LESC with MITM for Peripheral hardfault running on FreeRTOS Issue

[Eduardo Garcia]

Hello:

I am porting the LESC with MITM security protocol from your "ble_app_multirole_lesc" example inyo a project that runs FreeRTOS. This is on an NRF32540 With s140.

I am implementing MITM identical as in the example - a button determines if we accept or deny the bonding. However I have run into issues where a hardfault is triggered after a button admits/rejects the connection. The hardfault seems to be when FreeRTOS is switching contexts (function xPortPendSVHandler in port.c), and it happens in PendSV. My issue seems to be exactly the one described here (https://devzone.nordicsemi.com/f/nordic-q-a/45268/data-bus-hard-fault-when-handling-secure-connection-bonding) but unfortunately I could not find the solution. I preferred to make a new case rather than reviving a years old one.

Thinking it had to do something with the button interrupt throwing everything into haywire, I modified the import of the example to set a flag in the button interrupt. A thread will check for that flag and execute the bond check code (rather than it happening on an interrupt context). I did the following (other than porting the code from the example to my file):

• Notes on importing: You will need to do the change as mentioned here: https://devzone.nordicsemi.com/f/nordic-q-a/59655/lesc-and-freertos?ReplySortBy=CreatedDate&ReplySortOrder=Descending
• On my BLE file

static void bsp_event_handler(bsp_event_t event)

...

case BSP_EVENT_KEY_0:

    SetBondingAllowed(true); /*Set a flag so "on_num_comp_button_press" is called from
                            a thread later on rather than in the interrupt*/
                            
    break;

case BSP_EVENT_KEY_1:
    SetBondingAllowed(false);
    break;

• On my thread

void vApplicationIdleHook( void )
{
#if NRF_LOG_ENABLED
        if(flagBondConn)
      {
        flagBondConn = false;
        on_num_comp_button_press(bondResult);
      }
     vTaskResume(m_logger_thread);
#endif
}

My question is, is this correct? I see it currently working, but I am not sure if I am missing another example or patch that has implemented LESC with MITM running on FreeRTOS. We really need Freertos for the complexity of our project! Do you think the SD call to check the bond (sd_ble_gap_auth_key_reply) cannot be called on an FreeRTOS context?

Let me know what you think!

Thanks,

Eduardo

[bjorn-spockeli]

Do you think the SD call to check the bond (sd_ble_gap_auth_key_reply) cannot be called on an FreeRTOS context?

 The FreeRTOS tasks are running on Thread priority so there should not be any problem calling sd_ble_gap_auth_key_reply from a FreeRTOS task context. 

 So if you call sd_ble_gap_auth_key_reply from the bsp_event_handler, then you would hardfault?

A small question related to this - the MITM example makes whoever connects attempt to bond as soond as the connection starts. However on my port I have to initiate the bonding from the NRF Connect App. I still cannot read/write to characteristics, but I can't figure out why NRF Connect does not attempt bonding from the start on my port. Any tips on this would be appreciated. Thank you!

Let me know what you think!

 If nRF Connect does not sned  a Pairing request, then the peripheral can trigger this by calling sd_ble_gap_authenticate, see Peripheral Security Request

Best regards

Bjørn

[Eduardo Garcia]

Hi Bjørn:

That is correct - when porting the MITM project to FreeRTOS I would eventually get a hardfault if I kept the same scheme of calling the logic on the button press (on_num_comp_button_press(bondResult) called on BSP_EVENT_KEY_0/1).

When I changed it to the button interrupt setting a flag and the same MITM logic called from a task it would work. It seems identical to the issue on the link I posted - sadly the poster never came back to say if their issue was solved or not!

My application is not running anything else other than the SD thread, LESC thread, and the idle thread, where I put the code to handle the flag when the MITM button is pressed. 

[bjorn-spockeli]

Hi Eduardo, 

Apologies for the late reply. 

Would it be possible for you to share the project so that I may debug it here on my end?

Best regards

Bjørn