Concerto M3 seems have problem with watchdog timer or CMD file, it stuck in the code before main()

I use CCS 6.1.1 to develop my concerto device F25M35. One day I added some IPC messageQ code into my existing program some strange things occured. 

The project build successfully with no errors and warnings. But when I load it with XDS100V3 debugger. The CCS's run and stop button shows gray and the program never get to main(). Then I set change the debug configuration to something 'not run to main at program load or reset'. Then I step in the assembly code. I find that when I get to the ti_catalog_arm_cortexm3_concertoInit_Boot_initSharedRAMs(); the procedure fails.

Here is my configuration.

And here is where the program fails.

So I thought there maybe something wrong with my CMD file. Here is my CMD file. I have modified the shared memory.

MEMORY
{
    BOOTROM (RX)    : origin = 0x0,        length = 0x10000
    FLASH_BOOT (RWX): origin = 0x200030,   length = 0x4
    FLASH (RWX)     : origin = 0x200034,   length = 0x7FF9C
    C03SRAM (RWX)   : origin = 0x20000000, length = 0x08000
    S03SHRAM (RWX)  : origin = 0x20008000, length = 0x08000
    S47SHRAM (RWX)  : origin = 0x20010000, length = 0x08000
    CTOMMSGRAM (R)  : origin = 0x2007F000, length = 0x800
    MTOCMSGRAM (RW) : origin = 0x2007F800, length = 0x800
}

SECTIONS
{
    /* Allocate program areas: */
    .text       : > FLASH
    .binit      : > FLASH
    .cinit      : > FLASHwhe
    .pinit      : > FLASH
    .init_array : > FLASH
    .intvecs    : > FLASH
    .resetisr   : > FLASH
    .vtable     : > FLASH
    /* Initialized sections go in Flash */
    .const      : > FLASH

    /* Allocate uninitalized data sections: */

    .data       : > C03SRAM
    .bss        : >> C03SRAM | S47SHRAM
    .dma        : > C03SRAM
    .sysmem     : > C03SRAM
    .stack      : > C03SRAM
    .cio        : > C03SRAM
    .neardata   : > C03SRAM
    .rodata     : > C03SRAM
    .args       : > C03SRAM

    GROUP : > MTOCMSGRAM
    {
        PUTBUFFER
        PUTWRITEIDX
        GETREADIDX
    }

    GROUP : > CTOMMSGRAM
    {
        GETBUFFER : TYPE = DSECT
        GETWRITEIDX : TYPE = DSECT
        PUTREADIDX : TYPE = DSECT
    }

}

__STACK_TOP = __stack + 256;

Here is my memory allocation. I got to much .bss section.

Could anyone tell me how could I solve this problem?

Regards.