# 第二十章 函數指針
函數指針是指向函數的指針,和其他指針一樣,只是該指針指向函數代碼段的開始地址。函數指針經常用作回調1。
典型的例子如下:
```
C標準庫的 qsort()2, aexit()3;
*NIX OS的信號機制;
線程啟動:CreateThread()(Win32),pthread_create()(POSIX);
其他更多的Win32函數,比如EnumChildWindows()5。
```
qsort()函數是C/C++標準庫快速排序函數。該函數能夠排序任意類型的數據。qsort()調用比較函數。
比較函數被定義為如下形式:
```
#!cpp
int (*compare)(const void *, const void *)
```
我們稍作修改:
```
#!cpp
/* ex3 Sorting ints with qsort */
#include <stdio.h>
#include <stdlib.h>
int comp(const void * _a, const void * _b)
{
const int *a=(const int *)_a;
const int *b=(const int *)_b;
if (*a==*b)
return 0;
else
if (*a < *b)
return -1;
else
return 1;
}
int main(int argc, char* argv[])
{
int numbers[10]={1892,45,200,-98,4087,5,-12345,1087,88,-100000};
int i;
/* Sort the array */
qsort(numbers,10,sizeof(int),comp) ;
for (i=0;i<9;i++)
printf("Number = %d
",numbers[ i ]) ;
return 0;
}
```
## 20.1 MSVC
MSVC2010 /Ox選項編譯:
Listing 20.1: Optimizing MSVC 2010: /Ox /GS- /MD
```
#!bash
__a$ = 8 ; size = 4
__b$ = 12 ; size = 4
_comp PROC
mov eax, DWORD PTR __a$[esp-4]
mov ecx, DWORD PTR __b$[esp-4]
mov eax, DWORD PTR [eax]
mov ecx, DWORD PTR [ecx]
cmp eax, ecx
jne SHORT $LN4@comp
xor eax, eax
ret 0
$LN4@comp:
xor edx, edx
cmp eax, ecx
setge dl
lea eax, DWORD PTR [edx+edx-1]
ret 0
_comp ENDP
_numbers$ = -40 ; size = 40
_argc$ = 8 ; size = 4
_argv$ = 12 ; size = 4
_main PROC
sub esp, 40 ; 00000028H
push esi
push OFFSET _comp
push 4
lea eax, DWORD PTR _numbers$[esp+52]
push 10 ; 0000000aH
push eax
mov DWORD PTR _numbers$[esp+60], 1892 ; 00000764H
mov DWORD PTR _numbers$[esp+64], 45 ; 0000002dH
mov DWORD PTR _numbers$[esp+68], 200 ; 000000c8H
mov DWORD PTR _numbers$[esp+72], -98 ; ffffff9eH
mov DWORD PTR _numbers$[esp+76], 4087 ; 00000ff7H
mov DWORD PTR _numbers$[esp+80], 5
mov DWORD PTR _numbers$[esp+84], -12345 ; ffffcfc7H
mov DWORD PTR _numbers$[esp+88], 1087 ; 0000043fH
mov DWORD PTR _numbers$[esp+92], 88 ; 00000058H
mov DWORD PTR _numbers$[esp+96], -100000 ; fffe7960H
call _qsort
add esp, 16 ; 00000010H
...
```
第四個參數傳遞了一個地址標簽_comp,指向了comp()函數。
我們來看MSVCR80.DLL(包含C標準庫函數的MSVC DLL模塊)里該函數的內部調用:
Listing 20.2: MSVCR80.DLL
```
#!bash
.text:7816CBF0 ; void __cdecl qsort(void *, unsigned int, unsigned int, int (__cdecl *)(const void *, const void *))
.text:7816CBF0 public _qsort
.text:7816CBF0 _qsort proc near
.text:7816CBF0
.text:7816CBF0 lo = dword ptr -104h
.text:7816CBF0 hi = dword ptr -100h
.text:7816CBF0 var_FC = dword ptr -0FCh
.text:7816CBF0 stkptr = dword ptr -0F8h
.text:7816CBF0 lostk = dword ptr -0F4h
.text:7816CBF0 histk = dword ptr -7Ch
.text:7816CBF0 base = dword ptr 4
.text:7816CBF0 num = dword ptr 8
.text:7816CBF0 width = dword ptr 0Ch
.text:7816CBF0 comp = dword ptr 10h
.text:7816CBF0
.text:7816CBF0 sub esp, 100h
....
