上一篇文章《Linux Debugging:使用反汇编理解C++程序函数调用栈》没想到能得到那么多人的喜爱,因为那篇文章是以32位的C++普通函数(非类成员函数)为例子写的,因此只是一个特殊的例子。本文将函数调用时的参数传递方法进行一下总结。总结将为C++普通函数、类成员函数;32位和64位进行总结。
建议还是读一下Linux Debugging:使用反汇编理解C++程序函数调用栈,这样本文的结论将非常容易理解,将非常好的为CoreDump分析开一个好头。而且,它也是32位C++ 普通函数的调用的比较好的例子,毕竟从汇编的角度,将参数如何传递的进行了比较好的说明。
1. 32位程序普通函数
普通函数的意思是非class member functionvoid func2(int a, int b) { a++; b+ = 2; } int main() { func2( 1111, 2222); return 0; }
main函数的汇编:
main: pushl %ebp movl %esp, %ebp subl $8, %esp movl $2222, 4(%esp) movl $1111, (%esp) call func2(int, int) movl $0, %eax leave ret
1111是第一个参数,放到了esp指向的地址。2222是第二个参数,放到了高地址。因次我们可以知道,在函数func2中,通过ebp+8可以访问到第一个参数1111,通过ebp+12可以访问到第二个参数2222。
func2(int, int): pushl %ebp movl %esp, %ebp addl $1, 8(%ebp) addl $2, 12(%ebp) popl %ebp ret
下面我们使用gdb通过ebp打印一下传入的参数:
anzhsoft@ubuntu:~/linuxDebugging/parameter$ gdb a.out GNU gdb 6.8 Copyright (C) 2008 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-pc-linux-gnu"... (gdb) b func2 Breakpoint 1 at 0x8048597: file m32noclass.cpp, line 6. (gdb) r Starting program: /home/anzhsoft/linuxDebugging/parameter/a.out Breakpoint 1, func2 (a=1111, b=2222) at m32noclass.cpp:6 warning: Source file is more recent than executable. 6 a++; (gdb) p *(int*)($ebp+8) $1 = 1111 (gdb) p *(int*)($ebp+12) $2 = 2222
总结:
1. 参数通过栈查传递,底地址传递从左边开始的第一个参数
2. 使用gdb可以很方便打印传入参数
(gdb) p *(int*)($ebp+8) $1 = 1111 (gdb) p *(int*)($ebp+12) $2 = 2222
其实32位的程序也可以使用寄存器传递参数。请看下一节。
2. 32位普通函数-通过寄存器传递参数
可以使用GCC的扩展功能__attribute__使得参数传递可以使用寄存器。
修改第一节的函数:
#define STACKCALL __attribute__((regparm(3))) void STACKCALL func4(int a, int b, int c, int d) { a++; b += 2; c += 3; d += 4; } int main() { func4(1111, 2222, 3333, 4444); return 0; }__attribute__((regparm(3)))意思是使用寄存器
anzhsoft@ubuntu:~/linuxDebugging/parameter$ gdb a.out GNU gdb 6.8 Copyright (C) 2008 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-pc-linux-gnu"... (gdb) b main Breakpoint 1 at 0x80485bb: file m32noclass.cpp, line 14. (gdb) r Starting program: /home/anzhsoft/linuxDebugging/parameter/a.out Breakpoint 1, main () at m32noclass.cpp:14 14 func4(1111, 2222, 3333, 4444); (gdb) s func4 (a=1111, b=2222, c=3333, d=4444) at m32noclass.cpp:6 6 a++; (gdb) i r eax 0x457 1111 ecx 0xd05 3333 edx 0x8ae 2222 ebx 0xb3eff4 11792372 esp 0xbf8e9580 0xbf8e9580 ebp 0xbf8e958c 0xbf8e958c esi 0x0 0 edi 0x0 0 eip 0x80485a3 0x80485a3 <func4(int, int, int, int)+15> eflags 0x282 [ SF IF ] cs 0x73 115 ss 0x7b 123 ds 0x7b 123 es 0x7b 123 fs 0x0 0 gs 0x33 51 (gdb) p *(int*)($ebp+8) $1 = 4444 (gdb)
可以看到,前三个参数分别通过eax/edx/ecx传递,第四个参数通过栈传递。这种传递方式被称为fastcall
3. 32位 class member function 参数传递方式
我们通过宏USINGSTACK强制使用栈来传递参数。#define USINGSTACK __attribute__((regparm(0))) class Test { public: Test():number(3333){} void USINGSTACK func2(int a, int b) { a++; b += 2; } private: int number; }; int main(int argc, char* argv[]) { Test tInst; tInst.func2(1111, 2222); return 0; }
通过gdb来打印传入的参数:
anzhsoft@ubuntu:~/linuxDebugging/parameter$ gdb a.out GNU gdb 6.8 Copyright (C) 2008 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-pc-linux-gnu"... (gdb) b main Breakpoint 1 at 0x804859d: file m32class.cpp, line 18. (gdb) r Starting program: /home/anzhsoft/linuxDebugging/parameter/a.out Breakpoint 1, main (argc=1, argv=0xbf98a454) at m32class.cpp:18 18 Test tInst; (gdb) n 19 tInst.func2(1111, 2222); (gdb) s Test::func2 (this=0xbf98a39c, a=1111, b=2222) at m32class.cpp:9 9 a++; (gdb) p *(int*)($ebp+12) $1 = 1111 (gdb) p *(int*)($ebp+16) $2 = 2222 (gdb) p *this $3 = {number = 3333}
