Changing PHY from BLE peripheral results in error in pc-ble-driver-py

[Krakonos]

If the peripheral attempts to change PHY to 2MHz, I get a logger message:

ERROR Invalid received BLE event id: 0x21

I think this is BLE_GAP_EVT_PHY_UPDATE_REQUEST which seems to be not implemented in pc-ble-driver-py. Am I correct? Is it fixable?

Thanks, Ladislav

[bihanssen]

PHY update is supported in the lower layers (ble-driver), but functions and events have not been added to ble_driver.py.

[Krakonos]

Does that mean the pc-ble-driver should accept it automatically? Or do I need to implement the handler and respond myself? Because we have sniffed out the connection the the pc-ble-driver does not respond to the request.

[bihanssen]

The phy update procedure needs to be handled, it doesn't happen automatically in ble-driver. Specifically the phy_update function and phy_update_request event is needed. To follow the existing pattern, it needs to be implemented in ble_driver.py file.

[Krakonos]

I see. For future reference, this is actually a pretty big problem: if the peripheral device requests 2MHz PHY, it expects some response. The current implementation sends none, and so the peripheral disconnects after some time (in my case I observed around 40s, but I'm not sure what defines this interval).

I'll look into implementing the call.

[bihanssen]

That is correct, it will lead to a disconnect if the request is not handled. The timeout is 30 seconds and is specified in the Bluetooth specification, same as for all similar requests. A PR on the feature would be welcome.

[jangalda-nsc]

Hi, is it any plan to implement it? We encountered the same problem.

[Krakonos]

Hi! I never really finished due to different priorities. However, I have started. I'll try to upload the progress tomorrow and check how much is remaining (feel free to bug me if I forget).

[bihanssen]

An implementation of phy update has also been started in branch ble_gap_phy by @halkver.

[jangalda-nsc]

Hi, thanks for answers. I used @halkver's branch and it worked. The only change I needed to do is adding on_gap_evt_phy_update_request and on_gap_evt_phy_update empty methods to BLEDriverObserver class.

[Krakonos]

Great, that's good news. can you share your modifications? I'd like to test those. Also, @halkver might be able to use those to include in his code?

[jangalda-nsc]

You can find my fork here: https://github.com/jangalda-nsc/pc-ble-driver-py/tree/ble_gap_phy

[halkver]

Hi, I have added it to the observer class now. I originally only added it to the observers in my test and forgot the parent class, sorry! I did however encounter some problems with my implementation, it didnt matter which values I requested, the request always got received as tx_phy = 3 and rx_phy = 3. @jangalda-nsc you did not encounter any similar problems?

[jangalda-nsc]

@halkver I have not encountered such problem. In my case, the slave (based on ncs) sends LL_PHY_REQ with TX&RX = 2M, and a master (based on pc-ble-driver-py) responds with LL_PHY_UPDATE_IND with TX&RX = 2M. So, I request is not sent by pc-ble-driver-py based device

[jangalda-nsc]

@halkver Have you figured it out? Do you have any plans to push it to master and release new version of pc-ble-driver-py? If not, could you please prepare private build for me (I have problems with vcpkg and cannot build package from source code on my own)?

[halkver]

@jangalda-nsc We havent been able to fix it yet, but as for the time being it is merged in and will have to be looked into later. Looks like the problem only lies with app initiated phy update requests. There will be released a new version soon, @bihanssen might be able to follow you up on that.