C printf与汇编代码混合时出现分段错误（核心已转储）

Question

I tried some basic example of mixing C code and x86 Assembly code. It is simple Find smaller number example.

Here is Assembly code in bs file (I have to use intel syntax with no prefix and i386 architecture):

.intel_syntax noprefix
.arch i386

.data
    .globl out_format
    out_format:
        .asciz "min(%i, %i) = %i\n"

.text
    .globl min
    .type min, @function
    min:
        enter 0, 0
        mov ebx, [ebp + 8]
        cmp ebx, [ebp + 12]
        jle 1f
        mov eax, [ebp] + 12
        jmp 2f
        1:
            mov eax, ebx
        2:
            leave
        ret 8

.end

Here is the C code in ac file:

#include<stdio.h>

extern char out_format;

extern int min(int x, int y);

int main() {
    int x1, x2, r;
    scanf("%i%i", &x1, &x2);
    r = min(x1, x2);
    printf(&out_format, x1, x2, r);

    return 0;
}

I compile it with gcc on x64 Arch Linux:

gcc -m32 -Wall -Werror -g -o p b.s a.c

But when I start the executable file p, and insert 2 numbers and hit enter I get:

Segmentation fault (core dumped)

I researched all previous similar questions, and I tried several things but every time I get the same error. I also tried to debug it with gdb, and I see that Assembly function min returns the right result, but somehow it always stuck on C printf function.

I tried to not import string out_format from Assembly file, and define it directly in C file (I defined it as simple array of chars, and also as char pointer using malloc and free functions) but it didn't help.

Otherwise the program doesn't give such error if I intialize variable r with some integer instead with result of Assembly function min . But it doesn't have any sense to me as I already checked it with gdb and the Assembly function correctly intialize the variable r with correct smaller number.

Please, can anyone help me with this?

Thanks in advance

Answer 1

if min() works good, check out_format.

following code works.

printf("%d %d %d\n", x1, x2, r);

full code

#include<stdio.h>

int min(int x, int y)
{
    if (x<y) return x;
    return y;
}

int main() {
    int x1, x2, r;
    scanf("%i%i", &x1, &x2);
    r = min(x1, x2);
    printf("%d %d %d\n", x1, x2, r);
    return 0;
}

Answer 2

At the end I just want to leave the code that works.

File bs:

.intel_syntax noprefix
.arch i386

.data
    .globl out_format
    out_format:
        .asciz "min(%i, %i) = %i\n"

.text
    .globl min
    .type min, @function
    min:
        enter 0, 0
        push ebx
        mov ebx, [ebp + 8]
        cmp ebx, [ebp + 12]
        jle 1f
        mov eax, [ebp] + 12
        jmp 2f
        1:
            mov eax, ebx
        2:
            pop ebx
            leave
        ret 8

.end

File ac:

#include<stdio.h>

extern char out_format;

extern int __attribute__((stdcall)) min(int x, int y);

int main() {
    int x1, x2, r;
    scanf("%i%i", &x1, &x2);
    r = min(x1, x2);
    printf(&out_format, x1, x2, r);

    return 0;
}

Thanks everyone for their help.

Answer 3

The asm uses ret 8 to pop the args from the stack as it returns.

The standard 32-bit x86 Linux calling convention ( i386 System V ABI ) is caller-pops. Functions just use ret to return and the caller can clean up the stack if / when it wants. (eg use mov stores to write new args before another call instead of adjusting esp and using push again.)

As the OP's self answer points you, you could instead declare min as

extern int __attribute__((stdcall)) min(int x, int y);

so the compiler-generated code will agree with the asm on the calling convention. This calling convention is sometimes used on Windows.

Single-stepping with a debugger would have revealed this (if you knew what to look for). Always use a debugger to figure out where your code segfaults, then look at registers and work backwards to find out why.