Automatic reconnection when peripheral disconnects

[noonfom]

• bleak version: 0.10.0
• Python version: 3.7.3
• Operating System: Windows 10

Description

On Windows, a bleak client will automatically reconnect when the peripheral disconnects. This is undesirable behaviour in cases where it is required to completely terminate the connection with the host. For example, consider a DisconnectionService with a
a single characteristic: disconnection_reason_char. If a valid disconnection reason is written to the characteristic, the service immediately disconnects from the host. Such a service is useful when writing integration tests for BLE devices.

What I Did

I wrote a BLE application using mbed-os that contains a DisconnectionService. Then, I built and flashed the application onto a NUCLEO-WB55RG; however, any Mbed BLE-enabled development board should work. Lastly, I wrote a Python script to interact with the service using bleak.

The main run function is shown below. It scans for and attempts to connect to a device named "Disconnection", viz. the name of my peripheral. Once connected, it writes a valid disconnection reason to disconnection_reason_char of the DisconnectionService. Therefore, the peripheral immediately disconnects from the host.

async def run():
    devices = await BleakScanner.discover()
    for d in devices:
        if d.name == "Disconnection":
            client = BleakClient(d)
            connected = await client.connect()
            if connected:
                await client.write_gatt_char(UUID_DISCONNECTION_REASON_CHAR, USER_TERMINATION)
                time.sleep(60)
                connected = await client.is_connected()
                l.info(f'Connected: {connected}')

I set the log level to DEBUG and observed traces from bleak (see below). We see that _ConnectionStatusChanged_Handler gets flipped to 0 after disconnection_reason_char is written with 0x13. This is expected because that corresponds to user termination, a valid disconnection reason. However, ~15 seconds later _ConnectionStatusChanged_Handler gets flipped to 1; and, after the 1 minute pause, we see that the client has reconnected with the peripheral.

2021-03-08 15:43:52,832.832 DEBUG Using selector: SelectSelector
2021-03-08 15:43:53,442.442 DEBUG Received E9:1E:1F:57:F9:4C: Disconnection.
2021-03-08 15:43:53,443.443 DEBUG Received xx:xx:xx:xx:xx:xx: Unknown.
2021-03-08 15:43:55,516.516 DEBUG Received xx:xx:xx:xx:xx:xx: Unknown.
2021-03-08 15:43:55,517.517 DEBUG Received xx:xx:xx:xx:xx:xx: Unknown.
2021-03-08 15:43:56,876.876 DEBUG Received xx:xx:xx:xx:xx:xx: Unknown.
2021-03-08 15:43:56,877.877 DEBUG Received xx:xx:xx:xx:xx:xx: Unknown.
2021-03-08 15:43:57,899.899 DEBUG Connecting to BLE device @ E9:1E:1F:57:F9:4C
2021-03-08 15:44:04,561.561 DEBUG _ConnectionStatusChanged_Handler: 1
2021-03-08 15:44:04,563.563 DEBUG Get Services...
2021-03-08 15:44:05,954.954 INFO Services resolved for BleakClientDotNet (E9:1E:1F:57:F9:4C)
2021-03-08 15:44:06,077.077 DEBUG Write Characteristic f43620d0-779d-11eb-9439-0242ac130002 : bytearray(b'\x13')
2021-03-08 15:44:06,253.253 DEBUG _ConnectionStatusChanged_Handler: 0
2021-03-08 15:44:16,691.691 DEBUG _ConnectionStatusChanged_Handler: 1
2021-03-08 15:45:06,077.077 INFO Connected: True

Compiling and running the Mbed BLE application

1. Clone and enter the repository:

   git clone https://github.com/noonfom/mbed-os-example-disconnection-service
   cd mbed-os-example-disconnection-service
   

2. Link dependencies:

   mbed deploy
   

3. Build and flash for a given <target> and <toolchain>:

   mbed compile -t <target> -m <toolchain> -f
   

Attachments

automatic_reconnection_log.txt

P.S. apologies for all the Mbed-specific stuff.

[dlech]

                time.sleep(60)

This is blocking the async event loop from running. Use await asyncio.sleep(60) instead.

[noonfom]

@dlech Thanks for the quick reply. I applied your suggestion and it fixed the issue 🎉
However, let's say the blocking function is not time.sleep() but a wait_for_output() function that waits for a fixed period of time until a specific response is received on the serial line. What would you advise then?

[dlech]

There is https://pyserial-asyncio.readthedocs.io, but I don't think it works on Windows.

Another option is to use something like loop.run_in_executor() to run the blocking task in a background thread.

[dlech]

Or, based on your description, asyncio.Event might be more appropriate to get a signal from a background thread.