.text:7816CCE0 loc_7816CCE0: ; CODE XREF: _qsort+B1
.text:7816CCE0 shr eax, 1
.text:7816CCE2 imul eax, ebp
.text:7816CCE5 add eax, ebx
.text:7816CCE7 mov edi, eax
.text:7816CCE9 push edi
.text:7816CCEA push ebx
.text:7816CCEB call [esp+118h+comp]
.text:7816CCF2 add esp, 8
.text:7816CCF5 test eax, eax
.text:7816CCF7 jle short loc_7816CD04
```
第四個參數comp傳遞函數指針,comp()有兩個參數,參數被檢測后才執行。
這種使用函數指針的方式有一定的風險。第一種原因是如果你用qsort()調用了錯誤的函數指針,可能造成程序崩潰,并且這個錯誤很難被發現。
第二個原因是即使回調函數類型完全正確,使用錯誤的參數調用函數可能會導致更嚴重的問題。進程崩潰不是最大的問題,最大的問題是崩潰的原因—編譯器很難發現這種潛在的問題。
### 20.1.1 MSVC + OllyDbg
我們在OD中加載我們的例子,并在comp()函數下斷點。
我們可以看到第一次comp()調用時是如何比較的:fig.20.1.OD代碼窗口顯示了比較的值。我們還可以看到SP指向的RA地址在qsort()函數空間里(實際上位于MSVCR100.DLL)。
按F8直到函數返回到qsort()函數:fig20.2.這里比較函數被調用。
第二次調用comp()—當前比較的值不相同:fig203。

Figure 20.1: OllyDbg: first call of comp()

Figure 20.2: OllyDbg: the code in qsort() right a_er comp() call

Figure 20.3: OllyDbg: second call of comp()
### 20.1.2 MSVC + tracer
我們來看成對比較,來對10個數字進行排序:1892, 45, 200, -98, 4087, 5, -12345, 1087, 88,-100000.
我們找到comp()函數中的CMP指令地址,并在其地址0x0040100C上設置斷點。
```
#!bash
tracer.exe -l:17_1.exe bpx=17_1.exe!0x0040100C
```
斷點中斷是的寄存器地址:
```
#!bash
PID=4336|New process 17_1.exe
(0) 17_1.exe!0x40100c
EAX=0x00000764 EBX=0x0051f7c8 ECX=0x00000005 EDX=0x00000000
ESI=0x0051f7d8 EDI=0x0051f7b4 EBP=0x0051f794 ESP=0x0051f67c
EIP=0x0028100c
FLAGS=IF
(0) 17_1.exe!0x40100c
EAX=0x00000005 EBX=0x0051f7c8 ECX=0xfffe7960 EDX=0x00000000
ESI=0x0051f7d8 EDI=0x0051f7b4 EBP=0x0051f794 ESP=0x0051f67c
EIP=0x0028100c
FLAGS=PF ZF IF
(0) 17_1.exe!0x40100c
EAX=0x00000764 EBX=0x0051f7c8 ECX=0x00000005 EDX=0x00000000
ESI=0x0051f7d8 EDI=0x0051f7b4 EBP=0x0051f794 ESP=0x0051f67c
EIP=0x0028100c
FLAGS=CF PF ZF IF
...
