[Freeswitch-dev] Module segfault

[Michael Jerris]

would need to see the entire module likely to see, but I suspect however
this is being called, the channel is already destroyed

On Saturday, May 28, 2016, Juan Antonio Ibañez Santorum <
juanito1982 at gmail.com> wrote:

> Hello,
>
>    I built one module similar to mod_niblebill for billing puroses. It
> works ok but sometimes I get a segfault and I don't understand why. The
> code I use in my application for generating outgoing session is:
>
> ------------------------------
> status = switch_ivr_originate(session, &peer_session, &cause,
> cadena_marcado_tmp, 0, NULL, rt->callerid_name, rt->callerid_number, NULL,
> NULL, SOF_NONE, NULL);
>
> if(status == SWITCH_STATUS_SUCCESS){
>
>       if (switch_channel_test_flag(caller_channel, CF_PROXY_MODE)) {
> switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "%s: Llamada
> switch_ivr_signal_bridge\n", rt->uuid);
> switch_ivr_signal_bridge(session, peer_session);
> } else {
> switch_channel_t *channel = switch_core_session_get_channel(session);
> switch_channel_t *peer_channel =
> switch_core_session_get_channel(peer_session);
> char *a_key = (char *) switch_channel_get_variable(channel,
> "bridge_terminate_key");
> char *b_key = (char *) switch_channel_get_variable(peer_channel,
> "bridge_terminate_key");
> int ok = 0;
> //switch_input_callback_function_t func = NULL;
> rt->peer_channel = peer_channel;
> rt->peer_profile = switch_channel_get_caller_profile(peer_channel);
>
> if (a_key) {
> a_key = switch_core_session_strdup(session, a_key);
> ok++;
> }
> if (b_key) {
> b_key = switch_core_session_strdup(session, b_key);
> ok++;
> }
> if (switch_true(switch_channel_get_variable(caller_channel,
> SWITCH_BYPASS_MEDIA_AFTER_BRIDGE_VARIABLE)) ||
> switch_true(switch_channel_get_variable(peer_channel,
> SWITCH_BYPASS_MEDIA_AFTER_BRIDGE_VARIABLE))) {
> switch_channel_set_flag(caller_channel, CF_BYPASS_MEDIA_AFTER_BRIDGE);
> }
>
> switch_core_session_enable_heartbeat(session, rt->heartbeat_interval);
>
> // Bridge
> switch_ivr_multi_threaded_bridge(session, peer_session, NULL, NULL, NULL);
> if(peer_channel){
> if((peer_profile = switch_channel_get_caller_profile(peer_channel))){
> }
> rt->codec_out = switch_core_strdup(pool,
> switch_channel_get_variable(peer_channel, "write_codec"));
> }
>
> switch_core_session_disable_heartbeat(session);
> }
> }
> ------------------------------
>
> At the end of the application function y use this code to do some tasks if
> the call gets answered:
>
> ------------------------------
> ...
>
> if(dc->rt->peer_channel && switch_channel_test_flag(dc->rt->peer_channel,
> CF_ANSWERED)){
>     // Tasks if answered
>   }
> ...
> ------------------------------
>
>
> I don't understand why sometimes switch_channel_test_flag generate a
> segfault. GDB stack trace is:
>
>
> ------------------------------
> #0  0x00007f2947415420 in pthread_mutex_lock () from /lib64/libpthread.so.0
> #1  0x00007f2948e522f3 in switch_channel_test_flag (channel=0x7f2934558a30,
>     flag=CF_ANSWERED) at src/switch_channel.c:1539
> #2  0x00007f292950dcf6 in procesar_hangup (session=0x7f293826b998)
>     at
> /usr/src/instalacion/09freeswitch/freeswitch-1.4.23/src/mod/applications/mod_svbilling/mod_svbilling.c:4743
> #3  0x00007f29295153a8 in do_routing (session=0x7f293826b998)
>     at
> /usr/src/instalacion/09freeswitch/freeswitch-1.4.23/src/mod/applications/mod_svbilling/mod_svbilling.c:3937
> #4  0x00007f2948e7d4d0 in switch_core_session_exec (session=0x7f293826b998,
>     application_interface=0xd73208, arg=0x7f293429d550 "process")
>     at src/switch_core_session.c:2754
> #5  0x00007f2948e7da32 in
> switch_core_session_execute_application_get_flags (
>     session=0x7f293826b998, app=0x7f293429d540 "svbillingapp",
>     arg=0x7f293429d550 "process", flags=<value optimized out>)
>     at src/switch_core_session.c:2624
> #6  0x00007f2948e81794 in switch_core_standard_on_execute (
>     session=0x7f293826b998) at src/switch_core_state_machine.c:283
> #7  switch_core_session_run (session=0x7f293826b998)
>     at src/switch_core_state_machine.c:535
> #8  0x00007f2948e7c23e in switch_core_session_thread (
>     thread=<value optimized out>, obj=0x7f293826b998)
>     at src/switch_core_session.c:1607
> ...
> ------------------------------
>
> The generated core can be downloaded at:
>
> https://dl.dropboxusercontent.com/u/45510863/coretx/core.zip
>
> and the line generating the segfault inside switch_channel.c is
>
> ------------------------------
> switch_mutex_lock(channel->flag_mutex);
> ------------------------------
>
> I suppose I am doing something in a wrong way but I don't know which one.
> Do anyone knows why?
>
> Regards
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-dev/attachments/20160528/2fed4665/attachment-0001.html