STM32G0 and FLASH_TYPEPROGRAM_FAST fail

[JTred.2]

Hello,

I use a Nucleo-STM32G0B1 (512KB Flash).

I would like to flash same data in fast mode, but the HAL_FLASH_Program() function ending with error. The HAL_FLASH_GetError() = 0x2A0 (PGSERR and PGAERR).

Before writing I erase some flash pages.

If I clean only ONE page the HAL_FLASH_Program() works but if I clear more, it doesn't works.

This is my test code :

void write_test(void)
{
    FLASH_EraseInitTypeDef EraseInitStruct;
    uint32_t               PageError = 0;
    uint8_t                data[256];
    uint32_t               addrFlash = 0x8040100;
    uintptr_t              addrData  = (uintptr_t) data;
 
    // Init data
    for (int i = 0; i < 256; i++)
        data[i] = i;
 
    //====== ERASE
    if (HAL_FLASH_Unlock() != HAL_OK)
    {
        ERR("Flash unlock fail");
        return;
    }
    __HAL_FLASH_CLEAR_FLAG(FLASH_FLAG_OPTVERR);
    EraseInitStruct.TypeErase = FLASH_TYPEERASE_PAGES;
    EraseInitStruct.Banks     = FLASH_BANK_2;
    EraseInitStruct.Page      = 0;
    EraseInitStruct.NbPages   = 3; // If I set it to 1, HAL_FLASH_Program() works.
    if (HAL_FLASHEx_Erase(&EraseInitStruct, &PageError) != HAL_OK)
    {
        ERR("Fail to erase data, error=%X", PageError);
        HAL_FLASH_Lock();
        return;
    }
    HAL_FLASH_Lock();
 
    //===== FLASH
    if (HAL_FLASH_Unlock() != HAL_OK)
    {
        ERR("Flash unlock fail");
        return;
    }
    __HAL_FLASH_CLEAR_FLAG(FLASH_FLAG_OPTVERR);
    if (HAL_FLASH_Program(FLASH_TYPEPROGRAM_FAST, addrFlash, addrData) != HAL_OK)
    {
        ERR("FLash write fail 0x%X, error=%X", addrFlash, HAL_FLASH_GetError());
        HAL_FLASH_Lock();
        return;
    }
    HAL_FLASH_Lock();
}

[JTred.2]

Same as this one (without official answer).

The workaround was to write in FAST mode only on the last erased page.

It also works if you perform Bank Erase.

[KDJEM.1]

Hello @Community member​ and welcome to the Community 🙂,

Can you please share the other discussion in same context and without official answer because you pointed to the same post that you shared.

I am investigating in this issue to reproduce the case on my side.

Thank you.

kaouthar

[JTred.2]

Hello @KDJEM.1​

I corrected the link.

You can use my code to reproduce the problem, or just erase a page and flash in FAST mode in another page.

[KDJEM.1]

Hello @Community member​,

Unfortunately, I don't have currently Nucleo-STM32G0B1.

For that, I tested your code with nucleo-G071RB, and I don't find any problem.

If you use the FLASH_TYPEPROGRAM_DOUBLEWORD, do you have the same issue?

Thank you.

Kaouthar

[Antoine Odonne]

Hello,

I am afraid fast mode programming only works after a page erase if you intend to write the last erased page. I provide a bit more informations from our design team and will make sure it ends up in our Reference Manual:

Because locations are not checked before being programmed, it is important to prevent SW from programming the same row multiple times.

This may be achieved by SW by either performing a mass-erase on the relevant bank(s) or by just erasing the page which is about to be programmed.

If a mass-erase was performed, then all flash rows may be programmed, with the only constraint that row address must always be incremented across different row programming operations.

If a page erased was performed, then only flash rows belonging to that page may be fast-programmed (always incrementing row addresses).

From our RM programming sequence we start with this line below, I will request to modify it in order to explicit the ocnstraint above.

"1.  Perform a mass or page erase. If not, PGSERR is set."

Thank you all and best regards,

Antoine

[JTred.2]

Hello,

Thank you for the answer. I had changed my code to delete the pages as I used it.

And yes writing in FLASH_TYPEPROGRAM_DOUBLEWORD always works.

Regards,

Jerome