Redundant register values in GCC-compiled binary?

Question

Alright, so I'm trying to learn reverse engineering and x64 assembly. As an example I wrote this little test program in C:

#include <stdio.h>

int square(int num) {
        return num * num;
}

int main() {
        int ans = square(5);
        printf("%d", ans);
}

This results in the following assembly code for the square function:

push   rbp
mov    rbp,rsp
mov    DWORD PTR [rbp-0x4],edi
mov    eax,DWORD PTR [rbp-0x4]
imul   eax,eax
pop    rbp
ret

This seems kind of weird to me, as edi is the function argument, which I believe is then stored on the stack and loaded back into eax, where it is then multiplied? Why not skip lines 3 and 4 and just imul edi, edi?
So I opened up radare2 and did just that, but now the program returns seemingly random numbers, I'm guessing memory addresses?

Can someone explain to me why GCC uses this seemingly redundant register, and what I did wrong trying to patch the binary?

Ah, I just compiled it with default settings, output with O3 optimization makes much more sense. Thanks! — Julia van der Kris, May 19 '20 at 14:41

score 2 · Accepted Answer · answered May 19 '20 at 14:42

2

You didn't enable optimizations, so you're looking at debug code. With -O3, the output is:

One note about your question, you need to assign to eax, because eax stores the return value in an x86 application.

answered May 19 '20 at 14:42

Blindy

65,249
10
91
131

Right, that makes sense. And I found out what the problem was with my patch: `edi` is later overwritten by the other argument to printf: the pointer to the actual format string. The return value of square should be in `esi`. – Julia van der Kris May 19 '20 at 14:51
Which is fine, because `edi` is used to pass parameters. Keep in mind that `edi` and `esi` are the only registers that can run the `rep *sb` instructions, and so are good candidates for this. There are a whole bunch of conventions compilers use (almost) universally, if you really want to code assembly alongside the compiler you need to know them, or you'll generate invalid code. – Blindy May 19 '20 at 14:57
Actually programming in assembly is probably a bit ambitious for now, at the moment I'm mostly just trying to learn how to read assembly code generated by the compiler. But thanks for the help, I really appreciate it! – Julia van der Kris May 19 '20 at 15:05
1

Please don't post pictures of text. You can easily copy/paste text from Godbolt's asm output pane. – Peter Cordes May 19 '20 at 17:32
But without the color formatting. It's okay if you don't like pictures, feel free to add your own picture-free answer if it will improve everyone else's lives. – Blindy May 19 '20 at 17:36
Didn't see your reply at the time. When I copy text from Godbolt, I sometimes separate blocks with a blank line, or just comment that asm. That gives the same general value that color highlighting gave, but without being an image. – Peter Cordes Jan 03 '22 at 20:31

Redundant register values in GCC-compiled binary?

1 Answers1

Linked