Set the RTSP socket to close on exec() of child processes

[mikeshir]

... otherwise background run_this_before_play_begins | run_this_after_play_ends commands that are sleeping prevent the daemon from being restarted because the listening RTSP port is still in use.

diff --git a/rtsp.c b/rtsp.c
index e78456f..46de66c 100644
--- a/rtsp.c
+++ b/rtsp.c
@@ -1839,6 +1839,7 @@ void rtsp_listen_loop(void) {
continue;
}

• fcntl(fd, F_SETFD, FD_CLOEXEC);
  ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(yes));

ifdef IPV6_V6ONLY

[mikebrady]

Thanks. I'll get to this in a couple of days.

[mikebrady]

Hi there. I'm afraid I don't understand this. Could you explain it a bit more please?

[Update] Are you saying that a sleeping run_this_before_play_begins or a run_this_after_play_ends process will keep the port in use when main program is terminated?

[mikeshir]

Yes, that's what I'm saying. And /etc/init/d/sahirport-sync restart doesn't work as a result. I use the callbacks to turn my amps on and off and I sleep in the after_play callback. That fcntl() sets the file descriptor to close automatically if any of the exec...() functions are called.

[mikebrady]

Thanks for that – I've incorporated it into 2.8.4.3 which is out now, on the development branch. (BTW, using the same reasoning, I guess I should also be closing any open file descriptors. Is that correct, do you think?)

[mikeshir]

It shouldn't matter for open files unless you open them exclusively using file locking. I don't think you do since that one line fixed my problem. The listening socket on a specific port is the only exclusive resource that cannot be shared and therefore needs to be released.

[mikebrady]

Thanks.

[mikebrady]

There have been many updates since this, and I'm pretty sure the problem has been resolved. Please open a new issue if necessary.