Linux Process Address Space

  high address   +---------------+
                 |               |
                 |     Stack     |    int local_b
                 |               |
                 +---------------+
                 |       |       |
                 |       v       |
                 |               |
                 |               |
                 |       ^       |
                 |       |       |
                 +---------------+
                 |               |
                 |     Heap      |    int * heap_c = malloc()
                 |               |
                 +---------------+
                 |     Data      |    int global_a
                 +---------------+
                 |     Code      |
  low address    +---------------+

上图是 Linux 的进程地址空间，从低位到高位地址分别为：

Code Segment: 程序的代码，CPU 执行的指令部分，共享只读
Data Segment: 可细分为初始化数据段和未初始化数据段，常用于存储全局变量等
Stack: 函数以及自动变量(未加 static 的自动变量又称为局部变量)
Heap: 动态分配内存，如 malloc() 分配的内存

更为详细的介绍请见 Anatomy of a Program in Memory。

Fork

                  Parent Process             Child Process

  high address   +---------------+          +---------------+
                 |               |          |               |
                 |     Stack     |          |     Stack     |
                 |               |          |               |
                 +---------------+          +---------------+
                 |       |       |          |       |       |
                 |       v       |          |       v       |
                 |               |          |               |
                 |               |          |               |
                 |       ^       |          |       ^       |
                 |       |       |          |       |       |
                 +---------------+          +---------------+
                 |               |          |               |
                 |     Heap      |          |     Heap      |
                 |               |          |               |
                 +---------------+          +---------------+
                 |     Data      |          |     Data      |
                 +---------------+----------+---------------+
                 |                   Code                   |
  low address    +------------------------------------------+

fork 是 linux 中最重要的系统调用之一，用于创建一个新进程，它完全的复制父进程地址空间的 data segment、 heap 和 stack，但是和父进程共享一个 code segment，因为 code segment 通常为只读，从逻辑的角度来看，子进程和父进程的内存地址空间互相独立，子进程修改自己的 data segment，heap 和 stack 并不影响父进程内存空间。每次调用 fork，返回两次结果，其中父进程的返回值为子进程的 pid，子进程的返回值为 0。

#include<stdio.h>
#include<stdlib.h>
#include <unistd.h>

int global_a = 0;   // data segment

int main(void)
{
    int local_b = 0, status;    // stack
    int * heap_c = malloc(sizeof(int));     // heap
    pid_t pid;

    if(!fork()){
        global_a ++;
        local_b ++;
        *heap_c = 1;
        exit(0);
    }
    else{
        wait(&status);
    }

    printf("global_a = %d, local_b = %d, heap_c = %d\n", global_a, local_b, *heap_c);
}

程序的输出结果如下：

$ ./a.out
global_a = 0, local_b = 0, heap_c = 0

Note：为了减轻 fork 调用开销，实际采用 copy on write(COW) 技术。

Properties shared by both parent and child process

虽然父进程和子进程的地址空间是独立的，但是二者依旧共享很多其它的资源，以下摘自 Advanced Programming in the UNIX Environment, 3rd Edition：

File descriptor
Real user ID, real group ID, effective user ID, effective group ID
Supplementary group IDs
Process group ID
Session ID
Controlling terminal
The set-user-ID and set-group-ID flags
Current working directory
Root directory
File mode creation mask
Signal mask and dispositions
The close-on-exec flag for any open file descriptors
Environment
Attached shared memory segments
Memory mappings
Resource limits