Adapting nRF9160's coap_client sample in NCS

[Zeedel]

Hi,

I am adapting the nRF9160 coap_client sample, and have 2 questions:

1. If I do not care about responses what is the ideal / cleanest way of configuring the app or underlying socket to ignore / drop them?
2. On the other hand, if I choose to consume the responses, is it ok to have this handled in a one thread while coap requests are periodically dispatched in via a work_q that leverages a different thread? Could any concurrency issues or race conditions arise in the underlying socket or accompanying polling data structures? A draft of what I have in mind is as follows (note that the handler for send_telemetry_work recursively enqueues the work for subsequent coap requests):

       k_delayed_work_submit_to_queue(&application_work_q, &send_telemetry_work, K_SECONDS(0));
   
       int err, received;
       while (1) {
   		
   		err = wait(POLL_TIMEOUT_S);
   		if (err < 0) {
   			if (err == -EAGAIN) {
   				continue;
   			}
   
   			printk("Poll error, exit...\n");
   			break;
   		}
   
   		received = recv(sock, coap_buf, sizeof(coap_buf), MSG_DONTWAIT);
   		if (received < 0) {
   			if (errno == EAGAIN || errno == EWOULDBLOCK) {
   				printk("socket EAGAIN\n");
   				continue;
   			} else {
   				printk("Socket error, exit...\n");
   				break;
   			}
   		}
   
   		if (received == 0) {
   			printk("Empty datagram\n");
   			continue;
   		}
   
   		err = client_handle_telemetry_send_response(coap_buf, received);
   		if (err < 0) {
   			printk("Invalid response, exit...\n");
   			break;
   		}
   	}

Thanks.

[Heidi]

Hi!

1. There's no way to configure a socket to automatically drop the RX data. The easiest way to achieve such behavior would be a dummy recv() call whenever POLLIN is reported on a socket.

2. It is ok to poll() for data on one thread and send() requests from another. Then it is up to the application to synchronize requests with responses.

Best regards,

Heidi

[Zeedel]

Thanks for clarifying Heidi.