ESP32: check if OTA/new firmware is working properly

1

Hi,
I am not sure if this is PlatformIO issue or Espressif or Arduino.
I am compiling the project with:
build_flags = -D CONFIG_BOOTLOADER_APP_ROLLBACK_ENABLE=1
and:

platform = espressif32
framework = arduino

expecting, that I will be able to check post-OTA if the program is working OK or not, and in case not working ok the rollback would be possible.
For this I use example from Espressif github:

  esp_ota_img_states_t ota_state;
  Serial.print("ota_state:");Serial.println(esp_ota_get_state_partition(running, &ota_state));
  if (esp_ota_get_state_partition(running, &ota_state) == ESP_OK) {
      if (ota_state == ESP_OTA_IMG_PENDING_VERIFY) {
          // run diagnostic function ...
          bool diagnostic_is_ok = diagnostic();
          if (diagnostic_is_ok) {
              Serial.printf("Diagnostics completed successfully! Continuing execution ...\n");
              esp_ota_mark_app_valid_cancel_rollback();
          } else {
              Serial.printf("Diagnostics failed! Start rollback to the previous version ...\n");
              esp_ota_mark_app_invalid_rollback_and_reboot();
          }
      }
  }

but the outcome (after OTA) is always the same: ota_state = ESP_OK
that means: I am not able to control the rollback

it looks like OTA itself marks the partition to be valid or I don’t know what exactly who is marking it to be valid.
Anybody experienced it and found the solution?

2

Well if you are using the ArduinoOTA library, which in turns uses the Update library, when an OTA update is received correctly, it will do

I think this will code will mark it as valid (so no ESP_OTA_IMG_PENDING_VERIFY) and reboot the system.

Nonetheless, you should still be able to perform a rollback by calling into the rollback functions

i.e.,

          if (diagnostic_is_ok) {
              Serial.printf("Diagnostics completed successfully! Continuing execution ...\n");
              // Literally nothing to do
          } else {
              Serial.printf("Diagnostics failed! Start rollback to the previous version ...\n");
              if (Update.canRollBack()) {
                  Update.rollBack(); 
                  //reset microcontroller to reboot into rolled back partition
              } else {
                  Serial.printf("Can't do rollback!...\n");
              }
          }
1 Like
3

Thank you @maxgerhardt
I applied what you suggested:

{
  Serial.printf("[%s]: \n",__func__);
  const esp_partition_t *running = esp_ota_get_running_partition();
  Serial.printf("running partition: %s\n",running->label);
  // bool diagnostic_is_ok = diagnostic();
  bool diagnostic_is_ok = false;
  if (diagnostic_is_ok) {
      Serial.printf("Diagnostics completed successfully! Continuing execution ...\n");
  } else {
      Serial.printf("Diagnostics failed! Start rollback to the previous version ...\n");
      if (Update.canRollBack()) {
          Update.rollBack(); 
      } else {
          Serial.printf("Can't do rollback!...\n");
      }
  }  
}

but calling Update.rollBack() does literally nothing:

14:22:41.069 > [check_ota_partition]: 
14:22:41.071 > running partition: app0
14:22:41.071 > Diagnostics failed! Start rollback to the previous version ...
14:22:52.190 > [check_ota_partition]: 
14:22:52.190 > running partition: app0
14:22:52.193 > Diagnostics failed! Start rollback to the previous version ...
14:23:03.324 > [check_ota_partition]: 
14:23:03.324 > running partition: app0
14:23:03.327 > Diagnostics failed! Start rollback to the previous version ...
14:23:14.443 > [check_ota_partition]: 
14:23:14.443 > running partition: app0
14:23:14.445 > Diagnostics failed! Start rollback to the previous version ...
14:23:25.581 > [check_ota_partition]: 
14:23:25.581 > running partition: app0
14:23:25.583 > Diagnostics failed! Start rollback to the previous version ...
14:23:36.739 > [check_ota_partition]: 
14:23:36.740 > running partition: app0
14:23:36.742 > Diagnostics failed! Start rollback to the previous version ...
14:23:47.902 > [check_ota_partition]: 
14:23:47.902 > running partition: app0
14:23:47.904 > Diagnostics failed! Start rollback to the previous version ...
4

This part has to be implemented after calling Update.rollBack();, otherwise it will just mark it as “the next partition to boot to”, but not do anything. You could reuse the same reset code as in the ArduinoOTA class.

1 Like
5

bloody hell - you are completely right!
I was sure that:

Update.rollBack(); 
//reset microcontroller to reboot into rolled back partition

meant: “Update.rollBack(); does the rollback INCLUDING reset” :wink:

thanks a lot

Of course now, if the diagnostic_function() is blabla, ESP32 will periodically switch partitions :wink:
So I have to think what to check to make sure is it not ambiguous