Hi Wilson, On Fri, Sep 16, 2016 at 2:02 PM, Wilson <griscruf@xxxxxxxxxxx> wrote: > Hello, I have found a problem using Android 7.0 to connect to a bluez device over BLE. Android 6.0.1 is working. Android 7 definitely changed some things, but it is working with other (embedded) stacks like Wicentric and Searan. I first found the problem on bluez 5.37 and linux 3.14, but I still see it on bluez 5.41 and linux 4.4.0 (Ubuntu 16.04 on a Mac mini). I also tried the latest bluez in git. > > I'm running nRF Connect on the Android devices. I captured (latest) bluetoothd and btmon logs for both cases: > > Android6 bluetoothd: https://paste.gnome.org/pfkliqrqw > Android6 btmon: https://paste.gnome.org/pcyxvla79 > > Android7 bluetoothd: https://paste.gnome.org/pk4qj6wp8 > Android7 btmon: https://paste.gnome.org/p5e4fgoua Both seems to be connecting over BR/EDR and only finding GAP and GATT primary service so I wonder if you do have anything else registered there? > Using Android 6 I'm able to connect and "Refresh device cache" as much as I want and then disconnect. All the logs are shown clean in bluetoothd and nRF Connect is fast. However, using Android 7, I'm only able to connect and the cached services come up. If I refresh services, bluetoothd logs show them read: > > bluetoothd[2346]: src/device.c:gatt_debug() Read By Type - start: 0x0001 end: 0x0005 > bluetoothd[2346]: src/device.c:gatt_debug() Read By Type - start: 0x0001 end: 0x0005 > bluetoothd[2346]: src/device.c:gatt_debug() Read By Type - start: 0x0005 end: 0x0005 > bluetoothd[2346]: src/device.c:gatt_debug() Read By Type - start: 0x0006 end: 0x0009 > bluetoothd[2346]: src/device.c:gatt_debug() Read By Type - start: 0x0006 end: 0x0009 > bluetoothd[2346]: src/device.c:gatt_debug() Read By Type - start: 0x0008 end: 0x0009 > bluetoothd[2346]: src/device.c:gatt_debug() Find Info - start: 0x0009 end: 0x0009 > > but then nRF Connect is hung spinning. Disconnect does nothing. There are some differences in the logs but one major thing of note is at the end of the initial connection by Android 7 there is this: > > src/gatt-database.c:gatt_ccc_write_cb() CCC write called for handle: 0x0009 > > Is this a problem? I guess not, this must be for Service Changed which is normally subscribed. > Is a problem like this likely a bluetoothd problem or something in the linux driver? > > One thing I've noticed in OTA captures is that bluez, which is the slave, is requesting services from the master. In my application I do not have the need for this. It might help (might not) if I could disable this, and it would make the capture more clear. Is it possible to disable this? Well in GATT we can assume both roles, in fact some of the services do actually have useful information for the stack to read automatically like the device name for example, but I do recall Android having problems with that but that perhaps was related to MTU Exchange being sent over BR/EDR which was fixed recently. -- Luiz Augusto von Dentz -- To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
