[1.12.1] A crash occurs when the protocol stack is destroyed.

[susuyu]

When the UPNP protocol is started and the HTTP get request from the control point is received and processed, if the protocol stack is destroyed and resources are destroyed for example in web_server_destroy, the thread that processes the HTTP get instruction crashes when the gWebMutex is used.

[mrjimenez]

Hi,

Could you please post a running code example of the problem, so that we can objectively deal with it?

Regards,
Marcelo.

[susuyu]

This problem occurs occasionally. When a crash occurs, the stack is as follows:
backtrace:

      #00 pc 00054502  bionic/libc.so (abort+166)
      #01 pc 0009d783  bionic/libc.so (__fortify_fatal(char const*, ...)+26)
      #02 pc 0009cf8d  bionic/libc.so (HandleUsingDestroyedMutex(pthread_mutex_t*, char const*)+20)
      #03 pc 0009ce79  bionic/libc.so (pthread_mutex_lock+132)
      #04 pc 0001db21  libupnp.so (web_server_callback+2872)

The corresponding functions are web_server_callback() and UpnpFinish().
The location of crash is line 408 of alias_release, the crash code location is as follows:

static void alias_release(
	/*! [in] XML alias object. */
	struct xml_alias_t *alias)
{
	ithread_mutex_lock(&gWebMutex);  ---> crash
	/* ignore invalid alias */
	if (!is_valid_alias(alias)) {
		ithread_mutex_unlock(&gWebMutex);
		return;
	}

[mrjimenez]

There is not much I can imagine here that could be going wrong like this. The crash suggests that the web mutex is corrupted.

• Bionic is Ubuntu, right? Is it possible for you to link to 1.14.7 or at least to some 1.14.x release and enable debug symbols?
• What is the application that is using libupnp?
• If that is something you have written, is it possible to give us a minimal example that reproduces the crash?

Regards,
Marcelo.