Having trouble with nvs subsystem on same54 processor. (Using the same54_xpro eval board)

[CarloATX]

Hi everyone, I'm running into issues with storage on the same54_xpro eval board. In my personal program I require to have some data be stored on the board through power cycles and what I find is that after I updated from zephyr v3.5 to v3.7, nvs storage is failing to persist between power cycles.

When I run this program zephyr/samples/subsys/nvs/ on the eval board I get multiple warnings indicating issues with the zephyr/drivers/flash/flash_sam0.c driver:

As you can see, the program does not find the data that was stored in the initial run continuing to "store" the info and rebooting the eval board expecting to see it on the next run.

I noticed there was a commit in v3.6 that added some changes This pr .

This is the only pr in recent history between v3.5 and v3.7 that I can point to that caused this behavior but I have not fully confirmed this on my end. I just know that if I run the sample nvs program, it fails on the eval board but if I run the test suite for flash/common, it passes.

I hope someone can provide guidance/clarity on this matter. Thank you for any assistance.

[Laczen]

Hi @CarloATX, the log shows warnings that it is trying to access flash outside of the available flash. This can only be due to a length error, it would be good to slightly modify the warn message in flash_sam0.c at line 106 to also report the length that was requested.

[CarloATX]

Yes, I made sure to delete the memory before running again.

[Laczen]

I see that you have enabled 1 byte write mode. Is it the same for the standard 8 byte write mode?

[CarloATX]

I actually haven't explicitly enabled the 1 byte write mode. Is there a specific config I can use to change the write mode to 8 byte or should I be modifying the program directly somehow?

[Laczen]

That is strange, in a previous log it was showing a write block size of 8, now 1. According to the flash driver Kconfig it should default to 1 for certain devices. It should not change between compiles.
You can disable the option by setting CONFIG_SOC_FLASH_SAM0_EMULATE_BYTE_PAGES=n.

[CarloATX]

Hmm. There might be one thing that comes to mind that could cause this. There is this config CONFIG_SOC_FLASH_SAM0_EMULATE_BYTE_PAGES=y that is required for the sam0 soc and this might be modifying the write length. I may have added this in between compiles. as I was testing different combinations for building. I have to rebuild and confirm that this is the case however, it will take me a moment. right now I am building the nvs sample program on zephyr v4.0.0 but with the flash_sam0.c pr, reverted

[CarloATX]

That is strange, in a previous log it was showing a write block size of 8, now 1. According to the flash driver Kconfig it should default to 1 for certain devices. It should not change between compiles. You can disable the option by setting CONFIG_SOC_FLASH_SAM0_EMULATE_BYTE_PAGES=n.

It looks like the issue was most likely an error on my part although I was sure I was having this issue before even setting the emulate byte pages flag. I am now not having issues with the nvs subsystem and my personal code is working as intended.

setting CONFIG_SOC_FLASH_SAM0_EMULATE_BYTE_PAGES=n as you suggested resolved the issue. I am not sure why I did not notice this issue. I was so sure... 🤦‍♂️

[CarloATX]

Thank you so much for your help with this matter @Laczen
This is greatly appreciated that it was easily resolved with one config

[CarloATX]

on a side note, what is the emulate config used for now? Do you happen to know if there is any conflict with building mcuboot with this config and running an app that is built without this config?

Thanks again