nrf52832 software/firmware "lock up"

RE: nrf52832 software/firmware "lock up"

Sigurd — Thu, 04 May 2023 08:56:05 GMT

Hi,

[quote user="dlewis"]As for the watchdog timer, that is an avenue we are looking into to relieve some of the negative effects, but we still can hopefully determine the exact cause so that we can eliminate it all together. [/quote]

Assuming that the watchdog will reset the device when this "lock-up" happens, you can then use the code in this post here to save the location of the cpu program counter where the code was running when the watchdog reset happened, and then use addr2line with the .elf file to find the exact code line: Saving the location of Watchdog timeout

RE: nrf52832 software/firmware "lock up"

dlewis — Wed, 03 May 2023 17:10:52 GMT

Thanks for the response! I'd have to check to confirm that CONFIG_RESET_ON_FATAL_ERROR is set, but I believe it is because any other faults we experienced previously would automatically stop the debugger/restart the code in production configuration. As for the watchdog timer, that is an avenue we are looking into to relieve some of the negative effects, but we still can hopefully determine the exact cause so that we can eliminate it all together.

RE: nrf52832 software/firmware "lock up"

nordicator_prim3 — Tue, 02 May 2023 21:37:04 GMT

I don't have a good way for you to figure out what's actually happening but I do have a similar problem to you (only in that I have a strange bug that happens very rarely and can't recreate it). It would be great to actually find the bug or to find out whats really happening, but I've at least found a couple good things that helped me at least detect the inactivity and ensure that a reset occurs.

Do you have CONFIG_RESET_ON_FATAL_ERROR set to yes? This would make sure that if there was a hard fault then the device is sure to reset instead of hang and wait.

Also, check out using the zephyr watchdog timers. You can update this timer in a loop or based off of an interrupt from some preipheral that you expect to constantly see. If the watchdog timer doesn't get updated in a timely fashion (like if the thread that updates the timer stops running) then you can configure it to hard reset the device.

Good luck, I know these kinds of problems are annoying!