Zigbee Task halts on Leave Request

[sjf]

Hi,

We have been working with an end device based on nRF52840 with nRF5_SDK_for_Thread_and_Zigbee_v4.0.0_dc7186b and FreeRtos

This device has some bindings already set to the coordinator.

When it receives a Leave Request from coordinator, the Zigbee Stack gets stucked on zb_address_reset() function. Because of that, the Zigbee task remains blocked for the rest of the time.

Which should be the reason to halt at this point and how this could be avoided?

Regards,

Santiago

[Marte Myrvold]

Hi Santiago,

I've started looking into your issue.

Is the project on the end-device based on one of the examples in our SDK? If so, which one? Is the leave request part of the program and thus on purpose? If that's the case, how are you doing the leave request?

Could you please perform a sniffer trace and upload the sniffer log here?

Best regards,

Marte

[sjf]

It was based on an end device example but it has many modifications form that template. 

Basically we create some cluster bindings from the end device to the coordinator automatically once connected for the first time.

When the coordinator wants to remove the child, it sends a leave request (zdo_mgmt_leave_req()) with the following data:

p_req->dst_addr = <end device short address>

p_req->device_address = <end device ieee address>

p_req->rejoin = false

I'll try to get Wireshark log for you today/tomorrow

Regards,

[Marte Myrvold]

Hi Santiago,

Where is zb_address_reset() called, and what is it used for in your project?

Best regards,

Marte

[sjf]

Hi Marte, 

It is invoked internally by the Zigbee Task. We do not call it at all.

Consider that there are some bindings set on this device as extra information.

Please, we need to get a response as soon as possible from you/Zigbee team. 

Regards,

[Marte Myrvold]

Hi Santiago,

I have forwarded your problem and will get back to you as soon as I get a response.

In the meantime, what does it look like in your signal handler when the leave signal is received?

Best regards,

Marte

[sjf]

Hi Marte,

There is no signal triggered. It just halts

[Marte Myrvold]

Hi Santiago,

What does your code look like? After ZB_ZDO_SIGNAL_LEAVE is called.

Best regards,

Marte

[sjf]

ZB_ZDO_SIGNAL_LEAVE is never called. In fact, the description of this signal says that is called when the device leaves the network by sending leave command. 

Find code below:

case ZB_ZDO_SIGNAL_LEAVE:
        /* This signal is generated when the device itself has left the network by sending leave command.
          *
          * Note: this signal will be generated if the device tries to join legacy Zigbee network and the TCLK
          *       exchange cannot be completed. In such situation, the ZB_BDB_NETWORK_STEERING signal will be generated
          *       afterwards, so this case may be left unimplemented.
          */
        if (status == RET_OK)
        {
          zb_zdo_signal_leave_params_t * p_leave_params = ZB_ZDO_SIGNAL_GET_PARAMS(p_sg_p, zb_zdo_signal_leave_params_t);
          ZIGBEE_LOG_INFO("Network left (leave type: %d)", p_leave_params->leave_type);

          if(ZIGBEE_STATE_JOINING != zigbee.state)
          {
            App_Zigbee_Status(ZIGBEE_STATUS_DISCONNECTED);
            Zigbee_Fsm(ZIGBEE_STATE_DISCONNECTED);
          }
        }
        else
        {
          ZIGBEE_LOG_ERROR("Unable to leave network (status: %d)", status);
        }
 

      break;

[Marte Myrvold]

H Santiago,

The end device sends a leave command in your sniffer log, so it does leave, and therefore ZB_ZDO_SIGNAL_LEAVE is called. This signal is generated when the device itself has left the network by sending leave command. ZB_ZDO_SIGNAL_LEAVE is called even if the leave request is from the coordinator. You can test this yourself by using the CLI Agent example and issuing the 'zdo mgmt_leave' command, and you will see that "Network left" is printed.

Can you please debug your code using breakpoints so that you can see what's happening in your program? You can put the breakpoints at the if statements in the code above, as well as at App_Zigbee_Status() and Zigbee_Fsm().

What is zigbee.state, and what does the functions App_Zigbee_Status() and Zigbee_Fsm() do? 

Best regards,

Marte

[Marte Myrvold]

Hi Santiago,

I just got a response from the Zigbee team. They were able to reproduce the problem, with both nRF5 SDK T&Z and with NCS, so it seems the problem is with the Zigbee stack (ZBOSS). The problem has been reported to DSR in ZBOSS Open Initiative (ZOI).

Best regards,

Marte

[sjf]

Hi Marte,

Thanks for your response. Is there any workaround?

Regards,

[Marte Myrvold]

Hi Santiago,

I don't know of any workaround yet, but our Zigbee team is looking into it to try and see if there's an obvious fix besides waiting for a response from DSR. However, I can't give an estimate on when this will be done, as it might take some time. I'll let you know as soon as I know more.

Best regards,

Marte