Crashing Windows Bluetooth stack from ESP32?
Posted: Fri May 13, 2022 12:48 am
Anybody crashing their Windows stack with heavy HFP and/or SPP use on their ESP32? That's what I'm seeing, but I'm struggling to isolate where the problem lies: In the ESP or in Windows.
There are very few Google results suggesting the Windows stack is susceptible to crashing, but that seems to be happening for me:
- My BT mouse stops working, and toggles between "connected" and "paired" in Windows BT settings. Powering off WIndows BT and back on again usually fixes this.
- My Bluetooth comports (from several devices) cannot be connected to from Windows, and sometimes disappear entirely from any enumeration of comports, requiring a reboot to restore.
It usually take 5-6 hours to crash, with SPP & HFP both continuously connected, and only mild traffic on SPP. So testing is very difficult if I can only run one test per night!
(We had abysmal performance with ESP-IDF 4.3 where it would crash within 10 minutes, and seemed to be related to interrupts. We discovered a fix to the btController binary lib in 4.4 fixed this, but the source code for this is not published so I've no idea what the fix was)
There are very few Google results suggesting the Windows stack is susceptible to crashing, but that seems to be happening for me:
- My BT mouse stops working, and toggles between "connected" and "paired" in Windows BT settings. Powering off WIndows BT and back on again usually fixes this.
- My Bluetooth comports (from several devices) cannot be connected to from Windows, and sometimes disappear entirely from any enumeration of comports, requiring a reboot to restore.
It usually take 5-6 hours to crash, with SPP & HFP both continuously connected, and only mild traffic on SPP. So testing is very difficult if I can only run one test per night!
(We had abysmal performance with ESP-IDF 4.3 where it would crash within 10 minutes, and seemed to be related to interrupts. We discovered a fix to the btController binary lib in 4.4 fixed this, but the source code for this is not published so I've no idea what the fix was)