可以看到,class成员函数的第一个参数是对象的指针。通过这种方式,成员函数可以和非成员函数以类似的方式进行调用。
4. x86-64 class member function 的参数传递
在x86-64中,整形和指针型参数的参数从左到右依次保存到rdi,rsi,rdx,rcx,r8,r9中。浮点型参数会保存到xmm0,xmm1……。多余的参数会保持到栈上。
下面这个例子将传递九个参数。可以通过它来验证一下各个寄存器的使用情况:
class Test { public: Test():number(5555){} void func9(int a, int b, int c, int d,char*str, long e, long f, float h, double i) { a++; b += 2; c += 3; d += 4; } private: int number; }; int main(int argc, char* argv[]) { Test tInst; tInst.func9(1111, 2222, 3333, 4444, "hello, world!", 6666,7777, 8.888, 9.999); return 0; }
下面通过gdb验证各个寄存器的使用情况:
khawk-dev-zhanga12:~/study/c++callstack # gdb a.out GNU gdb (GDB) SUSE (7.0-0.4.16) Copyright (C) 2009 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 "x86_64-suse-linux". For bug reporting instructions, please see: <http://www.gnu.org/software/gdb/bugs/>... Reading symbols from /root/study/c++callstack/a.out...done. (gdb) b main Breakpoint 1 at 0x40071b: file m64class.cpp, line 19. (gdb) r Starting program: /root/study/c++callstack/a.out Breakpoint 1, main (argc=1, argv=0x7fffffffe188) at m64class.cpp:19 19 Test tInst; (gdb) n 20 tInst.func9(1111, 2222, 3333, 4444, "hello, world!", 6666,7777, 8.888, 9.999); (gdb) s Test::func9 (this=0x7fffffffe0a0, a=1111, b=2222, c=3333, d=4444, str=0x400918 "hello, world!", e=6666, f=7777, h=8.88799953, i=9.9990000000000006) at m64class.cpp:8 8 a++; (gdb) info reg rax 0x400918 4196632 rbx 0x400830 4196400 rcx 0xd05 3333 rdx 0x8ae 2222 rsi 0x457 1111 rdi 0x7fffffffe0a0 140737488347296 rbp 0x7fffffffe070 0x7fffffffe070 rsp 0x7fffffffe070 0x7fffffffe070 r8 0x115c 4444 r9 0x400918 4196632 r10 0xffffffffffffffff -1 r11 0x7ffff733d890 140737340758160 r12 0x400620 4195872 r13 0x7fffffffe180 140737488347520 r14 0x0 0 r15 0x0 0 rip 0x4007ff 0x4007ff <Test::func9(int, int, int, int, char*, long, long, float, double)+35> eflags 0x202 [ IF ] cs 0x33 51 ss 0x2b 43 ds 0x0 0 es 0x0 0 fs 0x0 0 gs 0x0 0 fctrl 0x37f 895 fstat 0x0 0 ftag 0xffff 65535 fiseg 0x0 0 fioff 0x0 0 foseg 0x0 0 fooff 0x0 0 fop 0x0 0 mxcsr 0x1f80 [ IM DM ZM OM UM PM ] (gdb) p *this $1 = {number = 5555} (gdb) p *(char*)$r9@13 $2 = "hello, world!" (gdb) p *(int*)($rbp+16) $4 = 6666 (gdb) p *(int*)($rbp+24) $5 = 7777 (gdb) p *(int*)$rdi $6 = 5555
r9存储的是指针型char *str的字符串。因为寄存器只能存储6个整形、指针型参数,注意,Test 对象的指针占用了一个。因此
参数e和f只能通过栈传递。注意每个地址空间占8个字节。因此rbp+2*8存储的是e,rbp+3*8存储的是f。
float h是通过xmm0,double i是通过xmm1传递的。这类寄存器的size大小是128bits,当然128个bits可以不填满。GDB将这些寄存器看成下面这些数据的联合:
union{ float v4_float[4]; double v2_double[2]; int8_t v16_int8[16]; int16_t v8_int16[8]; int32_t v4_int32[4]; int64_t v2_int64[2]; int128_t unit128; }xmm0,xmm1,xmm2,xmm3,xmm4,xmm5,xmm6,xmm7;
打印方式如下:
(gdb) p $xmm0.v4_float[0] $7 = 8.88799953 (gdb) p $xmm1.v2_double[0] $8 = 9.9990000000000006
总结:
在x86-64中,整形和指针型参数的参数从左到右依次保存到rdi,rsi,rdx,rcx,r8,r9中。浮点型参数会保存到xmm0,xmm1……。多余的参数会保持到栈上。
5. 总结
32位:
1)默认的传递方法不使用寄存器,使用ebp+8可以访问第一个参数,ebp+16可以访问第二个参数。。。
2) 可以使用GCC扩展__attribute__将参数放到寄存器上,依次使用的寄存器是eax/edx/ecx
3)对于class 的member function,调用时第一个参数为this(对象指针)地址。因此函数的第一个参数使用ebp+16才可以访问到。
4)ebp存储的是上层的bp的地址。ebp+4存储的是函数返回时的地址指令。
64位:
1)默认的传递方法是使用寄存器。当然也可以强制使用栈,方式:函数声明时使用
__attribute__((regparm(0)))2)整形和指针型参数的参数从左到右依次保存到rdi,rsi,rdx,rcx,r8,r9中。浮点型参数会保存到xmm0,xmm1……。多余的参数会保持到栈上。
3)xmm0……是比较特殊的寄存器,访问内容时需要注意。
尊重原创,转载请注明出处: anzhsoft http://blog.csdn.net/anzhsoft/article/details/18739193