```
過濾EAX和ECX得到:
```
EAX=0x00000764 ECX=0x00000005
EAX=0x00000005 ECX=0xfffe7960
EAX=0x00000764 ECX=0x00000005
EAX=0x0000002d ECX=0x00000005
EAX=0x00000058 ECX=0x00000005
EAX=0x0000043f ECX=0x00000005
EAX=0xffffcfc7 ECX=0x00000005
EAX=0x000000c8 ECX=0x00000005
EAX=0xffffff9e ECX=0x00000005
EAX=0x00000ff7 ECX=0x00000005
EAX=0x00000ff7 ECX=0x00000005
EAX=0xffffff9e ECX=0x00000005
EAX=0xffffff9e ECX=0x00000005
EAX=0xffffcfc7 ECX=0xfffe7960
EAX=0x00000005 ECX=0xffffcfc7
EAX=0xffffff9e ECX=0x00000005
EAX=0xffffcfc7 ECX=0xfffe7960
EAX=0xffffff9e ECX=0xffffcfc7
EAX=0xffffcfc7 ECX=0xfffe7960
EAX=0x000000c8 ECX=0x00000ff7
EAX=0x0000002d ECX=0x00000ff7
EAX=0x0000043f ECX=0x00000ff7
EAX=0x00000058 ECX=0x00000ff7
EAX=0x00000764 ECX=0x00000ff7
EAX=0x000000c8 ECX=0x00000764
EAX=0x0000002d ECX=0x00000764
EAX=0x0000043f ECX=0x00000764
EAX=0x00000058 ECX=0x00000764
EAX=0x000000c8 ECX=0x00000058
EAX=0x0000002d ECX=0x000000c8
EAX=0x0000043f ECX=0x000000c8
EAX=0x000000c8 ECX=0x00000058
EAX=0x0000002d ECX=0x000000c8
EAX=0x0000002d ECX=0x00000058
```
有34對。因此快速排序算法對10個數字排序需要34此對比操作。
### 20.1.3 MSVC + tracer (code coverage)
我們使用跟蹤特性收集寄存器的值并在IDA中查看。
跟蹤comp()函數所有指令:
```
tracer.exe -l:17_1.exe bpf=17_1.exe!0x00401000,trace:cc
```
IDA加載.idc腳本:fig20.4。
IDA給出了函數名字(PtFuncCompare)—IDA認為該函數指針被傳遞給qsort()。
可以看到a和b指向數組不同的位置,并且相差4-32bit的字節數。
0x401010 和 0x401012之間的指令從沒有被執行:事實上comp()從來不返回0,因為沒有相等的元素。

Figure 20.4: tracer and IDA. N.B.: some values are cutted at right
## 20.2 GCC
沒有太大的不同:
Listing 20.3: GCC
```
#!bash
lea eax, [esp+40h+var_28]
mov [esp+40h+var_40], eax
mov [esp+40h+var_28], 764h
mov [esp+40h+var_24], 2Dh
mov [esp+40h+var_20], 0C8h
mov [esp+40h+var_1C], 0FFFFFF9Eh
mov [esp+40h+var_18], 0FF7h
mov [esp+40h+var_14], 5
mov [esp+40h+var_10], 0FFFFCFC7h
mov [esp+40h+var_C], 43Fh
mov [esp+40h+var_8], 58h
mov [esp+40h+var_4], 0FFFE7960h
mov [esp+40h+var_34], offset comp
mov [esp+40h+var_38], 4
mov [esp+40h+var_3C], 0Ah
call _qsort
```
comp() 函數:
```
#!bash
public comp
comp proc near
arg_0 = dword ptr 8
arg_4 = dword ptr 0Ch
push ebp
mov ebp, esp
mov eax, [ebp+arg_4]
mov ecx, [ebp+arg_0]
mov edx, [eax]
xor eax, eax
cmp [ecx], edx
jnz short loc_8048458
pop ebp
retn
loc_8048458:
setnl al
movzx eax, al
lea eax, [eax+eax-1]
pop ebp
retn
comp endp
```
qsort()的實現在libc.so里,它實際上是qsort_r()的封裝。
我們通過傳遞函數指針然后調用快速排序:
Listing 20.4: (file libc.so.6, glibc version—2.10.1)
```
#!bash
.text:0002DDF6 mov edx, [ebp+arg_10]
.text:0002DDF9 mov [esp+4], esi
.text:0002DDFD mov [esp], edi
.text:0002DE00 mov [esp+8], edx
.text:0002DE04 call [ebp+arg_C]
...
```
### 20.2.1 GCC + GDB (with source code)
因為我們有例子的C源代碼,我們能在行數(11—第一次比較的地方)設置斷點(b)。編譯例子的時候使用了帶有調試信息的選項(-g),當前可以查看地址及行號,也可以打印變量(p):調試信息包含寄存器和變量值信息。
我們查看堆棧(bt),看到glibc使用的中間函數msort_with_tmp()。
Listing 20.5: GDB session
```
#!bash
dennis@ubuntuvm:~/polygon$ gcc 17_1.c -g
dennis@ubuntuvm:~/polygon$ gdb ./a.out
GNU gdb (GDB) 7.6.1-ubuntu
Copyright (C) 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /home/dennis/polygon/a.out...done.
(gdb) b 17_1.c:11
Breakpoint 1 at 0x804845f: file 17_1.c, line 11.
(gdb) run
Starting program: /home/dennis/polygon/./a.out
Breakpoint 1, comp (_a=0xbffff0f8, _b=_b@entry=0xbffff0fc) at 17_1.c:11
11 if (*a==*b)
(gdb) p *a
$1 = 1892
(gdb) p *b
$2 = 45
(gdb) c
Continuing.
