I want to understand AFL's code instrumentation in detail.
Compiling a sample program sample.c
int main(int argc, char **argv) {
int ret = 0;
if(argc > 1) {
ret = 7;
} else {
ret = 12;
}
return ret;
}
with gcc -c -o obj/sample-gcc.o src/sample.c
and afl-gcc -c -o obj/sample-afl-gcc.o src/sample.c
and disassembling both with objdump -d
leads to different Assembly code:
[GCC]
0000000000000000 <main>:
0: f3 0f 1e fa endbr64
4: 55 push %rbp
5: 48 89 e5 mov %rsp,%rbp
8: 89 7d ec mov %edi,-0x14(%rbp)
b: 48 89 75 e0 mov %rsi,-0x20(%rbp)
f: c7 45 fc 00 00 00 00 movl $0x0,-0x4(%rbp)
16: 83 7d ec 01 cmpl $0x1,-0x14(%rbp)
1a: 7e 09 jle 25 <main+0x25>
1c: c7 45 fc 07 00 00 00 movl $0x7,-0x4(%rbp)
23: eb 07 jmp 2c <main+0x2c>
25: c7 45 fc 0c 00 00 00 movl $0xc,-0x4(%rbp)
2c: 8b 45 fc mov -0x4(%rbp),%eax
2f: 5d pop %rbp
30: c3 retq
[AFL-GCC]
0000000000000000 <main>:
0: 48 8d a4 24 68 ff ff lea -0x98(%rsp),%rsp
7: ff
8: 48 89 14 24 mov %rdx,(%rsp)
c: 48 89 4c 24 08 mov %rcx,0x8(%rsp)
11: 48 89 44 24 10 mov %rax,0x10(%rsp)
16: 48 c7 c1 0e ff 00 00 mov $0xff0e,%rcx
1d: e8 00 00 00 00 callq 22 <main+0x22>
22: 48 8b 44 24 10 mov 0x10(%rsp),%rax
27: 48 8b 4c 24 08 mov 0x8(%rsp),%rcx
2c: 48 8b 14 24 mov (%rsp),%rdx
30: 48 8d a4 24 98 00 00 lea 0x98(%rsp),%rsp
37: 00
38: f3 0f 1e fa endbr64
3c: 31 c0 xor %eax,%eax
3e: 83 ff 01 cmp $0x1,%edi
41: 0f 9e c0 setle %al
44: 8d 44 80 07 lea 0x7(%rax,%rax,4),%eax
48: c3 retq
lea
until 0x30 lea
Re: Why the compilation with gcc and with afl-gcc is different, a short look at the afl-gcc source shows that by default it modifies the compiler parameters, setting -O3 -funroll-loops
(as well as defining __AFL_COMPILER
and FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
).
According to the documentation (docs/env_variables.txt):
By default, the wrapper appends -O3 to optimize builds. Very rarely, this will cause problems in programs built with -Werror, simply because -O3 enables more thorough code analysis and can spew out additional warnings. To disable optimizations, set AFL_DONT_OPTIMIZE.