S32K142 SRAM ECC problem

[davidjiang1973]

Hello,

My project needs to capture the SRAM non-correctable ECC event when the SRAM is corrupt for function safety. But now I have two problems when I use ERM to implement it. 

My solution is as follows:

1. Enable the ERM CH0 to capture the  SRAM_L non-correctable ECC event;

2. Enable the ERM CH1 to capture the  SRAM_U non-correctable ECC event;

3. The vector table is placed in the beginning of SRAM_L(i.e. __flash_vector_table__ is not defined) and some

     time-critical codes also run in SRAM_L.  All the variables are placed in SRAM_U;  

4. The EIM is also used to confirm that the ERM is working properly by calling the following function

    ERM_CheckByEIM() every 100ms  in which the single error is injected into SRAM_L & SRAM_U by EIM.

5. SCST_Test() from SCST library is called every 100ms;

The problems are as follows:

1. Problem 1: The ERM reports that SRAM_L generates the non-correctable ECC event, but SRAM_U is OK;

2. Problem 2: If the ERM CH0 is not enabled for SRAM_L non-correctable ECC event capture and RCM is enabled to delay the s/w reset(The register bits  'SIM->CHIPCTL.SRAML_RETEN/SRAMU_RETEN' have been set in SystemInit() after MCU reset and cleared in the ISR of RCM), the ERM reports that  the SRAM_U has the non-correctable ECC event too.

So I'm confused which causes the problems. 

/*------------------------------------------------------------------------------------------------------------------------*/

SCST_PLACE_IN_SECTION(safety_func_code) // This function is placed in SRAM_L to run.

void ERM_CheckByEIM(void)
{
 uint32_t *p;
 volatile uint32_t wReadData, wErrAddr;
 bool bTest1OK = false, bTest2OK = false;
 
 /*--- Read the the SRAM_L/SRAM_U to see the error. ---*/
 __disable_irq();
 
  /* Initialize the EIM for the single error injection. */
 EIM_DRV_Init(INST_EIM0, EIM_CHANNEL_COUNT0, eim1_ChannelConfig0);
 
 /* Clear the single error flag of CH0/1 */
 ERM_DRV_ClearEvent(INST_ERM0, 0, ERM_EVENT_SINGLE_BIT);
 ERM_DRV_ClearEvent(INST_ERM0, 1, ERM_EVENT_SINGLE_BIT);
 
 /* Read the SRAM_L twice to detect the sigle error by ERM. */
 p = (uint32_t *)0x1FFFFF00;
 wReadData = *p;
 wReadData = *p;
 /* Check the ERM sigle bit error flag to valiate that the ERM is working properly. */
 if (ERM_DRV_GetErrorDetail(INST_ERM0, 0, (uint32_t *)&wErrAddr) == ERM_EVENT_SINGLE_BIT)
 {
  if (wErrAddr == 0x1FFFFF00) bTest1OK = true;
 }
 
 /* Read the SRAM_U twice to detect the sigle error by ERM. */
 p = (uint32_t *)0x20000000;
 wReadData = *p;
 wReadData = *p;
 /* Check the ERM sigle bit error flag to valiate that the ERM is working properly. */
 if (ERM_DRV_GetErrorDetail(INST_ERM0, 1, (uint32_t *)&wErrAddr) == ERM_EVENT_SINGLE_BIT)
 {
  if (wErrAddr == 0x20000000) bTest2OK = true;
 } 
 
  /* Clear the single error flag of CH0/1 */
 ERM_DRV_ClearEvent(INST_ERM0, 0, ERM_EVENT_SINGLE_BIT);
 ERM_DRV_ClearEvent(INST_ERM0, 1, ERM_EVENT_SINGLE_BIT);
 
 __enable_irq();
 /*----------------------------------------------------*/
 
 /* Disable the EIM. */
 EIM_DRV_Deinit(INST_EIM0);
 
 /* Reset the MCU if the ERM is not working properly. */
 if ((bTest1OK == false) || (bTest2OK == false)) NVIC_SystemReset();   /* S/W reset */
}

/*------------------------------------------------------------------------------------------------------------------------*/

Thanks & BRs,

David

[davidjiang1973]

Hello Daniel,

Thanks a lot for your reply.

I tried your suggestion to place the code of 'ERM_CheckByEIM()' into Flash, but the problem still happened.

It's same even I didn't inject any error by EIM without calling the function ERM_CheckByEIM()''.

The user manual of S32K142 doesn't talk much about the ECC. So could you list all the factors which can trigger the ECC error of SRAM?

Thanks & BRs,

David

[danielmartynek]

Hello David,

ECC errors are detected when the SRAM is read and the data do not match the ECC checksum which is generated and stored while the SRAM is written.

Could you please check the EIM registers, it the injection really disabled?

At which point do you get the SRAM_L ECC error then?

Thanks,

BR, Daniel

[danielmartynek]

Hello David,

If I understand, you inject an ECC error on SRAM_L region and the function that does that is located in the SRAM_L region. I guess that the SRAM_L error is detected as soon as the global function is enabled.

Could you place the code into Flash?

Thanks,

BR, Daniel