| 03:06.03 | *** join/#asterisk-dev infobot (ibot@rikers.org) |
| 03:06.03 | *** topic/#asterisk-dev is Asterisk Development Discussion -=- http://www.asterisk.org/developers -=- Tier 2 and 3.14159265 support is in #asterisk -=- Hell hath frozen over: Asterisk is moving to Git -=- Follow on Twitter at @AsteriskDev |
| 06:02.29 | *** join/#asterisk-dev doome_ (~doome@94-21-135-49.pool.digikabel.hu) |
| 08:18.25 | *** join/#asterisk-dev tzafrir (~tzafrir@local.xorcom.com) |
| 08:18.25 | *** mode/#asterisk-dev [+o tzafrir] by ChanServ |
| 09:55.40 | dan_j | Hi. Trying to work out something. Where is the source code that creates a new pjsip transport causing it to accept incoming sip packets? |
| 10:11.39 | dan_j | think ive found it. |
| 10:11.54 | dan_j | config_transport.c ? |
| 11:41.37 | *** join/#asterisk-dev oej (~oej@h87-96-134-129.dynamic.se.alltele.net) |
| 11:41.37 | *** mode/#asterisk-dev [+o oej] by ChanServ |
| 11:52.15 | *** join/#asterisk-dev doome_ (~doome@94-21-135-49.pool.digikabel.hu) |
| 11:52.20 | *** join/#asterisk-dev oej (~oej@h87-96-134-129.dynamic.se.alltele.net) |
| 11:52.21 | *** mode/#asterisk-dev [+o oej] by ChanServ |
| 11:57.43 | *** join/#asterisk-dev oej (~oej@h87-96-134-129.dynamic.se.alltele.net) |
| 11:57.43 | *** mode/#asterisk-dev [+o oej] by ChanServ |
| 12:10.21 | file | dan_j, that is where the configuration handling occurs and where PJSIP is told to - yes |
| 12:11.24 | dan_j | Yes. I've realised that PJSIP is doing the actual work when it comes to sending/receiving SIP packets. |
| 12:11.46 | dan_j | Am I correct that res_rtp_asterisk manages the RTP data for both chan_sip and chan_pjsip ? |
| 12:12.08 | file | it manages the RTP data for everything in Asterisk, it is what listens and what does the encapsulation/sending/receiving |
| 12:15.36 | dan_j | Ok. Thanks. Thats what I thought. |
| 12:16.39 | dan_j | It's interesting that with chan_sip I dont seem to be having any issues with multihomed servers. But with chan_pjsip I do have the problem. Since they both use res_rtp_asterisk, I would expect them to act in the same way. |
| 12:17.08 | dan_j | Unless something is broken. I'm using chan_sip in asterisk 11, but when moving to pjsip, I've also upgraded to asterisk 13 |
| 12:17.10 | file | why? chan_pjsip is what decides what the RTP instance is bound to, as it gets passed into res_rtp_asterisk |
| 12:17.23 | file | if chan_sip is explicit and chan_pjsip isn't, then the behavior can differ |
| 12:18.37 | dan_j | So i should be looking in chan_pjsip and comparing it with chan_sip in order to find the issue? |
| 12:19.05 | file | yes, that would be where the problem is |
| 12:20.41 | file | well, probably in res_pjsip_sdp_rtp to be specific as that is actually where the instance is created |
| 12:22.10 | dan_j | Where is the instance created in chan_sip? |
| 12:22.15 | dan_j | *for chan_sip |
| 12:22.37 | file | somewhere in chan_sip.c |
| 12:22.43 | file | as that is where most of stuff is done |
| 12:24.25 | dan_j | Ok. I thought it might be in a separate file similar to res_pjsip_sdp_rtp |
| 12:24.30 | file | nope |
| 12:37.01 | dan_j | Does this function sound like the correct one to examine? create_outgoing_sdp_stream |
| 12:39.40 | file | maybe? I haven't looked at that code in months |
| 12:39.54 | file | ast_rtp_instance_new is what creates an RTP instance |
| 13:25.17 | dan_j | What does RAII_VAR do? |
| 13:26.12 | dan_j | Such as this example |
| 13:26.13 | dan_j | RAII_VAR(struct ast_sockaddr *, addrs, NULL, ast_free); |
