nRF51 to nRF52 migration, can't discover services

RE: nRF51 to nRF52 migration, can't discover services

Eliot Stock — Sat, 19 Aug 2017 15:08:31 GMT

@RK, Changing to using BLE_GAP_ADV_FLAGS_LE_ONLY_LIMITED_DISC_MODE made no difference here, sorry. It seems that both the Android and the iOS BLE stacks will happily do service discovery even with BLE_GAP_ADV_FLAG_BR_EDR_NOT_SUPPORTED. @Hung Bui, This now sounds plausible, thanks for your help.

RE: nRF51 to nRF52 migration, can't discover services

Eliot Stock — Sat, 19 Aug 2017 15:05:57 GMT

This was indeed a caching issue, so thanks @RK, @endnode, @Hung Bui and @Nguyen Hoan Hoang for your help with this. I was pretty sure the mobile device I was testing with had never had a chance to cache anything yet for this peripheral, but I was wrong. When I went and got another Android central and fired up nRF Connect, it discovered services, no problem.

RE: nRF51 to nRF52 migration, can't discover services

Hung Bui — Mon, 07 Aug 2017 10:41:22 GMT

Good catch RK. Also I saw that the address when advertising with hrs example was e6:79:ec:64:b2:c1 when the address when advertising with Eliot's code was e6:79:ec:64:b2:c0. Could be that the phone only cached ATT table of the e6:79:ec:64:b2:c0 device.

RE: nRF51 to nRF52 migration, can't discover services

RK — Sat, 05 Aug 2017 00:35:27 GMT

You've set neither general nor limited discoverable mode in your advertising flags. That makes the device non-discoverable even though android appears to be attempting to discover it and connect to it.

I'd change that.

RE: nRF51 to nRF52 migration, can't discover services

Eliot Stock — Fri, 04 Aug 2017 13:11:41 GMT

@RK That function is now called sd_ble_gap_addr_set(). Changing the MAC address has had no effect. @Hung Bui Rebooting the phone has had no effect either. Yes, I've added a capture with the HRS example (or the bootloader from it) above - it works.

RE: nRF51 to nRF52 migration, can't discover services

Nguyen Hoan Hoang — Fri, 04 Aug 2017 01:23:37 GMT

a side from that I don't see anything else that could go wrong except the caching thing. You could also compare with my code here that works both nRF51 (SDK12) & nRF52 (SDK13).

RE: nRF51 to nRF52 migration, can't discover services

RK — Thu, 03 Aug 2017 21:06:41 GMT

phone only caches when it bonds, don't know about android (and suspect the behaviour might be different on different phones and oses).

RE: nRF51 to nRF52 migration, can't discover services

Hung Bui — Thu, 03 Aug 2017 14:41:35 GMT

Usually turn off and on the phone should also clear the cached ATT table if there is no bonding. Have you tried to test with normal SDK's example ?

RE: nRF51 to nRF52 migration, can't discover services

RK — Thu, 03 Aug 2017 08:11:37 GMT

that function still exists

RE: nRF51 to nRF52 migration, can't discover services

Eliot Stock — Thu, 03 Aug 2017 07:30:48 GMT

@RK Thanks, but how do I now set the MAC address? sd_ble_gap_address_set() seems to have gone from the SDK. Is there an example app that sets it?

RE: nRF51 to nRF52 migration, can't discover services

Eliot Stock — Thu, 03 Aug 2017 07:06:30 GMT

Thanks, I was indeed missing this code, having migrated from SDK 12 to 13 at the same time as from the nRF51 to nRF52, but adding it hasn't fixed the problem. I see the same problem: the connection works, but no service discovery happens.

RE: nRF51 to nRF52 migration, can't discover services

RK — Tue, 01 Aug 2017 23:29:58 GMT

I have to go for caching as well - there's nothing on the nordic side in those traces, the connection has been made, the peripheral is responding on time when requested, it's the central which has sent out no single packet to ask any questions.

@endnode to remove bonding on iPhone you just select the one device in the list and say 'forget this device', no need to remove all connections, I've done this a lot

@eliot simple test - generate yourself a random MAC address before starting up the stack and set it, then test, if it's caching that'll prove it, if it's not I'm fairly out of ideas.

