Fix infinite loop in mqttsimple

[nmanne-anlgov]

If the connection is closed, recv returns 0 and gets stuck in an infinite loop.

[wolfSSL-Bot]

Can one of the admins verify this patch?

[embhorn]

Hi @nmanne-anlgov

Thanks for contacting wolfSSL Support. The socket timeout should handle the condition you described:

    /* Setup timeout */
    setup_timeout(&tv, timeout_ms);
    (void)setsockopt(*pSockFd, SOL_SOCKET, SO_RCVTIMEO, (char *)&tv,
            sizeof(tv));

wolfMQTT/examples/mqttsimple/mqttsimple.c

Lines 224 to 244 in 6db6c74

	/* Setup timeout */
	setup_timeout(&tv, timeout_ms);
	(void)setsockopt(*pSockFd, SOL_SOCKET, SO_RCVTIMEO, (char *)&tv,
	sizeof(tv));
	/* Loop until buf_len has been read, error or timeout */
	while (bytes < buf_len) {
	rc = (int)recv(*pSockFd, &buf[bytes], buf_len - bytes, 0);
	if (rc < 0) {
	rc = socket_get_error(*pSockFd);
	if (rc == 0)
	break; /* timeout */
	PRINTF("NetRead: Error %d", rc);
	return MQTT_CODE_ERROR_NETWORK;
	}
	bytes += rc; /* Data */
	}
	if (bytes == 0) {
	return MQTT_CODE_ERROR_TIMEOUT;
	}

Is the timeout not working with your setup?

Could you tell us a bit about your project using wolfMQTT?

Kind regards,
@embhorn - wolfSSL Support

[embhorn]

I was able to reproduce this issue by killing the broker task while the client was waiting for a ping response. I've added a fix here:
#407

Let me know if that resolves the issue for you, also.

Thanks,
@embhorn

[nmanne-anlgov]

Hi @embhorn,

Thanks for contacting wolfSSL Support. The socket timeout should handle the condition you described:

Is the timeout not working with your setup?

I think that timeout is for an individual recv call. If the other side doesn't send any message within the timeout, the recv call returns -1 and it exits the loop.

Upon successful completion, recv() returns the length of the message in bytes. If no messages are available to be received and the peer has performed an orderly shutdown, recv() returns 0. Otherwise, -1 is returned and errno is set to indicate the error.

But if the connection somehow gets closed, recv returns 0 immediately and it doesn't trigger the timeout or any of the other errors, causing this loop to go into a 100% CPU infinite loop.

In our project, we are using wolfMQTT on an embedded linux system to connect to a local broker from multiple processes running on it. These processes use MQTT to communicate among each other.
We had used the mqttsimple example as a base for implementing this communication in all the applications. After running for a while, some of these processes get stuck in a 100% CPU loop.
Using some debug printing, I realized its happening in mqtt_net_read, if they somehow disconnect from the broker. Adding this line solved that issue and I'm not seeing any more instances of 100% CPU usage.

Regards,
Nithin

[nmanne-anlgov]

I was able to reproduce this issue by killing the broker task while the client was waiting for a ping response. I've added a fix here: #407

Let me know if that resolves the issue for you, also.

Thanks, @embhorn

Thanks for the fix, yeah I think this should also solve the issue.

[embhorn]

Great! I'll close this PR in favor of the fix in #407