gcc常用的编译选项对代码的影响

by alert7
2001-12-21

测试环境 redhat 6.2

★ 前言

本文讨论gcc的一些常用编译选项对代码的影响。当然代码变了，
它的内存布局也就会变了，随之exploit也就要做相应的变动。
gcc的编译选项实在太多，本文检了几个最常用的选项。


★ 演示程序

[alert7@redhat62 alert7]$ cat > test.c
#include <stdio.h>
void hi(void)
{
printf("hi");
}

int main(int argc, char *argv[])
{
hi();
return 0;
}

★ 一般情况

[alert7@redhat62 alert7]$ gcc -o test test.c
[alert7@redhat62 alert7]$ wc -c test
11773 test
[alert7@redhat62 alert7]$ gdb -q test
(gdb) disass main
Dump of assembler code for function main:
0x80483e4 <main>: push %ebp
0x80483e5 <main+1>: mov %esp,%ebp
0x80483e7 <main+3>: call 0x80483d0 <hi>
0x80483ec <main+8>: xor %eax,%eax
0x80483ee <main+10>: jmp 0x80483f0 <main+12>
0x80483f0 <main+12>: leave
0x80483f1 <main+13>: ret
....
End of assembler dump.
(gdb) disass hi
Dump of assembler code for function hi:
0x80483d0 <hi>: push %ebp
0x80483d1 <hi+1>: mov %esp,%ebp
0x80483d3 <hi+3>: push $0x8048450
0x80483d8 <hi+8>: call 0x8048308 <printf>
0x80483dd <hi+13>: add $0x4,%esp
0x80483e0 <hi+16>: leave
0x80483e1 <hi+17>: ret
0x80483e2 <hi+18>: mov %esi,%esi
End of assembler dump.

来看看部分的内存映象
(内存高址)
+--------+
|bffffbc4| argv的地址(即argv[0]的地址)
0xbffffb84 +--------+
|00000001| argc的值
0xbffffb80 +--------+
|400309cb|main的返回地址
0xbffffb7c +--------+ <-- 调用main函数前的esp
|bffffb98| 调用main函数前的ebp
0xbffffb78 +--------+ <-- main函数的ebp
|080483ec| hi()的返回地址
0xbffffb74 +--------+
|bffffb78| 调用hi()前的esp
0xbffffb70 +--------+
|08048450| "hi"的地址
0xbffffb6c +--------+
| ...... |
(内存低址)

leave 指令所做的操作相当于MOV ESP,EBP 然后 POP EBP
ret 指令所做的操作相当于POP EIP

★ -O2 编译选项

-O2 Optimize even more. Nearly all supported optimizations that do
not involve a space-speed tradeoff are performed. Loop unrolling
and function inlining are not done, for example. As compared to -O,
this option increases both compilation time and the performance of
the generated code.

[alert7@redhat62 alert7]$ gcc -O2 -o test test.c
[alert7@redhat62 alert7]$ wc -c test
11757 test
[alert7@redhat62 alert7]$ gdb -q test
(gdb) disass main
Dump of assembler code for function main:
0x80483d8 <main>: push %ebp
0x80483d9 <main+1>: mov %esp,%ebp
0x80483db <main+3>: call 0x80483c8 <hi>
0x80483e0 <main+8>: xor %eax,%eax
0x80483e2 <main+10>: leave
0x80483e3 <main+11>: ret
...
0x80483ef <main+23>: nop
End of assembler dump.
(gdb) disass hi
Dump of assembler code for function hi:
0x80483c8 <hi>: push %ebp
0x80483c9 <hi+1>: mov %esp,%ebp
0x80483cb <hi+3>: push $0x8048440
0x80483d0 <hi+8>: call 0x8048308 <printf>
0x80483d5 <hi+13>: leave
0x80483d6 <hi+14>: ret
0x80483d7 <hi+15>: nop
End of assembler dump.

由于程序比较简单，再优化也没有好优化的了，所以跟-O出来的一样。

★ -fomit-frame-pointer 编译选项

-fomit-frame-pointer
Don't keep the frame pointer in a register for functions
that don't need one. This avoids the instructions to save,
set up and restore frame pointers; it also makes an extra
register available in many functions. It also makes
debugging impossible on most machines.

