Why do I have a segfault when using cmp instruction?

Question

I am using YASM on Linux x86_64.

I am following an introduction to assembly language on the internet. I've covered basic data types. Now I'm diving into loops contruction. Good, but my first attempt is a failure. This code segfault around the use of the cmd instruction:

    segment .data
a       dd  0x01,0x02,0x03,0x04
b       dd  0x03,0x03,0x03,0x03
product dd  0x00

    segment .text
    global _start
_start:

    mov     rax,0

begin_while:
    cmp     rax,4
    jnl     end_while

    inc     rax
    jmp     begin_while
end_while:

    mov     rax,1
    xor     rbx,rbx
    int     0x80

But when I add those few lines just after the _start label, then it behaves as expected. ie no segfault.

    push    rbp
    mov     rbp,rsp
    sub     rsp,16

The book I am reading uses this construct from time to time. Not everytime. I understand that is has to do with the procedure to follow when calling a function. I guess it might be related to the libc runtime. Anyway, I do not understand why it's needed. So far with all the few simple programs I wrote (not so much) I never had to use it. Just today, now that I am using jmp .

Does someone have a good explanation about it?

Answer 1

Well, you are using the x86 interrupt call

int     0x80 

which should be changed to

syscall

for x86_64 . You are relying on x86 compatibility while using 64-bit registers. As pointed out by Jester, your code will compile and run without error on x86_64 . (I have confirmed it both ways without error on AMD64 Linux) However, the extent to which this is true for all platforms is unclear.

In writing x86_64 code, should also change:

mov     rax,1

to

mov     rax, 0x3c  ; 60 decimal

to setup the proper x86_64 exit syscall instead of relying on x86 compatibility. (in x86_64, the syscall number 1 is __NR_write 1 , see: /usr/include/asm/unistd_64.h compared to /usr/include/asm/unistd_32.h )