| 13:32.56 | *** join/#asterisk-dev zerohalo (~zerohalo@2601:199:4200:d92e:202e:dfa3:7a86:681e) |
| 13:44.16 | dan_j | Ok. I think in res_pjsip_sdp_rtp load_module, address_ipv4 is initialised to 0.0.0.0 but never set to anything else. |
| 13:44.53 | dan_j | So when ast_rtp_instance_new is called, it's always with 0.0.0.0 as the bind address. |
| 13:45.55 | dan_j | In chan_sip, ast_rtp_instance_new is called with the chan_sip bindaddr variable, which has been updated correctly. |
| 13:46.16 | dan_j | Now I need to figure out how to get the correct bind address to ast_rtp_instance_new |
| 13:46.43 | dan_j | And if I want to submit a patch (my first one), i'll need to fix ipv6 too I assume. |
| 15:23.38 | *** join/#asterisk-dev oej (~oej@h87-96-134-129.dynamic.se.alltele.net) |
| 15:23.39 | *** mode/#asterisk-dev [+o oej] by ChanServ |
| 15:40.37 | *** join/#asterisk-dev oej (~oej@h87-96-134-129.dynamic.se.alltele.net) |
| 15:40.37 | *** mode/#asterisk-dev [+o oej] by ChanServ |
| 16:09.02 | *** join/#asterisk-dev [TK]D-Fender (~joe@64.235.216.2) |
| 18:11.35 | *** join/#asterisk-dev luckman212 (~luckman21@unaffiliated/luckman212) |
| 18:40.02 | dan_j | file: With PJSIP is the bind address stored anywhere within asterisk, or is it just passed to the PJSIP library and forgotten? |
| 18:52.04 | *** join/#asterisk-dev tzafrir (~tzafrir@bzq-179-40-172.cust.bezeqint.net) |
| 18:52.04 | *** mode/#asterisk-dev [+o tzafrir] by ChanServ |
| 19:34.16 | dan_j | file: I explained above that address_ipv4 in res_pjsip_sdp_rtp.c is set to '0.0.0.0' and then never changed. |
| 19:35.12 | dan_j | Does the fact that address_ipv4 exists mean that each transport has it's own instance of res_pjsip_sdp_rtp ? |
| 19:36.01 | file | I don't understand the question. A SIP transport and RTP are completely disassociated, they have no relation. |
| 19:36.46 | dan_j | chan_sip has a variable called bindaddr which is the IP address that chan_sip has been bound to and is listening on. |
| 19:37.01 | dan_j | In chan_sip, that same variable i used when creating a new rtp instance |
| 19:37.19 | dan_j | So the rtp and sip packets originate from the same IP |
| 19:37.22 | file | chan_sip is vastly different, and in the case of chan_pjsip you can have multiple configured transports |
| 19:37.42 | dan_j | True |
| 19:38.13 | dan_j | Forgot that chan_sip only allowed one transport and therefore could have global variables without any issues. |
| 19:39.05 | dan_j | I wonder what the purpose of this is in res_pjsip_sdp_rtp.c is then... |
| 19:39.07 | dan_j | static struct ast_sockaddr address_ipv4; |
| 19:39.18 | dan_j | A global within that file. |
| 19:39.34 | file | you have to pass an address into an RTP instance to bind it |
| 19:40.09 | file | and since res_pjsip_sdp_rtp binds to wildcard and lets the system decide the source, it saves doing the conversion |
| 19:41.12 | dan_j | Ok, however I need to get asterisk to do the binding to match the method that chan_sip uses. |
| 19:41.14 | dan_j | <PROTECTED> |
| 19:42.03 | dan_j | I'm currently looking at the create_rtp function which calls ast_rtp_instance_new |
| 19:42.21 | dan_j | create_rtp(struct ast_sip_session *session, struct ast_sip_session_media *session_media, unsigned int ipv6) |
| 19:43.01 | dan_j | I'm looking at ast_sip_session to see if it gives me any access to the transport being used, so I can obtain the IP that the transport is bound to. |
| 19:43.10 | dan_j | Then I can pass it to ast_rtp_instance_new |
| 19:43.17 | file | unless configured to do so a transport is not associated |
