Commissioner not responding to join request

[Mary]

nrf52840

NCS 1.9.1

We have a Thread project based on the Coap Client and Server example.  The Coap server acts as the Commissioner, and other devices join using a known pskd.  Sometimes this works fine, but after the server has been running a while (overnght) it does not respond to the join request.

A button press on the server allows joining:

static void on_button_changed(uint32_t button_state, uint32_t has_changed)
{
	uint32_t pressed = button_state & has_changed;
	uint32_t released = ~button_state & has_changed;

	// Button 1:  
	if (pressed & DK_BTN1_MSK) 
	{
		k_work_submit(&joining_work);		// calls activate_joining()
	}
}

// Called in workqueue joining_work
static void activate_joining(struct k_work *item)
{
	ARG_UNUSED(item);

	struct openthread_context *ot_context = openthread_get_default_context();
	otError err;

	otInstance *instance = ot_context->instance;

	otCommissionerState state;
	if ((state = otCommissionerGetState(instance)) == OT_COMMISSIONER_STATE_ACTIVE)
	{
		err = otCommissionerAddJoiner(instance, NULL, pskd, JOINER_TIMEOUT); 
		if (err != OT_ERROR_NONE)
		{
			printk("\r\notCommissionerAddJoiner() failed [%d]\r\n", err);
		}
		else
		{
			//k_timer_start(&led_timer, K_MSEC(100), K_MSEC(100));
			g_joining_allowed = true;

			printk("\r\nCommissioner joiner started for %d seconds...\r\n", JOINER_TIMEOUT);
		}
	}
	else
	{
		printk("\r\nCommissioner state = %d. Restarting...", state);

		err = otCommissionerStart(ot_context->instance, commissioner_state_cb, commissioner_joiner_cb, ot_context);
		if (err != OT_ERROR_NONE)
			printk("otCommissionerStart() failed [%d]\r\n", err);
		else
			start_joiner = true;		// trigger AddJoiner when "commissioner active" happens
	}
}

This happens successfully all the time.  I see the message "Commissioner joiner started for 120 seconds..." every time.

The Coap client uses auto joining by setting these configuration parameters.

# Thread Joiner

CONFIG_OPENTHREAD_JOINER=y

CONFIG_OPENTHREAD_JOINER_AUTOSTART=y

CONFIG_OPENTHREAD_JOINER_PSKD="J01NUS"

CONFIG_OPENTHREAD_MANUAL_START=n

On WireShark I can see the Discovery Request with "Joiner Flag: Intending" but there is no response from the Coap Server.  At the same time I have a device that is broadcasting Discovery Requests (without the joining flag set).  The Coap Server responds to these.  And also responds to requests for router information (otThreadGetRouterInfo() ) with this

--- Router fd70:a900:acc1:3d9f:0:ff:fe00:9000 ----

Ext Address: 0xfb58ff0e
Rloc16: 0x9000
IP6 Address List:
- fd70:a900:acc1:3d9f:0:ff:fe00:fc37    Commissioner
- fd70:a900:acc1:3d9f:0:ff:fe00:fc00    Leader
- fd70:a900:acc1:3d9f:0:ff:fe00:9000
- fd70:a900:acc1:3d9f:355e:d60e:5def:e87b    Coap Server
- fe80:0:0:0:cff:58fb:cdf3:6c77

So, the Coap Server is the Commissioner, and joining is allowed.  So, why isn't it responding to the joiner's request?

Why does this work sometimes and but not always?

Where can I look to diagnose this?

Mary

[Mary]

That is not our actual PSKd, I edited over it to keep ours confidential.

What NCS versions can be certified?

Mary

[Amanda Hsieh]

Hi, 

Certifications are listed in the nRF52840 nRF Connect SDK Thread CIDs table.

Certification for NCS 1.9.X has been deprecated due to the security issue and we didn't release a patch for this version.

It is recommended to use the 2.4.1 release (certification we have already received, but infocenter update is pending).

-Amanda H.

[Mary]

I have migrated the project to NCS 2.4.1, and I saw the same problem today.

The server/commissioner has been running overnight, ~16 hours.

Here is the Wireshark file:

Client_join_failed.pcapng

Mary

[Amanda Hsieh]

Hi, 

I'm just now noticing that you don't use

void openthread_api_mutex_lock(struct openthread_context *ot_context);

and

void openthread_api_mutex_unlock(struct openthread_context *ot_context);

when you're calling the openthread API directly. Not sure if this is the cause of the problem. Please verify the problem still exists after fixing this.

-Amanda H.

[Mary]

I added the  mutex lock/unlock around the calls to otCommissionerAddJoiner() and any other calls that change something.  

Some days later, the problem reappeared.  I tested on two units running identical code.  I verified that using nrfjprog --verify.  One unit allowed a device to join, and the other didn't.  I did a factory reset on the one that failed, and then joining succeeded.

Mary