Crash when WiFi modem sleep is disabled before BLE initialization

[thorrak]

Something I noticed recently while trying to debug mDNS issues in one of my projects is that if WiFi modem sleep is disabled, either via the ESP-IDF function esp_wifi_set_ps(WIFI_PS_NONE); or one of the variants of the the Arduino function WiFi.setSleep(false); then initializing NimBLE crashes with an error similar to the following:

abort() was called at PC 0x4017f830 on core 1


Backtrace:0x40083fb1:0x3ffd35e00x40095e35:0x3ffd3600 0x4009bae1:0x3ffd3620 0x4017f830:0x3ffd36a0 0x4017f65b:0x3ffd36c0 0x4013e261:0x3ffd36e0 0x400f48f2:0x3ffd3700 0x400decb5:0x3ffd3750 0x400d95c9:0x3ffd3790 0x401101f6:0x3ffd37b0 

  #0  0x40083fb1:0x3ffd35e0 in panic_abort at /home/runner/work/esp32-arduino-lib-builder/esp32-arduino-lib-builder/esp-idf/components/esp_system/panic.c:402
  #1  0x40095e35:0x3ffd3600 in esp_system_abort at /home/runner/work/esp32-arduino-lib-builder/esp32-arduino-lib-builder/esp-idf/components/esp_system/esp_system.c:128
  #2  0x4009bae1:0x3ffd3620 in abort at /home/runner/work/esp32-arduino-lib-builder/esp32-arduino-lib-builder/esp-idf/components/newlib/abort.c:46
  #3  0x4017f830:0x3ffd36a0 in coex_core_enable at ??:?
  #4  0x4017f65b:0x3ffd36c0 in coex_enable at ??:?
  #5  0x4013e261:0x3ffd36e0 in esp_bt_controller_enable at /home/runner/work/esp32-arduino-lib-builder/esp32-arduino-lib-builder/esp-idf/components/bt/controller/esp32/bt.c:1848
  #6  0x400f48f2:0x3ffd3700 in NimBLEDevice::init(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) at .pio/libdeps/lcd_ssd1306/NimBLE-Arduino/src/NimBLEDevice.cpp:770

Digging in, the specific line in esp-idf which is causing the crash is this one enabling software coexistence.

Interestingly, there is actually an error message logged if the WiFi modem sleep is disabled after NimBLE initialization, following BLE activity and immediately preceding a crash:

E (10532) wifi:Error! Should enable WiFi modem sleep when both WiFi and Bluetooth are enabled!!!!!!

I'm not sure what the best solution to this would be -- there is a function esp_wifi_get_ps to query the "sleep state" that returns a value that can be tested against. The value that will guarantee a crash is WIFI_PS_NONE -- the default value (and the one that you want for software coexistence) is WIFI_PS_MIN_MODEM. That said, I don't know what the effect would be of querying this if coexistence isn't needed/available in a sketch -- an alternative would be just adding a comment to the code referencing the power saving functions as a possible cause of a crash at the specific line that the trace references.

[h2zero]

Thanks!

This is probably something that should be in the docs somewhere.

I don't really want to add any code to check this setting to the library, it's better for the user to be aware of the requirements instead.

[thorrak]

Yeah — it feels to me like something that should be addressed in esp-idf, somewhere in coex_enable(), either by raising an error, changing the sleep mode, or both. I went ahead and raised an issue over there for this: https://github.com/espressif/esp-idf/issues/9595