---

author: hjjdebug
date: 2026年 01月 08日 星期四 12:45:41 CST
descrip: 计算机中的符号是什么意思?

---

符号, 符号化调试, 计算机编译,连接,调试离不开符号,到底什么是符号呢?

甲. 什么叫符号?

研究符号, 也应该从hello-world 开始, 这里的符号,就是elf文件中的符号

0 下面是测试代码

$ cat test.c#include"stdio.h"intmain(){printf("hello\n");return0;}

把它编译成test 可执行文件
$ gcc -g -no-pie -o test test.c

对于名叫test 的elf 文件,

1 用 $ nm test 可列出其所有符号,非常简明

$ nm test0000000000601030B __bss_start0000000000601030b completed.76980000000000601020D __data_start0000000000601020W data_start0000000000400440t deregister_tm_clones0000000000400430T _dl_relocate_static_pie00000000004004b0 t __do_global_dtors_aux0000000000600e18t __do_global_dtors_aux_fini_array_entry0000000000601028D __dso_handle0000000000600e20d _DYNAMIC0000000000601030D _edata0000000000601038B _end0000000000400574T _fini00000000004004e0t frame_dummy0000000000600e10t __frame_dummy_init_array_entry00000000004006c4 r __FRAME_END__0000000000601000d _GLOBAL_OFFSET_TABLE_ w __gmon_start__000000000040058c r __GNU_EH_FRAME_HDR00000000004003c8 T _init0000000000600e18t __init_array_end0000000000600e10t __init_array_start0000000000400580R _IO_stdin_used0000000000400570T __libc_csu_fini0000000000400500T __libc_csu_init U __libc_start_main@@GLIBC_2.2.500000000004004e7T main U puts@@GLIBC_2.2.50000000000400470t register_tm_clones0000000000400400T _start0000000000601030D __TMC_END__

一下子多出来这么多符号, 其实从源代码中,我们只知道main 符号 和 printf 符号,
printf 在这里被puts@@GLIBC_2.2.5 替代了.
其它的先不管了.
由此看出符号的几个特点:

1. 符号首先要有名字,用以区别是这个符号还是哪个符号
   计算机中用一个4字节数表示.
   为什么是4字节?
   因为所有的名字字符串构成一个字符串节,排列在一起,这个4字节32位数,
   表示这个名字的首个字符在字符串节中的偏移位置, 4字节足够使用了

2. 符号要有值. 这是符号存在的意义.
   例如:
   00000000004004e7 T main
   其值,在64bits 机器上占用8bytes 地址值,也足够大了

3. 符号有类型. 例如这里的T,t,D,d,w,B,R 等等

2. 用 $ readelf -s test 有对符号更细致的描述.

它会列出动态符号与静态符号集合.

3. 用 $ readelf --dyn-syms test 会只列出动态符号

$ readelf--dyn-syms test Symbol table'.dynsym'contains4entries:Num:Value Size Type Bind Vis Ndx Name0:00000000000000000NOTYPE LOCAL DEFAULT UND1:00000000000000000FUNC GLOBAL DEFAULT UND puts@GLIBC_2.2.5(2)2:00000000000000000FUNC GLOBAL DEFAULT UND __libc_start_main@GLIBC_2.2.5(2)3:00000000000000000NOTYPE WEAK DEFAULT UND __gmon_start__

关注 puts(即转义的printf), 及__libc_start_main
动态符号, 其值暂时为0, 将来由loader 去确定符号具体的数值即内存偏移地址.
静态符号太长 66项,

