RAM state after reset by WatchDog

[AKYR]

Hello Nordic team,

I'm looking into solution to pass data to my application after a Reset using the RAM of my nrf52840.

Looking at the documentation, this picture troubles me: 

I'm interested into the RAM state after a WatchDog reset, and it appears that RAM should be reset on the picture. But just after there is a note telling that the RAM is never reset ? 

So I made some tests using  a non_init section as described in this bug : https://devzone.nordicsemi.com/f/nordic-q-a/34395/status-of-sram-during-an-over-the-air-update/132658#132658

After a WD reset, the data i wrote in the section are still here.

So what is the case of corrumption the note talks about ?  Is data written in a non_init section reliable affter a WD reset ? 

Thanks in advance,

Aloïs KYROU

[Heidi]

Hi!

It looks like you got lucky. The documentation states that the RAM may be corrupted, but that doesn't mean it will happen every time. So the data is not reliable after a WD reset. 

Best regards,

Heidi

[AKYR]

Hey Heidi,

Thanks for this, I didn't saw you answered !

So for the product we are designing the most probable cause of a reboot is the Watchdog, but we would like to save a few data for the next boot...

This is the only solution I see, RAM struct in .non_init section with a CRC calculation to ensure data is not corrupted, can you think of any other way to pass data to next boot after Watchdog reboot ? 

Thanks in advance !

Aloïs KYROU.

[Heidi]

Hi, you could write to FLASH instead.

I would recommend using the Flash Data Storage library in the SDK (instead of the Flash Storage library) if you're updating after every boot. Flash has an endurance of 10000 write/erase cycles. With FDS you can fill up a whole page before having to erase, so you get basic wear leveling. 

There is also an FDS example available in the SDK.

Best regards,

Heidi