Feed external hardware watchdog on Heltec Tower V2 (aka MeshTower V2)

[beala]

Like clockwork, my MeshTower V2 restarts approximately every 9 minutes and 30 seconds. The variant.h for this device gives a clue as to why:

MeshCore/variants/heltec_tower_v2/variant.h

Lines 85 to 88 in fec88e1

	#define HAS_HARDWARE_WATCHDOG
	#define HARDWARE_WATCHDOG_DONE (0 + 9)
	#define HARDWARE_WATCHDOG_WAKE (0 + 10)
	#define HARDWARE_WATCHDOG_TIMEOUT_MS (8 * 60 * 1000)

Both pins and a timeout for a watchdog are defined, but never used. After one of these reboots, I can see this over the serial port, which is consistent with an external watchdog device resetting the MCU.

DEBUG: PWRMGT: Reset = Reset Pin (0x1)

This feeds the doggo and toggles the DONE pin every 60 seconds inside of a SoftwareTimer to prevent the watchdog from resetting the device.

It would probably be better to feed the watchdog inside the main.cpp but there's no existing hook for this as far as I can tell. Simply spinning up a timer for this specific board and making no big architecture changes seemed like the easier and less controversial change. Let me know if you have thoughts.

Testing

I flashed this to my device and verified the device hasn't rebooted with an uptime of 13 minutes, an uptime never before reached on this device.

If you'd like to test this yourself, here are build instructions for a repeater:

# Checkout my PR
git fetch upstream pull/2904/head:pr-2904
git checkout pr-2904

# Build with a custom version string and serial debug logging turned on
PLATFORMIO_BUILD_FLAGS='-D FIRMWARE_VERSION=\"v1.16.0-watchdog-fix\" -D MESH_DEBUG=1' pio run -e Heltec_tower_v2_repeater

Head over to https://meshcore.io/flasher and use the "Custom firmware" option to flash .pio/build/Heltec_tower_v2_repeater/firmware.zip

If there are issues, you can run ver on the repeater CLI to verify v1.16.0-watchdog-fix is the version string. That verifies that you flashed your build.

You can connect to the repeater over serial and look for something like DEBUG: PWRMGT: Fed hardware watchdog (t=541020 ms) to verify that the fix is actually running. How to connect is platform specific. On Linux, you can use screen:

# Substitute in the correct /dev/... path.
screen /dev/serial/by-id/usb-Heltec_HT-n5262_5F3611E4949BEF11-if00 115200

cc @Quency-D who PR'd the original support for this device. #2752

[beala]

It looks like there's an unmerged PR for V1 of this device (#1446) to solve this same issue. Time is a flat circle. Unfortunately, it doesn't look like merging V1's PR will fix V2 because the #defines are named slightly differently.

So merging this PR is still useful, but I defer to Heltec on how they want to fix this. I just want my device to work :)

[beala]

Thanks for taking a look @Quency-D! Can I interpret the thumbs up reactji as a stamp of approval from Heltec?

If not, I'm happy to hand over the reigns. Just open another PR and I'll close this one.

[Quency-D]

Thanks for taking a look @Quency-D! Can I interpret the thumbs up reactji as a stamp of approval from Heltec?

If not, I'm happy to hand over the reigns. Just open another PR and I'll close this one.

This is a good quick solution. However, adding it to the main loop would be a better way to utilize the watchdog timer's full capabilities.

[stilldavid]

I have received my MeshTower V2 and ran into this issue as well. I can confirm that this branch does fix the watchdog reset - uptime is 12+ hours and counting on my unit.