Regression from RDP level 1 to 0

[Gpeti]

Hello,

I'm trying to test RDP levels configuration on STM32H753.

Here is the code I've tested (on a Nucleo 743 with STM32CubeIDE):

int main(void)
{
 HAL_Init();
 SystemClock_Config();
 MX_GPIO_Init();
 MX_USART3_UART_Init();
 
 OBInit.Banks = FLASH_BANK_1;
 HAL_FLASHEx_OBGetConfig(&OBInit);
 RdpLevel = OBInit.RDPLevel;
 
 BSP_PB_Init(BUTTON_USER,BUTTON_MODE_GPIO);
 BSP_LED_Init(LED1);
 
 if ( RdpLevel == OB_RDP_LEVEL_0 )
 {
	/* Wait for User push-button press before starting the Communication */
	while (BSP_PB_GetState(BUTTON_USER) != GPIO_PIN_SET)
	{
 
	}
	// toggle LED1 to show RDP level
	BSP_LED_Toggle(LED1);
 
	// program OB
	HAL_FLASH_OB_Unlock();
	HAL_FLASH_Unlock();
 
	OBInit.OptionType = OPTIONBYTE_RDP;
	OBInit.RDPLevel = OB_RDP_LEVEL_1;
	HAL_FLASHEx_OBProgram(&OBInit);
 
	/* Start the Option Bytes programming process */
	if (HAL_FLASH_OB_Launch() != HAL_OK)
	{
		/* User can add here some code to deal with this error */
		while (1)
		{
			BSP_LED_Toggle(LED1);
 
		}
	}
 
 }
 else
 {
 
	 /* Wait for User push-button press before starting the Communication */
		while (BSP_PB_GetState(BUTTON_USER) != GPIO_PIN_SET)
		{
 
		}
		BSP_LED_Toggle(LED2);
 
		// program OB
		HAL_FLASH_OB_Unlock();
		HAL_FLASH_Unlock();
 
		OBInit.OptionType = OPTIONBYTE_RDP;
		OBInit.RDPLevel = OB_RDP_LEVEL_0;
		HAL_FLASHEx_OBProgram(&OBInit);
 
		/* Start the Option Bytes programming process */
		if (HAL_FLASH_OB_Launch() != HAL_OK)
		{
			/* User can add here some code to deal with this error */
			while (1)
			{
				BSP_LED_Toggle(LED2);
 
			}
		}
 }
 while (1)
 {
 }
}

The switch to Level 1 works fine. I can check that I am in Level 1 for example because the debugger doesn't work anymore and the LED 2 switches on. However the regression to Level 0 doesn't seem to work. I know the Flash is supposed to be erased but I cannot do anything after the "regression": cannot connect to a debugger or to STM32CubeProgrammer, no LED switches, etc...

Do you see something wrong in the code ?

[Tomas DRESLER]

Can you please check with CubeProgrammer with the option 'connect under reset', possibly after a power cycle of the board, what's the status of Option Bytes?

[Tesla DeLorean]

Pretty sure you'd need to power cycle. Reset doesn't clear option latching, protects against hacking attacks.​

[Gpeti]

I had tried a power cycle, even if I didn't write it. But perhaps I didn't wait long enough for the Flash to be erased. What happens in case of tearing during the mass erase ? I have been working on this kind of issues and I know it can be tricky.

In the meantime I unlocked the chip through the USB bootloader and the BOOT0 pin. This time I cannot switch from level 0 to Level 1. I wonder if there isn't something wrong with my code. Step by step, I can see the register FLASH_OPTSR_PROG is correctly written, I set OPTSTART, wait for the operation to be done, I get no error message or timeout but the option byte is not written to OPTSR_CUR.

[Gpeti]

More clarifications on my previous message. I double checked at register level my code to switch from Level 0 to Level 1. After busy bit is cleared, I get no error message and when I read back OPTSR_CUR it still contains the old level 0 value.

BUT when I power off the chip and plug it to STCubeProgrammer, I can see that the chip is indeed in Level 1.

It looks like the RDP modification was taken into account in NVM but not reloaded to OPTSR_CUR. Does it make sense ?

[Gpeti]

I think I found out the problem. Looks it was due to the problem described in errata sheet: "Option byte loading can be done with the user wait-state configuration"

I changed the WS from 4 to 7 and it "fixed" the issue: I can switch from Level 0 to level 1 and then from Level 1 to Level 0 (with a power off in between).

There is still something strange: the change of RDP level seems to be taken into account immediatly when the option byte is programmed: the debug session is immediatly closed (on Keil through STLink). I thought the RDP level was active at the next POR or reset.