忽略帧指针。这样在程序就不需要保存，安装，和恢复ebp了。这样ebp也就是一个
free的register了，在函数中就可以随便使用了。

[alert7@redhat62 alert7]$ gcc -fomit-frame-pointer -o test test.c
[alert7@redhat62 alert7]$ wc -c test
11773 test
[alert7@redhat62 alert7]$ gdb -q test
(gdb) disass main
Dump of assembler code for function main:
0x80483e0 <main>: call 0x80483d0 <hi>
0x80483e5 <main+5>: xor %eax,%eax
0x80483e7 <main+7>: jmp 0x80483f0 <main+16>
0x80483e9 <main+9>: lea 0x0(%esi,1),%esi
0x80483f0 <main+16>: ret
....
End of assembler dump.
(gdb) disass hi
Dump of assembler code for function hi:
0x80483d0 <hi>: push $0x8048450
0x80483d5 <hi+5>: call 0x8048308 <printf>
0x80483da <hi+10>: add $0x4,%esp
0x80483dd <hi+13>: ret
0x80483de <hi+14>: mov %esi,%esi
End of assembler dump.

在main()和hi()中都去掉了以下指令
push %ebp
mov %esp,%ebp//这两条指令安装
leave//这条指令恢复
来看看部分的内存映象
(内存高址)
+--------+
|bffffbc4| argv的地址(即argv[0]的地址)
0xbffffb84 +--------+
|00000001| argc的值
0xbffffb80 +--------+
|400309cb|main的返回地址
0xbffffb7c +--------+
|080483e5| hi()的返回地址
0xbffffb78 +--------+
|08048450| "hi"字符串的地址
0xbffffb74 +--------+
| ...... |
(内存低址)
没有保存上层执行环境的ebp.

★ -fomit-frame-pointer && -O2

-fomit-frame-pointer编译选项去掉了
push %ebp
mov %esp,%ebp//这两条指令安装
leave//这条指令恢复
-O2编译选项去掉了
add $0x4,%esp

两个加起来会不会这四条指令一起去掉，从而使stack不平衡呢？

[alert7@redhat62 alert7]$ gcc -fomit-frame-pointer -O2 -o test test.c
[alert7@redhat62 alert7]$ wc -c test
11741 test
[alert7@redhat62 alert7]$ gdb -q test
(gdb) disass main
Dump of assembler code for function main:
0x80483d8 <main>: call 0x80483c8 <hi>
0x80483dd <main+5>: xor %eax,%eax
0x80483df <main+7>: ret
End of assembler dump.
(gdb) disass hi
Dump of assembler code for function hi:
0x80483c8 <hi>: push $0x8048430
0x80483cd <hi+5>: call 0x8048308 <printf>
0x80483d2 <hi+10>: add $0x4,%esp
0x80483d5 <hi+13>: ret
0x80483d6 <hi+14>: mov %esi,%esi
End of assembler dump.
来看看部分的内存映象
(内存高址)
+--------+
|bffffbc4| argv的地址(即argv[0]的地址)
0xbffffb84 +--------+
|00000001| argc的值
0xbffffb80 +--------+
|400309cb|main的返回地址
0xbffffb7c +--------+
|080483dd| hi()的返回地址
0xbffffb78 +--------+
|08048430| "hi"字符串的地址
0xbffffb74 +--------+
| ...... |
(内存低址)

此时就没有把add $0x4,%esp优化掉，如果优化掉的话，整个stack就
会变的不平衡，从而会导致程序出错。

★ -fPIC 编译选项

-fPIC If supported for the target machine, emit position-independent
code, suitable for dynamic linking,even if branches need large
displacements.
产生位置无关代码（PIC），一般创建共享库时用到。
在x86上，PIC的代码的符号引用都是通过ebx进行操作的。