Symbol table'.symtab'contains66entries:Num:Value Size Type Bind Vis Ndx Name0:00000000000000000NOTYPE LOCAL DEFAULT UND1:00000000004002380SECTION LOCAL DEFAULT12:00000000004002540SECTION LOCAL DEFAULT23:00000000004002740SECTION LOCAL DEFAULT34:00000000004002980SECTION LOCAL DEFAULT45:00000000004002b80SECTION LOCAL DEFAULT56:00000000004003180SECTION LOCAL DEFAULT67:00000000004003560SECTION LOCAL DEFAULT78:00000000004003600SECTION LOCAL DEFAULT89:00000000004003800SECTION LOCAL DEFAULT910:00000000004003b00SECTION LOCAL DEFAULT1011:00000000004003c80SECTION LOCAL DEFAULT1112:00000000004003e00SECTION LOCAL DEFAULT1213:00000000004004000SECTION LOCAL DEFAULT1314:00000000004005740SECTION LOCAL DEFAULT1415:00000000004005800SECTION LOCAL DEFAULT1516:000000000040058c0SECTION LOCAL DEFAULT1617:00000000004005c80SECTION LOCAL DEFAULT1718:0000000000600e100SECTION LOCAL DEFAULT1819:0000000000600e180SECTION LOCAL DEFAULT1920:0000000000600e200SECTION LOCAL DEFAULT2021:0000000000600ff00SECTION LOCAL DEFAULT2122:00000000006010000SECTION LOCAL DEFAULT2223:00000000006010200SECTION LOCAL DEFAULT2324:00000000006010300SECTION LOCAL DEFAULT2425:00000000000000000SECTION LOCAL DEFAULT2526:00000000000000000SECTION LOCAL DEFAULT2627:00000000000000000SECTION LOCAL DEFAULT2728:00000000000000000SECTION LOCAL DEFAULT2829:00000000000000000SECTION LOCAL DEFAULT2930:00000000000000000SECTION LOCAL DEFAULT3031:00000000000000000FILE LOCAL DEFAULT ABS crtstuff.c32:00000000004004400FUNC LOCAL DEFAULT13deregister_tm_clones33:00000000004004700FUNC LOCAL DEFAULT13register_tm_clones34:00000000004004b00FUNC LOCAL DEFAULT13__do_global_dtors_aux35:00000000006010301OBJECT LOCAL DEFAULT24completed.769836:0000000000600e180OBJECT LOCAL DEFAULT19__do_global_dtors_aux_fin37:00000000004004e00FUNC LOCAL DEFAULT13frame_dummy38:0000000000600e100OBJECT LOCAL DEFAULT18__frame_dummy_init_array_39:00000000000000000FILE LOCAL DEFAULT ABS test.c40:00000000000000000FILE LOCAL DEFAULT ABS crtstuff.c41:00000000004006c40OBJECT LOCAL DEFAULT17__FRAME_END__42:00000000000000000FILE LOCAL DEFAULT ABS43:0000000000600e180NOTYPE LOCAL DEFAULT18__init_array_end44:0000000000600e200OBJECT LOCAL DEFAULT20_DYNAMIC45:0000000000600e100NOTYPE LOCAL DEFAULT18__init_array_start46:000000000040058c0NOTYPE LOCAL DEFAULT16__GNU_EH_FRAME_HDR47:00000000006010000OBJECT LOCAL DEFAULT22_GLOBAL_OFFSET_TABLE_48:00000000004005702FUNC GLOBAL DEFAULT13__libc_csu_fini49:00000000006010200NOTYPE WEAK DEFAULT23data_start50:00000000000000000FUNC GLOBAL DEFAULT UND puts@@GLIBC_2.2.551:00000000006010300NOTYPE GLOBAL DEFAULT23_edata52:00000000004005740FUNC GLOBAL DEFAULT14_fini53:00000000000000000FUNC GLOBAL DEFAULT UND __libc_start_main@@GLIBC_54:00000000006010200NOTYPE GLOBAL DEFAULT23__data_start55:00000000000000000NOTYPE WEAK DEFAULT UND __gmon_start__56:00000000006010280OBJECT GLOBAL HIDDEN23__dso_handle57:00000000004005804OBJECT GLOBAL DEFAULT15_IO_stdin_used58:0000000000400500101FUNC GLOBAL DEFAULT13__libc_csu_init59:00000000006010380NOTYPE GLOBAL DEFAULT24_end60:00000000004004302FUNC GLOBAL HIDDEN13_dl_relocate_static_pie61:000000000040040043FUNC GLOBAL DEFAULT13_start62:00000000006010300NOTYPE GLOBAL DEFAULT24__bss_start63:00000000004004e723FUNC GLOBAL DEFAULT13main64:00000000006010300OBJECT GLOBAL HIDDEN23__TMC_END__65:00000000004003c80FUNC GLOBAL DEFAULT11_init