Breakpoint 1, comp (_a=0xbffff104, _b=_b@entry=0xbffff108) at 17_1.c:11
11 if (*a==*b)
(gdb) p *a
$3 = -98
(gdb) p *b
$4 = 4087
(gdb) bt
#0 comp (_a=0xbffff0f8, _b=_b@entry=0xbffff0fc) at 17_1.c:11
#1 0xb7e42872 in msort_with_tmp (p=p@entry=0xbffff07c, b=b@entry=0xbffff0f8, n=n@entry=2)
at msort.c:65
#2 0xb7e4273e in msort_with_tmp (n=2, b=0xbffff0f8, p=0xbffff07c) at msort.c:45
#3 msort_with_tmp (p=p@entry=0xbffff07c, b=b@entry=0xbffff0f8, n=n@entry=5) at msort.c:53
#4 0xb7e4273e in msort_with_tmp (n=5, b=0xbffff0f8, p=0xbffff07c) at msort.c:45
#5 msort_with_tmp (p=p@entry=0xbffff07c, b=b@entry=0xbffff0f8, n=n@entry=10) at msort.c:53
#6 0xb7e42cef in msort_with_tmp (n=10, b=0xbffff0f8, p=0xbffff07c) at msort.c:45
#7 __GI_qsort_r (b=b@entry=0xbffff0f8, n=n@entry=10, s=s@entry=4, cmp=cmp@entry=0x804844d <comp
>,
arg=arg@entry=0x0) at msort.c:297
#8 0xb7e42dcf in __GI_qsort (b=0xbffff0f8, n=10, s=4, cmp=0x804844d <comp>) at msort.c:307
#9 0x0804850d in main (argc=1, argv=0xbffff1c4) at 17_1.c:26
(gdb)
```
### 20.2.2 GCC + GDB (no source code)
更多時候我們沒有源碼,我們可以反匯編comp()函數(disas),找到CMP指令地址并設置斷點(b)。每次中斷后,dump所有寄存器的值(info registers),堆棧信息(bt),但是沒有comp()函數對應的行號信息。
Listing 20.6: GDB session
```
#!bash
dennis@ubuntuvm:~/polygon$ gcc 17_1.c
dennis@ubuntuvm:~/polygon$ gdb ./a.out
GNU gdb (GDB) 7.6.1-ubuntu
Copyright (C) 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /home/dennis/polygon/a.out...(no debugging symbols found)...done.
(gdb) set disassembly-flavor intel
(gdb) disas comp
Dump of assembler code for function comp:
0x0804844d <+0>: push ebp
0x0804844e <+1>: mov ebp,esp
0x08048450 <+3>: sub esp,0x10
0x08048453 <+6>: mov eax,DWORD PTR [ebp+0x8]
0x08048456 <+9>: mov DWORD PTR [ebp-0x8],eax
0x08048459 <+12>: mov eax,DWORD PTR [ebp+0xc]
0x0804845c <+15>: mov DWORD PTR [ebp-0x4],eax
0x0804845f <+18>: mov eax,DWORD PTR [ebp-0x8]
0x08048462 <+21>: mov edx,DWORD PTR [eax]
0x08048464 <+23>: mov eax,DWORD PTR [ebp-0x4]
0x08048467 <+26>: mov eax,DWORD PTR [eax]
0x08048469 <+28>: cmp edx,eax
0x0804846b <+30>: jne 0x8048474 <comp+39>
0x0804846d <+32>: mov eax,0x0
0x08048472 <+37>: jmp 0x804848e <comp+65>
0x08048474 <+39>: mov eax,DWORD PTR [ebp-0x8]
0x08048477 <+42>: mov edx,DWORD PTR [eax]
0x08048479 <+44>: mov eax,DWORD PTR [ebp-0x4]
0x0804847c <+47>: mov eax,DWORD PTR [eax]
0x0804847e <+49>: cmp edx,eax
0x08048480 <+51>: jge 0x8048489 <comp+60>
0x08048482 <+53>: mov eax,0xffffffff
0x08048487 <+58>: jmp 0x804848e <comp+65>
0x08048489 <+60>: mov eax,0x1
0x0804848e <+65>: leave
0x0804848f <+66>: ret
End of assembler dump.
