STM32U575VGT HAL_FLASHEx_Erase() not working (page is not erased)

[Adriano Melis]

Hello all,

this code snippet is not working: the page is not erased, although the HAL does not report errors.

No RTOS is running.

 

if (HAL_ICACHE_Disable() != HAL_OK) {

Error_Handler();

}

 

status = HAL_FLASH_Unlock();

if (status != HAL_OK) {

Error_Handler();

}

 

erase_init_struct.TypeErase = FLASH_TYPEERASE_PAGES;

erase_init_struct.Banks = FLASH_BANK_1;

erase_init_struct.Page = 127;

erase_init_struct.NbPages = 1;

 

status = HAL_FLASHEx_Erase(&erase_init_struct, &page_error);

if (status != HAL_OK) {

Error_Handler();

}

 

Any suggestion?

 

 

 

[TDK]

How are you verifying the page is not erased? What tool and memory addresses are you looking at?

[Adriano Melis]

I am monitoring the page content using the Memory renderer of STM32Cube Ide, there are values different from the erased value 0xFFFFFFFF

[Adriano Melis]

 

[Sarra.S]

Hello @Adriano Melis 

The page 127 in bank1 is in 0x080F E000 - 0x080F FFFF 

the screenshot you showed, does not show page 127 

 

 

[Adriano Melis]

 

[Adriano Melis]

my bad. this is an actual screenshot: Flash erase has been executed, FLASH content is unchanged.

(status value is HAL_OK)

 

[Sarra.S]

Hello @Adriano Melis, 

I tried to reproduce your issue using the FLASH_EraseProgram example in STM32U5CubeFW, however, page 127 was erased successfully and I couldn't reproduce it 

Maybe share a minor project to reproduce it on my side. 

[Adriano Melis]

hello, these are additional info:

• the MCU is STM32U575VGT6 on a proprietary board clocked at 100 MHz: are there known problems on this particular device? The same code is OK on the Nucleo STM32U575ZI-Q
• FLASH erase on a single page works fine from Cube Programmer
• when the page is pre-erased with Cube Programmer writing data is successful

[Sarra.S]

Hello @Adriano Melis 

I suppose you're setting Flash latency to 3WS? 

[Adriano Melis]

Attached two cube projects, mcb-fw-FLASH-IAP is for U575VGT (NOT working), IAP is for U575ZI (working on the Nucleo board).

As you can see the code is the same.

 

[Adriano Melis]

Hello all,

I found the source of the problem: the DUALBANK bit in FLASH_OPTR was set to 1.

On this device this arranges the FLASH page access in two banks of 64 pages each.