[alert7@redhat62 alert7]$ gcc -fPIC -o test test.c
[alert7@redhat62 alert7]$ wc -c test
11805 test
[alert7@redhat62 alert7]$ gdb -q test
(gdb) disass main
Dump of assembler code for function main:
0x80483f8 <main>: push %ebp
0x80483f9 <main+1>: mov %esp,%ebp
0x80483fb <main+3>: push %ebx
0x80483fc <main+4>: call 0x8048401 <main+9>
0x8048401 <main+9>: pop %ebx//取得该指令的地址
0x8048402 <main+10>: add $0x1093,%ebx//此时ebx里面存放着是GOT表的地址
0x8048408 <main+16>: call 0x80483d0 <hi>
0x804840d <main+21>: xor %eax,%eax
0x804840f <main+23>: jmp 0x8048411 <main+25>
0x8048411 <main+25>: mov 0xfffffffc(%ebp),%ebx
0x8048414 <main+28>: leave
0x8048415 <main+29>: ret
...
End of assembler dump.
(gdb) disass hi
Dump of assembler code for function hi:
0x80483d0 <hi>: push %ebp
0x80483d1 <hi+1>: mov %esp,%ebp
0x80483d3 <hi+3>: push %ebx
0x80483d4 <hi+4>: call 0x80483d9 <hi+9>
0x80483d9 <hi+9>: pop %ebx
0x80483da <hi+10>: add $0x10bb,%ebx
0x80483e0 <hi+16>: lea 0xffffefdc(%ebx),%edx
0x80483e6 <hi+22>: mov %edx,%eax
0x80483e8 <hi+24>: push %eax
0x80483e9 <hi+25>: call 0x8048308 <printf>
0x80483ee <hi+30>: add $0x4,%esp
0x80483f1 <hi+33>: mov 0xfffffffc(%ebp),%ebx
0x80483f4 <hi+36>: leave
0x80483f5 <hi+37>: ret
0x80483f6 <hi+38>: mov %esi,%esi
End of assembler dump.
来看看部分的内存映象

(内存高址)
+--------+
|bffffbc4| argv的地址(即argv[0]的地址)
0xbffffb84 +--------+
|00000001| argc的值
0xbffffb80 +--------+
|400309cb|main的返回地址
0xbffffb7c +--------+ <-- 调用main函数前的esp
|bffffb98| 调用main函数前的ebp
0xbffffb78 +--------+ <-- main函数的ebp
|401081ec| 保存的ebx
0xbffffb74 +--------+
|0804840d| （存放过call 0x8048401的下一条指令地址）
0xbffffb70 +--------+
|bffffb78| 调用hi()前的esp
0xbffffb6c +--------+
|08049494| GOT表地址
0xbffffb68 +--------+
|08048470|（存放过call 0x80483d9的下一条指令地址）
0xbffffb64 +--------+
| ...... |
(内存低址)

★ -static 编译选项

-static
On systems that support dynamic linking, this prevents
linking with the shared libraries. On other systems,
this option has no effect.
把一些函数都静态的编译到程序中，而无需动态链接了。

[alert7@redhat62 alert7]$ gcc -o test -static test.c
[alert7@redhat62 alert7]$ wc -c test
962808 test
[alert7@redhat62 alert7]$ gdb -q test
(gdb) disass main
Dump of assembler code for function main:
0x80481b4 <main>: push %ebp
0x80481b5 <main+1>: mov %esp,%ebp
0x80481b7 <main+3>: call 0x80481a0 <hi>
0x80481bc <main+8>: xor %eax,%eax
0x80481be <main+10>: jmp 0x80481c0 <main+12>
0x80481c0 <main+12>: leave
0x80481c1 <main+13>: ret
...
End of assembler dump.
(gdb) disass hi
Dump of assembler code for function hi:
0x80481a0 <hi>: push %ebp
0x80481a1 <hi+1>: mov %esp,%ebp
0x80481a3 <hi+3>: push $0x8071528
0x80481a8 <hi+8>: call 0x804865c <printf>
0x80481ad <hi+13>: add $0x4,%esp
0x80481b0 <hi+16>: leave
0x80481b1 <hi+17>: ret
0x80481b2 <hi+18>: mov %esi,%esi
End of assembler dump.
[alert7@redhat62 alert7]$ ldd test
not a dynamic executable
-static出来的代码已经没有PLT了，GOT虽然有，已经全部为0了。

还有-m486等

这引起是跟CPU相关的
一般在本机上运行时可以选择
如果想做发布版的最好不要选择

眼花