S32K344 - C40_Ip_MainInterfaceWrite only write all data if I run step by step in debugger

[FredM]

Hello,

Please I need help to understand and fix this issue. 

I write on the Flash memory sector 0x00600000U (C40_CODE_ARRAY_0_BLOCK_2_S256) custom data that I "memcpy" in a dynamic buffer and then I write the buffer variable using 

C40_Ip_MainInterfaceWrite(ulAddress, ulSize, pvSource, FLS_MASTER_ID);

The code works but only if I run it in debugger using breakpoints, I move forward step by step and I can see the whole memory sector with the correct content of the buffer variable.

If I restart the board I can see the 0x00600000U with persistent data and it's correct.

This is my issue:

But If I run the code without any breakpoint the 0x00600000U has no data written and/or fails written correct bytes values. It's breaking my mind since 3 days ago. Please can you help me to understand what I am missing?

I put a global variable failCounter and is increased only if I run the code without any breakpoint.

I tried to do these 2 previous solved questions:

- https://community.nxp.com/t5/S32K/The-S32K344-flash-reads-and-writes-are-abnormal/m-p/1735231

- https://community.nxp.com/t5/S32K/S32K344-C40-IP-Hardware-Fault-Problem/td-p/1697432

This is the CODE TO WRITE buffer variable into the FLASH MEM:

uint8_t functionToWriteOnFlash(void * buffer){

    // Unlock sector if needed

    if (STATUS_C40_IP_SECTOR_PROTECTED == C40_Ip_GetLock(FLS_SECTOR_ADDR)) {

        C40_Ip_ClearLock(FLS_SECTOR_ADDR, FLS_MASTER_ID);

    }

    // NOTE: Is this Necessary to rewrite the same previous sector ?

    eraseSector();

    size_t writeAddress = HEADER_ADDRESS;

    size_t remainingSize = totalSize;

    size_t chunkSize = 16;

    size_t currentSize = 0;

    uint8_t *bufferPtr = buffer;

    uint8_t resp;

    // Move through buffer 16 Bytes every iteration until end

    while (remainingSize > 0) {

        if (remainingSize < chunkSize && remainingSize % 16 != 0) {

            currentSize = (remainingSize > 16) ? (remainingSize - (remainingSize % 16)) : remainingSize;

        } else {

            currentSize = (remainingSize > chunkSize) ? chunkSize : remainingSize;

        }

        resp = writeNvs(writeAddress, currentSize, bufferPtr);

        writeAddress += currentSize;

        bufferPtr += currentSize;

        remainingSize -= currentSize;

        if (0 != resp) {

            return resp;

        }

    }

    if (remainingSize > 0) {

        resp = writeNvs(writeAddress, remainingSize, bufferPtr);

        if (0 != resp) {

            return resp;

        }

    }

    C40Status = C40_Ip_SetLock(FLS_SECTOR_ADDR,FLS_MASTER_ID);

    return 0;

}

static uint8_t writeNvs(size_t ulAddress, size_t ulSize, void *pvSource) {

    // WRITE DATA ONLY WORKS WHEN I DEBUG STEP BY STEP

    C40_Ip_MainInterfaceWrite(ulAddress, ulSize, pvSource, FLS_MASTER_ID);

    do {

        C40Status = C40_Ip_MainInterfaceWriteStatus();

    } while (STATUS_C40_IP_BUSY == C40Status);

    Nvs_Assert(STATUS_C40_IP_SUCCESS == C40Status);

    // Compare the data in memory with source

    // I ONLY GET failCounter = 0 IF I USE THE DEBUG STEP BY STEP

    C40Status = C40_Ip_Compare(ulAddress, ulSize, pvSource);

    if(STATUS_C40_IP_SUCCESS != C40Status){

        failCounter++;

    }

    return Nvs_Assert(STATUS_C40_IP_SUCCESS == C40Status);

}

[danielmartynek]

Hi @FredM,

There is no need to place the code to SRAM if you program another block, e.g. programming the 0x00600000 block from the 0x00400000U block.

This looks like a cache problem.

The driver synchronizes the cache if this option is set:

Regards,

Daniel