RM57L843: Errors inside HL_sys_startup.c

Hi Koichiro,

Would you check if there is any possibility of customer's code reading a variable without first initializing it? Do they perform an auto-initialization of the CPU RAM before accessing it?

Which compiler is used? 

Hi QJ,

It is not quite clear what below question exactly means.

QJ Wang said:

Would you check if there is any possibility of customer's code reading a variable without first initializing it? Do they perform an auto-initialization of the CPU RAM before accessing it?

At the beginning  of HL_sys_startup.c, RAM is accessed by switch(resetReason).
Is this problem?

Thanks and regards,
Koichiro Tashiro

Hi Koichiro,

Can you please declare resetReason as register?

Hi QJ,

Could you explain what to do more in detail?
resetReason is declared as below in HL_sys_startup.c line#88.

SL_ResetReason     resetReason;        /* Reset reason */

I do not know what is "SL_ResetReason"

QJ Wang said:

Can you please declare resetReason as register?

Could you specify how to do that?

Thanks and regards,
Koichiro Tashiro

QJ,

You mentioned about RAM initialization. Please tell us how to initialize RAM to avoid the issue?

Thanks and regards,
Koichiro Tashiro

Hi Koichiro,

I am not able to reproduce this issue. I'd like to check if the following methods can resolve the issue:

1. Initialize the MCU SRAM before checking the reset resource:

    SL_Init_Memory(RAMTYPE_RAM);

   _SL_Init_EnableEventExport();

   resetReason = SL_Init_ResetReason();

   switch(resetReason)
   {

       .....

   }

2. Declare the resetReason as register. The keyword 'register' means that the variable is stored in a processor register instead of memory.

    register resetSource_t   rstSrc;
    rstSrc = SL_Init_ResetReason();

   switch(rstSrc)
   {
        case ...

The start-up sequence in the Safety Library demo app does need to be changed.

My recommendation is :

void _c_int00(void)

{

    SL_Init_R5Registers();

    SL_Init_StackPointers();

    SL_Init_Memory(RAMTYPE_RAM);

    _SL_Init_EnableEventExport();

    resetReason = SL_Init_ResetReason();

...

In the original sequence, _SL_Init_EnableEventExport() call may cause the double-bit ECC error. The first instruction in this function is a store to the stack, which will cause the double-bit ECC error. However since the event-bus export is not enabled at this point, there is no response or indication for this ECC error.

The function call resetReason = SL_Init_ResetReason(); can also cause 2-bit ECC error. This is a C function, which also uses the stack to save and restore CPU registers. The double-bit ECC error caused by this stack write is now exported to the ESM. This is the cause of the ESM group 3 error that you see on power-up.

Please call the memory initialization function before the call to enable event export. This memory initialization function in the demo app is not using stack operations as the return value is defined as a "register", so it uses a CPU register to hold this variable.

Hi QJ,

Thanks. Customer will try your last recommendation.
One question:

QJ Wang said:

This memory initialization function in the demo app is not using stack operations as the return value is defined as a "register", so it uses a CPU register to hold this variable.

As I said, resetReason is declared as below in the demo code. SL_ResetReason is enum, not "register".

SL_ResetReason     resetReason;        /* Reset reason */

Is this OK to leave this as-is?
No need to apply below changes you mentioned before?

QJ Wang said:

    register resetSource_t   rstSrc;
    rstSrc = SL_Init_ResetReason();

   switch(rstSrc)
   {
        case ...

Thanks and regards,
Koichiro Tashiro

Koichiro Tashiro said:

Is this OK to leave this as-is?
No need to apply below changes you mentioned before?

If the MCU SRAM is initialized before resetReason = SL_Init_ResetReason();, you can use it as-is.

Or 

you can declare resetReason as register 

But I suggest to init SRAM before any read/write from/to the memory location.

QJ,

Customer tried below changes, but it does not help.

QJ Wang said:

void _c_int00(void)

{

    SL_Init_R5Registers();

    SL_Init_StackPointers();

    SL_Init_Memory(RAMTYPE_RAM);

    _SL_Init_EnableEventExport();

    resetReason = SL_Init_ResetReason();

...

Did you try above changes worked on your side?
Again, customer uses TI hardware(HDK) and software (HL_sys_startup.c), so it should work without an issue.
Is there any missing initialization, etc.?

Thanks and regards,
Koichiro Tashiro

Hi QJ,

Customer investigated it deeper and found the code does not return from SL_InitMemory(RAMTYPE_RAM); and it stops while loop at sl_init.c line#172.

	/*SAFETYMCUSW 440 S MR: 11.3 <APPROVED> Comment_5*/
	/*SAFETYMCUSW 9 S MR: 12.2 <APPROVED> Comment_2*/
    /*SAFETYMCUSW 28 D <APPROVED> Comment_6*/
    /*SAFETYMCUSW 134 S MR: 12.2 <APPROVED> Comment_3*/
    while (MEM_HW_INIT_COMPLETE != BF_GET(sl_systemREG1->MSTCGSTAT, BF_MINIDONE_START, BF_MINIDONE_LENGTH)) {

    }

Tried many time, there is no problem to run RAMTYPE_RAM auto-init:

SL_InitMemory(RAMTYPE_RAM);

Hi QJ,

I see. Let me check with customer again.

Thanks and regards,
Koichiro Tashiro