Given a scenario where MixMonitor was initiated over AMI it
was possible for the channel and MixMonitor thread to remain
alive past hang up of the channel. This scenario required
the AMI initiated MixMonitor to retrieve the channel, a
hangup to occur on the channel in another thread, and then
for MixMonitor to actually start. If this occurred the
MixMonitor thread would remain alive indefinitely and
the channel reference would remain.
This change ensures that audiohooks are never able to
be attached to channels that have been hung up. An
additional fix has also been done in app_mixmonitor to
properly release the channel reference if this occurs.
ASTERISK-28780
Change-Id: I8044c06daa06f0f16607788c596f55623be26f58
if (startmon(chan, &mixmonitor->audiohook)) {
ast_log(LOG_WARNING, "Unable to add '%s' spy to channel '%s'\n",
mixmonitor_spy_type, ast_channel_name(chan));
+ ast_autochan_destroy(mixmonitor->autochan);
ast_audiohook_destroy(&mixmonitor->audiohook);
mixmonitor_free(mixmonitor);
return -1;
{
ast_channel_lock(chan);
+ /* Don't allow an audiohook to be attached to a channel that is already hung up.
+ * The hang up process is what actually notifies the audiohook that it should
+ * stop.
+ */
+ if (ast_test_flag(ast_channel_flags(chan), AST_FLAG_ZOMBIE)) {
+ ast_channel_unlock(chan);
+ return -1;
+ }
+
if (!ast_channel_audiohooks(chan)) {
struct ast_audiohook_list *ahlist;
/* Whoops... allocate a new structure */