RE: nRF51 to nRF52 migration, can't discover services

Nguyen Hoan Hoang — Tue, 01 Aug 2017 19:20:45 GMT

Do you have this function in your code ?

/**@brief Function for initializing the GATT library. */
void gatt_init(void)
{
    ret_code_t err_code;

    err_code = nrf_ble_gatt_init(&s_Gatt, gatt_evt_handler);
    APP_ERROR_CHECK(err_code);

    err_code = nrf_ble_gatt_att_mtu_periph_set(&s_Gatt, 64);
    APP_ERROR_CHECK(err_code);
}

to be called in your ble initialization after gap_param_init

#if (NRF_SD_BLE_API_VERSION > 3)
    gatt_init();
#endif

Could that be the problem ?

RE: nRF51 to nRF52 migration, can't discover services

endnode — Tue, 01 Aug 2017 19:12:19 GMT

These two captures are targeting two advertisers each having different MAC address. Are you sure that one of them wasn't cached as GATT Server on mobile side in any previous encounter? Also advertising data look like you are using DFU nor HRS demo for comparison (shouldn't matter but it is confusing). Otherwise there is no obvious reason why should the phone (assuming it's the same device?) do GATT Service Discovery on one target and not on the other....

RE: nRF51 to nRF52 migration, can't discover services

Eliot Stock — Tue, 01 Aug 2017 18:20:36 GMT

@Hung Bui Added two more captures above. How do I get logging out of the BLE sources in the SDK? What next?

RE: nRF51 to nRF52 migration, can't discover services

Hung Bui — Mon, 31 Jul 2017 14:14:14 GMT

I don't see any problem from the peripheral side according to your sniffer trace. You may want to record another one. Also try to test with the SDK example for nRF52 for comparison.

RE: nRF51 to nRF52 migration, can't discover services

endnode — Sat, 29 Jul 2017 15:20:24 GMT

Just experience, 9 from 10 cases of "missing GATT Service Discovery on mobile phone side" is typically this;) If you say these devices never communicated before and this is definitely first connection which occurs then it must be something else. Although the trace doesn't provide much other hints...

RE: nRF51 to nRF52 migration, can't discover services

Eliot Stock — Sat, 29 Jul 2017 15:15:34 GMT

Thanks, but what makes you think this is to do with GATT caching? There's never been a successful connection and service discovery between this central (Android phone) and this peripheral (nRF52).

RE: nRF51 to nRF52 migration, can't discover services

endnode — Sat, 29 Jul 2017 10:05:04 GMT

bike-tracker-nRF52-discovery-from-nrf-connect-android-app-2017-07.pcapng shows CONNECT_REQ but no single PDU afterwards so it either failed or sniffer failed to follow the packets. Second bike-tracker-nRF52-discovery-from-custom-android-app-2017-07.pcapng is better but as you said, no GATT procedure visible, just empty PDUs except LL_FEATURE_REQ/RSP procedure. Note that basically all mobile phones today use GATT caching on Client side so to see whole Service discovery "like for the first time" you need to either change GAP Peripheral adv. (MAC) address or you need to erase cache on mobile side (which can be harder then it sounds, e.g. on iPhone you need to erase network settings which hurts because it deletes also other valuable pairings/keys including WiFi if I'm not mistaken, on Android it usually is enough to find "BT Share" or similar process and wipe all data).

RE: nRF51 to nRF52 migration, can't discover services

Eliot Stock — Sat, 29 Jul 2017 09:29:47 GMT

I've added Wireshark captures above, but don't see the service discovery starting. I can see connection and bonding on one of the captures, then nothing. From the logcat on the Android side, I can see BluetoothGatt.discoverServices() starting in our custom Android app. I'm using Segger RTT. The only logging I see regarding BLE stuff right now is coming from my own code, not the SDK sources.

RE: nRF51 to nRF52 migration, can't discover services

endnode — Fri, 28 Jul 2017 19:45:04 GMT

Radio trace from BLE sniffer usually shows where GATT Service discovery procedure differs very quickly, I would start from there. When it comes to logging, are you using UART or SEGGER RTT? Is there anything going out over debug line and you just miss some particular module debug output (which?) or it doesn't output anything at all?