ble_app_multilink_central example breaks with addition of app_timer.c

Question

I've been slowly and incrementally adding changes to the ble_app_multilink_central example from nRF51-SDK-6.0.0. This is an exercise to understand where my other app stopped working.

As soon as I add app_timer.c to my project and also declare a timer id from the header file (which is just a uint32_t typedef) my central app breaks. I haven't added any timer init code at all. I've tested this dozens of times and it goes from a successful connection to an ASSERT every time.

Here is the debug output from a successful connection:

[APPL]:[0x00] >> DM_EVT_LINK_SECURED
[APPL]:[0x00] << DM_EVT_DEVICE_CONTEXT_LOADED
[APPL]:[0x00] >> DM_EVT_CONNECTION
[APPL]:[58 9D 13 DF B5 D5]: Connection Established

[APPL]:[CI 0x00]: Requesting GATT client create
[DB]: Starting discovery of service with UUID 0x4a30 for Connection handle 0
[APPL]:[0x00] << DM_EVT_CONNECTION
[DB]: Discovery of service with UUID 0x4a30 completed with success for Connectionhandle 0
[APPL]:[0x00] >> DM_LINK_SECURED_IND, result 0x00000000
[APPL]:[0x00] << DM_LINK_SECURED_IND
[APPL]:[0x00] >> DM_EVT_SECURITY_SETUP_COMPLETE, result 0x00000000
[APPL]:[0x00] << DM_EVT_SECURITY_SETUP_COMPLETE
[APPL]:[0x00] >> DM_EVT_DEVICE_CONTEXT_STORED
[APPL]:[0x00] << DM_EVT_DEVICE_CONTEXT_STORED

I add timer.c and the following line to main.c:

static app_timer_id_t                     m_soh_timer_id;                      /**< Advertisement sleep timer. */

and my connection setup looks like this:

[APPL]:[0x00] >> DM_EVT_LINK_SECURED
[APPL]:[0x00] << DM_EVT_DEVICE_CONTEXT_LOADED
[APPL]:[0x00] >> DM_EVT_CONNECTION
[APPL]:[58 9D 13 DF B5 D5]: Connection Established

[APPL]:[CI 0x00]: Requesting GATT client create
[APPL]: ASSERT: ..\client_handling.c, 86, error 0x00000011

The assert line indicated is from the following function:

/**@brief Function for service discovery.
 *
 * @param[in] p_client Client context information.
 */
static void service_discover(client_t * p_client)
{
    uint32_t   err_code;

    p_client->state = STATE_SERVICE_DISC;

    err_code = ble_db_discovery_start(&(p_client->srv_db),
                                      p_client->srv_db.conn_handle);
    APP_ERROR_CHECK(err_code);
}

The error returned is: NRF_ERROR_BUSY. Looking at the code, it appears that it believes that there is a discovery_in_progress

uint32_t ble_db_discovery_start(ble_db_discovery_t * const p_db_discovery,
                                uint16_t                   conn_handle)
{
    if (p_db_discovery == NULL)
    {
        return NRF_ERROR_NULL;
    }

    if (!m_initialized)
    {
        return NRF_ERROR_INVALID_STATE;
    }

    if (m_num_of_modules_reg == 0)
    {
        // No user modules were registered. Hence no services to discover.
        return NRF_ERROR_INVALID_STATE;
    }

    if (p_db_discovery->discovery_in_progress)
    {
        return NRF_ERROR_BUSY;
    }

    ble_db_discovery_srv_t * p_srv_being_discovered;

    m_num_of_discoveries_made             = 0;
    m_pending_usr_evt_index               = 0;

    p_db_discovery->curr_srv_ind          = 0;
    p_db_discovery->conn_handle           = conn_handle;

    p_srv_being_discovered = &(p_db_discovery->services[p_db_discovery->curr_srv_ind]);

    p_srv_being_discovered->srv_uuid = m_registered_modules[p_db_discovery->curr_srv_ind].srv_uuid;
    
    DB_LOG("[DB]: Starting discovery of service with UUID 0x%x for Connection handle %d\r\n",
                    p_srv_being_discovered->srv_uuid.uuid, p_db_discovery->conn_handle);
    
    uint32_t err_code;

    err_code = sd_ble_gattc_primary_services_discover(p_db_discovery->conn_handle,
                                                      SRV_DISC_START_HANDLE,
                                                      &(p_srv_being_discovered->srv_uuid));
    if (err_code != NRF_SUCCESS)
    {
        return err_code;
    }
    p_db_discovery->discovery_in_progress = true;

    return NRF_SUCCESS;
}

I don't understand the cause/effect happening here! Has anyone else seen this or can't anyone shed any light into what's going on here?

Thank you for any insight.

Erik Sanchez · Accepted Answer

OK, here is a proper answer. 
 I was able to get past the NRF_ERROR_BUSY issue by adding initialization to m_client in the 
 void client_handling_init(void) 
 function as shown below: 
 for (i = 0; i < MAX_CLIENTS; i++)
 {
 memset(&m_client[i], 0, sizeof(client_t));
 m_client[i].state = IDLE;
 m_client[i].srv_db.discovery_in_progress = false;
 }
 
 This got me past discovery. However, I was still seeing odd behavior now and then. I looked through the client_handling.c file and added error checking where I thought that it was lacking or error prone. 
 Changed the following function to return if a client was not found. 
 static void db_discovery_evt_handler(ble_db_discovery_evt_t * p_evt)
{
 // Find the client using the connection handle.
 client_t * p_client;
 uint32_t index;
 bool is_valid_srv_found = false;

 index = client_find(p_evt->conn_handle);
 
 if (index == MAX_CLIENTS)
 {
 // We didn't find the client.
 return;
 }

...
 
 Did the same thing with this function: 
 void client_handling_ble_evt_handler(ble_evt_t * p_ble_evt)
{
 client_t * p_client = NULL;

 uint32_t index = client_find(p_ble_evt->evt.gattc_evt.conn_handle);
 if (index == MAX_CLIENTS)
 {
 // We didn't find the client.
 return;
 }

...
 
 My central app now seems stable.

ble_app_multilink_central example breaks with addition of app_timer.c

Top Replies