a patch, I've used another approach: -- Instead of having two transports, disconnect UDP and then connect the TURN one, I've created an "special" TURN transport that contains a UDP transport inside. This is like "inheriting" the UDP transport, in some way. -- It has an additional function, and by calling it I open the TURN and this special transport redirects the SIP traffic through the TURN server. I think this would be a "safer" way, as it is not usual for PJSIP applications to be registering/unregistering transports during the normal execution of the program. Regards, Guilherme Balena Versiani. On Wed, Jun 6, 2012 at 11:48 PM, Benny Prijono <bennylp at teluu.com> wrote: > On Thu, Jun 7, 2012 at 2:38 AM, Guilherme Balena Versiani <guibv at nymgo.com > > wrote: > >> Hello, >> >> >> My idea is basically pass the SIP traffic through a TURN server to >> overcome some firewall limitations. >> >> I'm using PJSUA, registering an UDP transport >> using pjsua_transport_create as usual, and, when I discover a timeout 408 >> during the first REGISTER, my idea was to close the previously created UDP >> transport and register a new one. I've implemented a sip_transport_turn.c >> based on the sip_transport_udp.c, and during the initialization it sends an >> allocate request to TURN to listen to SIP packets. In the on_rx_data >> callback, I am checking if this is the first time I am sending packets to >> this destination, and, if so, I'm using the pj_turn_sock_set_perm to add >> permission to send data. >> >> Everything is great in theory, but I am facing a lot of problems in PJSUA >> code (I'm using PJSIP 2.0): >> >> 1 - I'm trying to close the UDP transport in the on_reg_state >> callback from pjsua_acc, using the function pjsua_transport_close; >> 1.1 - If I try to pass PJ_TRUE to force the destruction of the UDP >> transport, there is an assertion error being thrown >> in pjsip_transport_destroy: >> >> PJ_ASSERT_RETURN(pj_atomic_get(tp->ref_cnt) == 0, PJSIP_EBUSY); >> >> > Just to clarify, you are trying to close UDP SIP transport, not your TURN > UDP SIP transport, right? The force=TRUE option, I'm afraid is not working. > You should use FALSE and let pjsip destroy the transport once the reference > counter is zero. > > >> 1.2 - If I try to pass PJ_FALSE, then the UDP transport is smoothly >> destroyed after a while, but in this case PJSUA maintains an invalid >> pointer in pjsua_var structure (specifically in tpdata member). >> >> > This shouldn't happen unless the transport is destroyed prematurely. > > Best regards > Benny > > > As there is no examples of real use case scenarios using PJNATH and SIP >> together, I'm clueless. >> >> >> Regards, >> Guilherme Balena Versiani. >> >> _______________________________________________ >> Visit our blog: http://blog.pjsip.org >> >> pjsip mailing list >> pjsip at lists.pjsip.org >> http://lists.pjsip.org/mailman/listinfo/pjsip_lists.pjsip.org >> >> > > _______________________________________________ > Visit our blog: http://blog.pjsip.org > > pjsip mailing list > pjsip at lists.pjsip.org > http://lists.pjsip.org/mailman/listinfo/pjsip_lists.pjsip.org > > -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.pjsip.org/pipermail/pjsip_lists.pjsip.org/attachments/20120606/8ee88181/attachment.html>
