sd_ble_gap_rssi_start() returns NRF_ERROR_INVALID_ADDR

[Tim]

Any idea why sd_ble_gap_rssi_start(handle, 0, 0); returns NRF_ERROR_INVALID_ADDR? The documentation doesn't list that as a possible return value and I can't look in the code to see what is causing it.

I'm not doing anything fancy, I basically do this in my connection callback (I'm using mBed):

void onConnectionInterrupt(const Gap::ConnectionCallbackParams_t* params)
{
	int rv = sd_ble_gap_rssi_start(params->handle, 0, 0);
	if (rv != NRF_SUCCESS)
		printf("Failed: %d\n", rv);
}

Which prints Failed: 16.

Bonus question: Is it necessary to call sd_ble_gap_rssi_stop() in the disconnection interrupt, or is it handled cleanly by the softdevice?

My main question though is why the RSSI updates don't work. I'm using an nRF52-DK.

Edit: I tried with an nRF51-DK and this time it prints Failed: 8 which is NRF_ERROR_INVALID_STATE. There's clearly a bug in the softdevice somewhere, even it if is just returning the wrong error.

Edit 2:: Here is a minimal example that reproduces the problem. Note that this doesn't work in the mBed online compiler. In fact event the BLE_Button example doesn't seem to work at the moment - the board freezes when you connect. You might want to take a look at that!

Anyway, this code 'works' when compiled with the mBed CLI, in that it prints Failed: 8 when you connect.

#include "mbed.h"
#include "ble/BLE.h"
#include "nrf_error.h"
#include "nrf_svc.h"

const int SD_BLE_GAP_RSSI_START = 0x84;
SVCALL(SD_BLE_GAP_RSSI_START, uint32_t, sd_ble_gap_rssi_start(uint16_t conn_handle, uint8_t threshold_dbm, uint8_t skip_count));

const static char DEVICE_NAME[] = "RSSI Test";

Serial pc(USBTX, USBRX);

void connectionCallback(const Gap::ConnectionCallbackParams_t *params)
{
    pc.printf("Connected.\nStarting RSSI updates on: %d\n", params->handle);
    
    int rv = sd_ble_gap_rssi_start(params->handle, 0, 0);
    if (rv == NRF_SUCCESS)
    {
        pc.puts("Success!\n");
    }
    else
    {
        pc.printf("Failed: %d\n", rv);
    }
}


void disconnectionCallback(const Gap::DisconnectionCallbackParams_t *params)
{
    pc.puts("Disconnected.\n");
    BLE::Instance().gap().startAdvertising();
}

void bleInitComplete(BLE::InitializationCompleteCallbackContext *params)
{
    BLE& ble = params->ble;
    ble_error_t error = params->error;

    if (error != BLE_ERROR_NONE) {
        pc.puts("Error initialising BLE.\n");
        return;
    }
    
    if (ble.getInstanceID() != BLE::DEFAULT_INSTANCE) {
        pc.puts("BLE not default instance.\n");
        return;
    }

    ble.gap().onConnection(connectionCallback);
    ble.gap().onDisconnection(disconnectionCallback);

    ble.gap().accumulateAdvertisingPayload(GapAdvertisingData::BREDR_NOT_SUPPORTED | GapAdvertisingData::LE_GENERAL_DISCOVERABLE);
    ble.gap().accumulateAdvertisingPayload(GapAdvertisingData::COMPLETE_LOCAL_NAME, (uint8_t *)DEVICE_NAME, sizeof(DEVICE_NAME));
    ble.gap().setAdvertisingType(GapAdvertisingParams::ADV_CONNECTABLE_UNDIRECTED);
    ble.gap().setAdvertisingInterval(100);
    ble.gap().startAdvertising();
}

int main()
{
    pc.printf("-------\n\n");
    
    BLE &ble = BLE::Instance();
    ble.init(bleInitComplete);
    
    while (!ble.hasInitialized())
    {
    }
    
    for (;;)
    {
        ble.waitForEvent();
    }
}

Edit 3: I tested it with the official nRF5x SDK and it seems to actually work. Using the ble_peripipheral/ble_app_template example, and adding these lines to the BLE_GAP_EVT_HANDLER case:

		err_code = sd_ble_gap_rssi_start(p_ble_evt->evt.gap_evt.conn_handle, 0, 0);
		NRF_LOG_INFO("RSSI Start: %d %d\r\n", err_code, p_ble_evt->evt.gap_evt.conn_handle);

It prints RSSI Start: 0 0. So obviously there is some difference between mBed and the official SDK. Is there a way to pay for someone from Nordic to take a look at this?

[Hung Bui]

@Tim: I'm not very familiar with mbed, but have you made sure params->handle was a correct connection handle ? Usually you receive INVALID_STATE when the device is actually not connected or wrong connection handle.

Since you have the nRF51DK, you can also try to test with our SDK and check if you have the same issue.

[Tim]

With the above code it prints Starting RSSI updates on: 0. Is 0 a valid connection handle? Is that what you'd expect for the very first connection? If I disconnect and reconnect it always prints 0. That makes sense to me - in peripheral mode there's only ever one connection.

[Tim]

Also the docs say it would return BLE_ERROR_INVALID_CONN_HANDLE in that case. NRF_ERROR_INVALID_STATE apparently means Disconnection in progress. Invalid state to perform operation. but I'm calling it in the onConnection() handler! (I did also try calling it a bit after it is connected but I got the same result.)

[Hung Bui]

connection handle = 0 sounds valid. I don't know what could be wrong here.

One thing I don't know how do you plan to capture BLE_GAP_EVT_RSSI_CHANGED event ? Do you plan to modify the mbed wrap layer ? Sorry I don't have much experience with mbed.

[Tim]

Yes I modified the mbed layer.