CLI over TCP returns only one line

Hallo!
I am trying to implement a command line interface using the libraries provided in FreeRTOS. I managed to run with the UDP connection, however I encounter two bugs when implementing the same functionality with a TCP:

  • the first is that the TCP connection stops working after 8 commands sent.
  • the second is that no matter the pdFALSE value only one line gets printed out over the TCP. If a command returns more lines, only the first one gets printed out.

None of these bugs were present when using UDP. I tried separating the ‘listening’ and ‘respond’ functionality in separate tasks, but it does not solve the problem.
I am using lwIP stack for the network communication.
Does anybody have an idea why the task freezes after 8 commands have been recieved? Or maybe even how to get more lines printed out? The tcp connection should still be active - there is a separate function to close a connection without deleting it. I am not using it here, so - as far I understand - I should be able to write() to the connection as long as it is up.

My code:

 void vTCPCommandConsoleTask(void* pvParamaters){

struct netconn *pxNewConnection, *tcpconn;
struct netbuf *buf, *pxRxBuffer, *buf_send;
err_t connection_err, send_err, bind_err, listen_err, accept_err, recv_err;
ip_addr_t board_addr;
unsigned int payload_len;
char *data, *payload_data;
BaseType_t xMoreDataToFollow;
static signed char cOutputString[ configCOMMAND_INT_MAX_OUTPUT_SIZE ], cLocalBuffer[ configCOMMAND_INT_MAX_OUTPUT_SIZE ]; 

/* create a new connection */
tcpconn = netconn_new(NETCONN_TCP);

/* bind the connection to a local IP */
IP4_ADDR((&board_addr.u_addr.ip4), netif_cfg[0]->ip_addr[0], netif_cfg[0]->ip_addr[1], netif_cfg[0]->ip_addr[2], netif_cfg[0]->ip_addr[3]);
bind_err = netconn_bind(tcpconn, &board_addr, SRC_PORT_NUM );

/* start listening and accept */
if (bind_err== ERR_OK) {
    listen_err = netconn_listen(tcpconn);
	for( ;; ){
    	      accept_err = netconn_accept(tcpconn, &pxNewConnection);
                  if(accept_err == ERR_OK){
    		       /*recieve the packet */
    		 	recv_err = netconn_recv(pxNewConnection, &pxRxBuffer);
    		 	if ( recv_err == ERR_OK ){
    		 	        /* Get the payload and length */
    		 		payload_len = pxRxBuffer->p->len;
    		 		payload_data = pxRxBuffer->p->payload;

    		 		/* Copy the recieved message into the command string */
    		 		signed char cInputString[payload_len];
    		 		strncpy(cInputString, payload_data, payload_len);

    		 		/* create new buffers and send them for each line of the output */
    		 		do{
    		 			/*Pass the string to FreeRTOS+CLI. */
    		 			xMoreDataToFollow = FreeRTOS_CLIProcessCommand( cInputString, cOutputString, configCOMMAND_INT_MAX_OUTPUT_SIZE );

    		 			/*create a buffer to send the data and fill it with output */
    		 			buf_send = netbuf_new();
    		 			data = netbuf_alloc(buf_send, sizeof(cOutputString));
    		 			memcpy(data, cOutputString, sizeof(cOutputString));

    		 			/* Send the output generated by the command's implementation. */
    		 			netconn_write( pxNewConnection, data , (u16_t) configCOMMAND_INT_MAX_OUTPUT_SIZE, NETCONN_COPY );

    		 			/* Free the buffer */
    		 			netbuf_delete(buf_send);
    		 			} while( xMoreDataToFollow != pdFALSE ); /* keep sending until the command does not generate any more output. */

    		 	} //rev_err == ERR_OK

    		  } //accept_err

    } //while(1)

} //bind_err

}

This looks to be very lwIP specific so would recommend asking on the lwIP mailing list.

The bug was in the specific implementation of the tcp client I have been using. This problem has been solved by sending an empty string, while in the loop. Adding char test_data[1] = " "; netconn_write(pxNewConnection, test_data, 1, NETCONN_COPY); just before exiting the command execution loop made all the output strings to get printed out.

The code above had also other bugs, that made it crash on the MPC5748G. All of these bugs were caused by memory overflow. First of all every new call to nectonn_recv() would allocate a new buffer. The problem has been solved by adding netbuf_delete(pxRxBuffer); at the end of the receive loop. This way the program makes sure that the memory will be used over again.
Furthermore, he netconn_accept() call invokes a new connection every time it is called. Therefore another memory overflow can be possible, and a call to netconn_close(pxNewConnection); netconn_delete(pxNewConnection); needs to be made at the end of the accept loop to free the memory.

Thanks for taking the timer to report back.