NRF_UICR->CUSTOMER read/write on 52832

[QuentinFarizon]

Hello,
I'm trying to port my code (working in production) from 52811 to 52832.

I have several crashes on a piece of code at the start of my application that check the NRF_UICR, and based on the value does something then erases it.

When I say crash, it restarts the application, without entering app_error_fault_handler 

// Check first two bytes of NRF_UICR->CUSTOMER
if ( NRF_UICR->CUSTOMER[0] != 0x0 || NRF_UICR->CUSTOMER[1] != 0x0 ) {
    
    // Read NRF_UICR->CUSTOMER[i], i from 0 to 8
    (...)
    
    //Erase first two bytes of NRF_UICR->CUSTOMER
    NRF_NVMC->CONFIG = NVMC_CONFIG_WEN_Wen;
    while (NRF_NVMC->READY == NVMC_READY_READY_Busy){}
    NRF_UICR->CUSTOMER[0] = 0x0;
    NRF_UICR->CUSTOMER[1] = 0x0;
    NRF_NVMC->CONFIG = NVMC_CONFIG_WEN_Ren;
    while (NRF_NVMC->READY == NVMC_READY_READY_Busy){}
}

 

What could be different between 52811 and 52832 on this matter ?

Best regards,
Quentin

[Kenneth]

What is the NRF_POWER->RESETREAS on startup? Make sure to clear all bits in RESETREAS after reading by writing 1 to them, else they may be kept on next startup.

Kenneth

[QuentinFarizon]

I don't have anything on my code about that. Is this is a known difference between 52811/52832 ?

[Kenneth]

I don't understand what the problem may be, but using that register will help find what the reset reason is, that will help us understand next step.

Kenneth

[QuentinFarizon]

RESEATREAS is 0x4 -> Reset from soft

[Kenneth]

Since this is not triggered from app_error_fault_handler(), then I suspect it is triggered by HardFault_process().

Try to add/change #define HARDFAULT_HANDLER_ENABLED 1 to sdk_config.h.

In addition:

Hardfault_handler_<toolchain>.c needs to be included in order for HardFault_c_handler() (by hardfault_implementation.c) to be invoked on a fault exception. Did you include one of the handler files from \components\libraries\hardfault\nrf52\handler (different versions depending on toolchain)?

Edit: The code snippet you shared seems to work on an nRF52832 I have here without any issues.

Best regards,
Kenneth

[QuentinFarizon]

On 52832 I have set CONFIG_GPIO_AS_PINRESET to 1
Could this have an impact ? 

I have added the hardfault handler but I don't see any logs from there, nor do any breakpoints break here. 

[Kenneth]

There is two reasons why the soft reset bit may be set in RESETREAS:

1. Somewhere in your project you are calling NVIC_SystemReset() or sd_nvic_SystemReset().

2. You did not clear the RESEATREAS by writing set bits to 1 on previous reset (this can be checked by reading the bits after clearing to make sure they are all 0):
https://infocenter.nordicsemi.com/topic/com.nordic.infocenter.nrf52832.ps.v1.1/power.html#register.RESETREAS 

Can you just double check if any of the above is the case here? 

QuentinFarizon said:

On 52832 I have set CONFIG_GPIO_AS_PINRESET to 1
Could this have an impact ? 

I don't see how that should affect the soft reset being set, but you can try to remove and see if that changes things.

[QuentinFarizon]

Thanks for your help.

To be sure, I've commented out all calls to NVIC_SystemReset and sd_nvic_SystemReset (including in SDK files).
I have also checked that RESEATREAS is reset by reading after resetting.
I have tried removing CONFIG_GPIO_AS_PINRESET configuration.

The reboot still occurs in all cases (always with RESETREAS 0x4).

[Kenneth]

I think there must be something we are overlooking here, can it be that you have a bootloader that is handling/doing something here? Can you share the implementation you have of the RESETREAS in the start of main()?

[QuentinFarizon]

I have the error with and without the bootloader (and my bootloader is unmodified from the examples).

Here is the RESETREAS code :

    uint32_t reset_reason = NRF_POWER->RESETREAS;
    NRF_LOG_INFO("Reset_reason: 0x%08x.\n", reset_reason);
    NRF_POWER->RESETREAS = 0xffffffff;
    reset_reason = NRF_POWER->RESETREAS;
    NRF_LOG_INFO("Reset_reason resetted: 0x%08x.\n", reset_reason);