| 19:43.25 | file | PJSIP evaluates and determines the transport to use when constructing and sending a message |
| 19:44.04 | dan_j | oh dear. |
| 19:45.23 | file | the dialog may contain some information such as the local IP address last used/to be used on certain headers, but I haven't looked into that for these purposes |
| 19:46.45 | dan_j | So, to review so i know i've understood, pjsip transports can be bound using the transport bind option in pjsip.conf. But when a channel is created, asterisk has no idea which transport was used and therefore cant retrieve any information about the transport nor what IP that transport was bound to? |
| 19:47.23 | file | correct, it can ask PJSIP based on the available information what it would likely be - which is based on the target URI being dialed |
| 19:51.42 | dan_j | If I want to write the following to debug, do I need to wrap it in anything? Asterisk is crashing when I try it. |
| 19:52.03 | dan_j | ast_log(LOG_NOTICE, "Dans Debug transport name %s\n", session_media->transport); |
| 19:52.18 | dan_j | Getting segmentation fault |
| 19:52.35 | dan_j | I remember last time, I had to duplicate it. |
| 19:52.39 | dan_j | or something like htat |
| 19:52.40 | dan_j | thta |
| 19:52.42 | dan_j | that* |
| 19:53.18 | dan_j | session_media->transport is defined as |
| 19:53.19 | dan_j | pj_str_t transport |
| 19:53.23 | file | you have to use ast_copy_pj_str to copy it into a temporary buffer as it is not NULL terminated |
| 19:53.36 | file | and that will not contain the name of the SIP transport |
| 19:54.34 | dan_j | Wish I had a way to browse what each accessible variable contained at runtime. Would make understanding things quite a bit easier. |
| 19:58.55 | dan_j | I wonder if i can force the system not to use the incorrect IP by using something like iptables. |
| 20:19.14 | dan_j | file: based on what you said, this seems to be the pjsip function you mentioned. |
| 20:19.15 | dan_j | pjsip_endpt_acquire_transport (pjsip_endpoint *endpt, pjsip_transport_type_e type, const pj_sockaddr_t *remote, int addr_len, const pjsip_tpselector *sel, pjsip_transport **p_tp) |
| 20:19.46 | dan_j | Is there a method or a way to convert ast_sip_endpoint to a pjsip_endpoint ? |
| 20:20.16 | dan_j | session->endpoint is type ast_sip_endpoint |
| 21:30.04 | dan_j | file: just realised that someone else has also recently reported the same issue https://issues.asterisk.org/jira/browse/ASTERISK-25632 |
| 22:40.07 | *** join/#asterisk-dev jameswf (uid27319@gateway/web/irccloud.com/x-wqaglhxfwbolodkg) |
| 23:02.51 | file | does some JIRA stuff then |
| 23:06.25 | stefan27 | in dtls_srtp_setup(rtp, srtp, instance) we export keys for srtp from rtp->dtls when dtls is completed on rtp, but when dtls is complete on rtcp we do nothing with the keying material of rtp->rtcp->dtls? |
| 23:12.54 | stefan27 | does asterisk require that rtcp and rtp are not muxed onto same port? |
| 23:13.10 | file | stefan27, I think it expects to use the same key... libsrtp that is... |
| 23:13.21 | file | or at the very least our interface to it does |
| 23:15.09 | stefan27 | yeah I couldnt make out from libsrtp.pdf how to make it use seperate keys |
| 23:16.19 | stefan27 | asterisk has no rtcp-mux support? |
| 23:16.35 | file | it sends on separate ports |
| 23:22.07 | stefan27 | all right, so there's not much workaround available? I guess I should make a jira anyways |
| 23:23.26 | stefan27 | or ask the libsrtp people |
| 23:23.54 | stefan27 | anyways thanks and gn! |