TMS320F28379S: TMS320F28379S: Cannot execute program after execution with different entry point

BTW, here are the codes for F2837xS_CodeStartBranch.asm

;//###########################################################################

;// $TI Release: F2837xS Support Library v210 $

;// $Release Date: Tue Nov  1 15:35:23 CDT 2016 $

;// $Copyright: Copyright (C) 2014-2016 Texas Instruments Incorporated -

;//             http://www.ti.com/ ALL RIGHTS RESERVED $

;//###########################################################################

***********************************************************************

WD_DISABLE  .set  1    ;set to 1 to disable WD, else set to 0

    .ref _c_int00

    .global code_start

***********************************************************************

* Function: codestart section

*

* Description: Branch to code starting point

***********************************************************************

    .sect "codestart"

code_start:

    .if WD_DISABLE == 1

        LB wd_disable       ;Branch to watchdog disable code

    .else

        LB _c_int00         ;Branch to start of boot._asm in RTS library

    .endif

;end codestart section

***********************************************************************

* Function: wd_disable

*

* Description: Disables the watchdog timer

***********************************************************************

    .if WD_DISABLE == 1

    .text

wd_disable:

    SETC OBJMODE        ;Set OBJMODE for 28x object code

    EALLOW              ;Enable EALLOW protected register access

    MOVZ DP, #7029h>>6  ;Set data page for WDCR register

    MOV @7029h, #0068h  ;Set WDDIS bit in WDCR to disable WD

    EDIS                ;Disable EALLOW protected register access

    LB _c_int00         ;Branch to start of boot._asm in RTS library

    .endif

;end wd_disable

    .end

Hi,

Yagamy Huang said:

Furthermore, I follow up previous thread to disable Watch dog in /Applications/ti/controlSuite/device_support/F2837xS/v210/F2837xS_common/source/F2837xS_CodeStartBranch.asm, then the address of entry point is still 0x00094a4f, but it can boot successfully (after powered off and on).

Once you disable the WD, code always works. Is that correct ? If yes then we suggest to keep the WD disable and enable in your main application and make sure to service it with in it's TIMEOUT period. 

Regards,

Vivek Singh

Hi Vivek,

Thanks for feedback. We will try your suggested way to enable WD in our main application.

But, be curious, why different address (especially odd number of address) can make WD timeout? All of those codes (F2837xS_CodeStartBranch.asm, boot28.asm) are from TI CodeSuite. We didn't modify them... Can you share more insights about odd address issue for entry point?

My wild guess is that boot28.asm tries to initialize memory map, interrupt vector based on the address of entry point, and some of the memcpy/memset operations need to be word aligned, 16 bits... Then, odd address would make mis-alignment issue, and then jump to wrong address of main()...

Br,

Yagamy

Hi,

Yagamy Huang said:

But, be curious, why different address (especially odd number of address) can make WD timeout?

May be I misunderstood this. I thought even with same odd address as entry point, if you disable the WD, it works fine. Is that not the case ? 

Regards,

Vivek Singh

Your understandings are correct. Let me organize what we've tested:

1. WD disabled, odd address as entry point => OK
2. WD disabled, even address as entry point => OK
3. WD enabled, odd address as entry point => NG
4. WD enabled, even address as entry point => OK

My question is that, why 4 can work but 3 cannot work. In my codes, when I just add one line of function call to GPIO_WritePin, that increases that object file 4 bytes, then it makes linker to rearrange object files and finally place EntryPoint _c__init00() to an odd address.

Although I can always disable WD in F2837xS_CodeStartBranch.asm to make situations only 1 and 2, but I still want to learn more why 4 can work but 3 cannot work because we need to make technical explanations to our customer to ease their concerns to disable WD in F2837xS_CodeStartBranch.asm.

They think, disabling WD is workaround, and there must be some side-effects if we don't know the true root-cause. They can accept the workaround if the root-cause is one of following reasons:

• odd number as entry point address, as Errata of F28379S
• known issue in toolchain, either assembler, linker, or compiler
• known issue in ControlSuite

Br,

Yagamy

I don't think it has anything related to entry point because if that is causing issue then WD disable will not help. I think it's because of addition of the code which is causing WD to timeout. My guess is that once you add that line, for some reason it's increasing the execution time before WD get disable and that is enough for WD to time out. You can profile you code execution from entryPoint _c__init00() to function in main() which disables the WD in both case using CCS and see if there is any difference. 

Regards,

Vivek Singh

Could you please guide me how to profile code execution from entryPoint to main()?

BTW, our program also disables WD at very early stage of main() execution. It looks like:

void main() {

    InitSysCtrl(); /* from F2837xS_SysCtrl.c */

    ...

}

void InitSysCtrl(void) {
    //
    // Disable the watchdog
    //
    DisableDog();
    ...
}

You can see watch dog is disabled at very early stage since executing main(). The implementation of _c_init00() is from TI. We didn't modify it. Why does the execution of _c_init00() take longer time to trigger WD when we just increase 4 bytes (by adding a function call) in the object files?