GNU assembler Directives in X-Loader start.s

Question

I'm analysing the TI X-Loader code to get a deeper understanding about the omap cortex boot sequence.

I started with the start.S file in the /cpu/omap3 folder. The first lines are as follow:

#include <config.h>
#include <asm/arch/cpu.h>

.globl _start
_start: 
    b   reset
    ldr pc, _hang
    ldr pc, _hang
    ldr pc, _hang
    ldr pc, _hang
    ldr pc, _hang
    ldr pc, _hang
    ldr pc, _hang

_hang:
    .word do_hang

    .word 0x12345678
    .word 0x12345678
    .word 0x12345678
    .word 0x12345678
    .word 0x12345678
    .word 0x12345678
    .word 0x12345678 /* now 16*4=64 */l

.global _end_vect
_end_vect:

The exception vectors (except the Reset Vector) will jump to the address "_hang". But there is no op-code there, but only the address of the label do_hang (found later in the code and which is a dead loop function). My understanding of .word is that it copies the value (in our case the address of do_hang) into that specific location in the binary, in our case at the address "_hang". When the PC is loaded with _hang, program execution jumps to address _hang and there it doesn't finds an op-code but a value which is an address (do_hang).

My questions:

• Shouldn't there be the command "b do_hang" instead of ".word do_hang"?
• When the program counter points to a location which doesn't contains an opcode but an address shouldn't this cause the system to crash?
• Shouldn't .word directive make sense only in the .data section?

Thanks in advance for your help.

Martin

Answer 1

Ok! I have got it,

the command is LDR PC _hang! It doesn't jumps to _hang but it loads the value in _hang into PC. The value in _hang is the address of the dead loop function do_hang.

One question still remains: why not simply "b do_hang" in each vector? Well, I guess we will never know.

Regards

Martin