vRC Oct 15 Torch cannot exit firmware passthrough

[nseidle]

The technique to allow firmware to be loaded onto the UM980 (Torch) works fine on v2.2. On d2.3 Oct-15, exiting passthrough mode causes reset:

The file creation and checking was reworked from the original.

[PaulZC]

I suspect we have two things trying to access LittleFS at the same time?

I did test the the reworked um980FirmwareRemoveUpdate on Torch and I didn't see any badness at the time. But that was around August 15th and many things have changed since then.

OK. With fresh eyes, allowing a task access to LittleFS to delete a file is potentially dangerous, if other loop code could already be accessing LittleFS.

Is there a scenario where loop code has called e.g. recordSystemSettingsToFileLFS, and the task calls um980FirmwareRemoveUpdate while the settings are still being recorded?

createSettingsString can also access LittleFS to read the station coordinates. updateSettingWithValue can also access LittleFS to write the station coordinates.

I can think of two solutions. Either add a semaphore for LittleFS, or let the loop do the LittleFS file removal and reboot. A semaphore would be good, just in case we encounter the same two-things-accessing-LittleFS situation again in the future. But letting the loop do the remove and reboot is cleaner.

I'll work on a PR and test on Torch.

[PaulZC]

Ah! Hang on a minute...

In bluetoothCommandTask you're calling processCommand. processCommand has plenty of scope to access LittleFS on Torch. Could that be it? Did you have Bluetooth connected at the same time when the crash happened?

Maybe we do need that semaphore!?

[PaulZC]

Changes made in #764

But I suspect a LittleFS semaphore would be a good thing to have...

[nseidle]

Thanks for the fix @PaulZC!