Why does adding a test instruction cause a segmentation fault in GNU assembly?

Question

I'm having trouble tracking down a segmentation fault. After compiling my c program to assembly, I'm editing it and adding a few things. I added some code, including this section: ``` .SB1: call fib jmp .LBL2 ``` That part works fine. But now I want to call test and maybe jump to another label. Right now, I'm just playing with it to see if I can learn how things work (taking baby steps). So I changed the code to this: ``` .SB1: call fib test %esp, 0xfffffff jz .SB2 jmp .LBL2 .SB2: jmp .LBL2 ``` But now I get a segmentation fault. Anyone know why? If you need more information or want the code to reproduce it let me know.

Accepted Answer

In AT&T syntax, a literal needs to be prefixed with `$`. Then, also the operands need to be switched:

```
test    $0xfffffff, %esp
```

Without the `$` prefix, the assembler assumes a memory address, and accessing address `0xfffffff` is most likely out of your mapped memory, which causes the segmentation fault.

Why does adding a test instruction cause a segmentation fault in GNU assembly?

Solution

Problem