(gdb) b *0x08048469
Breakpoint 1 at 0x8048469
(gdb) run
Starting program: /home/dennis/polygon/./a.out
Breakpoint 1, 0x08048469 in comp ()
(gdb) info registers
eax 0x2d 45
ecx 0xbffff0f8 -1073745672
edx 0x764 1892
ebx 0xb7fc0000 -1208221696
esp 0xbfffeeb8 0xbfffeeb8
ebp 0xbfffeec8 0xbfffeec8
esi 0xbffff0fc -1073745668
edi 0xbffff010 -1073745904
eip 0x8048469 0x8048469 <comp+28>
eflags 0x286 [ PF SF IF ]
cs 0x73 115
ss 0x7b 123
ds 0x7b 123
es 0x7b 123
fs 0x0 0
gs 0x33 51
(gdb) c
Continuing.
Breakpoint 1, 0x08048469 in comp ()
(gdb) info registers
eax 0xff7 4087
ecx 0xbffff104 -1073745660
edx 0xffffff9e -98
ebx 0xb7fc0000 -1208221696
esp 0xbfffee58 0xbfffee58
ebp 0xbfffee68 0xbfffee68
esi 0xbffff108 -1073745656
edi 0xbffff010 -1073745904
eip 0x8048469 0x8048469 <comp+28>
eflags 0x282 [ SF IF ]
cs 0x73 115
ss 0x7b 123
ds 0x7b 123
es 0x7b 123
fs 0x0 0
gs 0x33 51
(gdb) c
Continuing.
Breakpoint 1, 0x08048469 in comp ()
(gdb) info registers
eax 0xffffff9e -98
ecx 0xbffff100 -1073745664
edx 0xc8 200
ebx 0xb7fc0000 -1208221696
esp 0xbfffeeb8 0xbfffeeb8
ebp 0xbfffeec8 0xbfffeec8
esi 0xbffff104 -1073745660
edi 0xbffff010 -1073745904
eip 0x8048469 0x8048469 <comp+28>
eflags 0x286 [ PF SF IF ]
cs 0x73 115
ss 0x7b 123
ds 0x7b 123
es 0x7b 123
fs 0x0 0
gs 0x33 51
(gdb) bt
#0 0x08048469 in comp ()
#1 0xb7e42872 in msort_with_tmp (p=p@entry=0xbffff07c, b=b@entry=0xbffff0f8, n=n@entry=2)
at msort.c:65
#2 0xb7e4273e in msort_with_tmp (n=2, b=0xbffff0f8, p=0xbffff07c) at msort.c:45
#3 msort_with_tmp (p=p@entry=0xbffff07c, b=b@entry=0xbffff0f8, n=n@entry=5) at msort.c:53
#4 0xb7e4273e in msort_with_tmp (n=5, b=0xbffff0f8, p=0xbffff07c) at msort.c:45
#5 msort_with_tmp (p=p@entry=0xbffff07c, b=b@entry=0xbffff0f8, n=n@entry=10) at msort.c:53
#6 0xb7e42cef in msort_with_tmp (n=10, b=0xbffff0f8, p=0xbffff07c) at msort.c:45
#7 __GI_qsort_r (b=b@entry=0xbffff0f8, n=n@entry=10, s=s@entry=4, cmp=cmp@entry=0x804844d <comp
>,
arg=arg@entry=0x0) at msort.c:297
#8 0xb7e42dcf in __GI_qsort (b=0xbffff0f8, n=10, s=4, cmp=0x804844d <comp>) at msort.c:307
#9 0x0804850d in main ()
```
- 第一章 CPU簡介
- 第二章 Hello,world!
- 第三章? 函數開始和結束
- 第四章 棧
- Chapter 5 printf() 與參數處理
- Chapter 6 scanf()
- CHAPER7 訪問傳遞參數
- Chapter 8 一個或者多個字的返回值
- Chapter 9 指針
- Chapter 10 條件跳轉
- 第11章 選擇結構switch()/case/default
- 第12章 循環結構
- 第13章 strlen()
- Chapter 14 Division by 9
- chapter 15 用FPU工作
- Chapter 16 數組
- Chapter 17 位域
- 第18章 結構體
- 19章 聯合體
- 第二十章 函數指針
- 第21章 在32位環境中的64位值
- 第二十二章 SIMD
- 23章 64位化
- 24章 使用x64下的SIMD來處理浮點數
- 25章 溫度轉換
- 26章 C99的限制
- 27章 內聯函數
- 第28章 得到不正確反匯編結果
- 第29章 花指令
- 第30章 16位Windows
- 第31章 類
- 三十二 ostream