拣我们已知的. main, puts
能看懂名字的. _start, _init, __bss_start, _end, _edata, test.c, crtstuff.c,
其它就先不管了.

4 符号的构成要素

用readelf, 我们看到符号不仅有value, 而且还有size,
不仅有type, 而且还有Bind, Visibility, 还有index.
名称呢? 有的符号没有名称.
下面就要从计算机的角度来解释一下什么叫符号了.
不是含糊的概念, 而是准确的概念.
以64位cpu 为例.

1. 前面说了, name 占4个bytes
2. value 是一个8bytes 地址值
3. size 是一个8bytes, 的大小值, 例如 main 函数,占用23bytes 大小
4. index 是什么意思? 是说该符号属于哪个section, index 是section的index,用word表示,2bytes
5. 把type 和 binding 合成一个byte, type 占低4bits, bind 占高4bits

5 符号在计算机中是如下定义的.

typedefstruct{Elf64_Word st_name;/* Symbol name (string tbl index) */4字节unsignedcharst_info;/* Symbol type and binding */1字节unsignedcharst_other;/* Symbol visibility */1字节 Elf64_Section st_shndx;/* Section index */2字节 Elf64_Addr st_value;/* Symbol value */8字节 Elf64_Xword st_size;/* Symbol size */8字节}Elf64_Sym;对大小还不敢确认,看看其elf64.h 中的定义typedefuint32_tElf64_Word;typedefuint16_tElf64_Section;typedefuint64_tElf64_Addr;typedefuint64_tElf64_Xword;

6 补充1: type 的 定义

switch(sym_type) { case 0: return "NOTYPE"; case 1: return "OBJECT"; case 2: return "FUNC"; case 3: return "SECTION"; case 4: return "FILE"; case 6: return "TLS"; case 7: return "NUM"; case 10: return "LOOS"; case 12: return "HIOS"; default: return "UNKNOWN"; }

7 补充2: bind 的 定义

switch(sym_bind) { case 0: return "LOCAL"; case 1: return "GLOBAL"; case 2: return "WEAK"; case 3: return "NUM"; case 10: return "UNIQUE"; case 12: return "HIOS"; case 13: return "LOPROC"; default: return "UNKNOWN";

8 补充3: visibility 的 定义, 空间留了一个byte, 256种可能,但实际只有4种

switch(sym_vis) { case 0: return "DEFAULT"; //可见 case 1: return "INTERNAL"; case 2: return "HIDDEN"; case 3: return "PROTECTED"; default: return "UNKNOWN"; }

9 补充4: section header index, 留了2个bytes, 默认是索引号, 但也有几个特例.

switch(shdr_idx) { case SHN_ABS: return "ABS"; case SHN_COMMON: return "COM"; case SHN_UNDEF: return "UND"; case SHN_XINDEX: return "COM"; default: return std::to_string(shdr_idx); }

#define SHN_UNDEF 0 /* Undefined section/
#define SHN_ABS 0xfff1 /Associated symbol is absolute/
#define SHN_COMMON 0xfff2 /Associated symbol is common/
#define SHN_XINDEX 0xffff /Index is in extra table. */

10. 小结:

我们从形态,构造,功能方面用白话介绍了符号的概念, 希望对想了解编译,连接,调试的朋友有所帮助.
想了解重定位的概念, 请参考链接:
elf 格式 relocation 概念

参考代码: https://gitee.com/hejinjing/elf-parser.git