操作系统--进程间通信(一)

IPC 运用范畴

avatar

配图与本文无关,纯属有趣

相对于做点什么,我们更想睡懒觉,但是相对于睡觉,我想写总结!
今天把博客的 git 分支搞混了,浪费了一点时间。那么今天介绍一下进程间通信(IPC) 。进程间通信,我们从 << UNIX网络编程卷2:进程间通信 >> 最后一章 Sun RPC 谈起,那为什么呢? 因为我们去掌握一个东西,肯定是因为它有价值,我们才去学习,如果一些技术现在都不用了,淘汰了,那我们就没有必要去深究,只需学习其思想,浅尝辄止即可。

最后一张概述部分是这样介绍的,构筑一个应用程序时,我们首先在以下两者之间作出选择:

(1) 构筑一个庞大的单一程序,完成全部工作;
(2) 把整个应用程序散步到彼此通信的多个进程中。
如果我们选择后者,接下去的抉择是:

(2a) 假设所有进程运行在同一台主机上 (允许 IPC 用于这些进程间的通信);
(2b) 假设某些进程会运行在其他主机上 (要求使用进程间某种形式的网络通信)。

这本著作主要关注的是 2a 这种情况,也就是 使用消息传递共享内存、并可能使用某种形式的同步来进行同一台主机上的进程间 IPC。同一进程内不同线程间的 IPC 以及不同进程内各个线程间的 IPC只是这种情形的特殊情况。

而对于不同部分之间需要网络通信的应用程序则大多数使用 显示网络编程explicit network programming)方式编写,也就是我们现在流行的 Socket编程。

那么现在是什么情况呢? 由于互联网大火,导致现在主流热门的进程间通信方式变成了使用 Socket 套接字通信。这个也可以理解,由于互联网就是传统的 C/S(服务器端-客户端)架构,对于客户端开发来说,主要就是移动应用开发、PC端应用开发、Web 浏览器开发,这些是互联网主要的流量入口,应用和浏览器需要通信的对象就是服务器,服务器要做的也是和客户端更好的进行通信,所以如果需要从事互联网的工作,那么对 Socket 编程相对于其他 IPC 手段就需要更熟悉,特别是服务器后台开发,客户端还好,客户端业务逻辑和 UI 体验相对重要,至于一些系统层技术,开发框架都已经封装的很好了,只需掌握使用方法就可以,像androidIOS开发。但是如果你是从事系统开发,那么所有的 IPC 方式都需要了解,特别是消息传递共享内存需要深入了解。

具体的 IPC 手段

信号

软件中断通知事件处理 例如:SIGFPE, SIGKILL, SIGUSR1, SIGSTOP, SIGCONT
接收到信号会发生什么:

  • Catch:指定信号处理函数被调用
  • Ignore:依靠操作系统的默认操作
    example:Abort,memory dump,suspend or resume process
  • Mask:闭塞信号,因此不会发生
    可能是暂时的(处理同样类型的信号)

不足:不能传输要交换的任何数据

特点:

  • signals are only delivered between processes when either: 仅在同一个用户的进程间传递,或者超级用户发出信号
    1. the processes involved are run by the same user
    2. the raising process is run by a superuser
  • signal types and type names are defined in signal.h 定义在 signal.h 文件
  • signals are not queued 信号没有排队
  • handlers “terminate process” (SIG_DFL) and “ignore signal” (SIG_IGN) are already defined in signal.h 常用终止进程和忽略信号定义在 signal.h

unix 主要接口信息:

1
2
3
4
5
6
7
8
9
10
#include: signal.h

functions:void raise(int sig_type)
description:用信号sig_type发信号。

functions:void signal(int sig_type, void (*sig_handler)(int signal_type))
description:sig_type是signal.h中定义的信号值(或名称)之一。 sig_handles是一个指向带有int参数的函数的指针。当进程被指定类型的信号命中时,此函数将在进程上下文中运行。

functions:int kill(pid_t dest_pid, int sig_type)
description:向使用PID dest_pid运行的进程发送sig_type类型的信号。 dest_pid的特定类型0和-1 make kill()分别发送与发送方的同一组执行的所有进程,并发信号通知系统中运行的所有进程(仅适用于超级用户凭据)。

管道和FIFO

1.管道,通常指无名管道,是 UNIX 系统IPC最古老的形式
特点:

  • pipes are uni-directional 半双工
  • with file descriptors, they can only be shared between processes with parent-child relationship 只能用于父子进程间
  • atomicity is guaranteed for messages shorter than PIPE_BUF bytes 特定大小缓存保持原子性
  • alf-open pipes are told widowed (or broken) pipes. Writing on them causes a write error, reading from them always returns 0 bytes. 坏的管道,会出现写错误,读取时,返回0字节.

unix 主要接口信息:

1
2
3
4
#include: unistd.h
functions: int pipe(int *fd_couple)
description:创建一个管道并将其文件描述符存储到fd_couple [0](读取结束)和fd_couple [1](写入结束)。
return: 成功返回 0,错误返回 -1

模拟使用场景,实现 Linux 系统调用 popen,主要代码:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
FILE * vpopen(const char * cmdstring, const char * type)
{
int pfd[2];
FILE * fp;
pid_t pid;

if ((type[0] != 'r' && type[0] != 'w') || type[1] != 0)
{
errno = EINVAL;
return (NULL);
}

if(pipe(pfd) != 0)
{
return NULL;
}

if((pid = vfork()) < 0) // vfork 相对于fork区别是: vfork需要等待子进程执行完后才开始执行,两者共享地址空间
{
return NULL;
}
else if(pid == 0) // 子进程
{
if(*type == 'r')
{
close(pfd[0]);
if (pfd[1] != STDOUT_FILENO)
{
dup2(pfd[1], STDOUT_FILENO);
close(pfd[1]);
}
}
else
{
close(pfd[1]);
if (pfd[0] != STDIN_FILENO)
{
dup2(pfd[0], STDIN_FILENO);
close(pfd[0]);
}
}

if(execl("/bin/bash", "bash", "-c", cmdstring, (char *) 0) < 0)
{
return NULL;
}
}

wait(0); // 等子进程结束,回收子进程

if (*type == 'r')
{
close(pfd[1]);
if ((fp = fdopen(pfd[0], type)) == NULL)
{
return (NULL);
}
}
else
{
close(pfd[0]);
if ((fp = fdopen(pfd[1], type)) == NULL)
{
return (NULL);
}
}

return (fp);
}

2.有名管道,FIFO,
特点:

  • 每个FIFO有一个路径名与之关联,允许无亲缘关系进程间的通信
  • 拥有标准管道的属性和特征

unix 主要接口信息:

1
2
3
4
*  #include: sys/types.h and sys/stat.h
* functions: int mkfifo(char *path, mode_t mode)
* description:path是要创建的FIFO文件的(路径+)名称。 mode是文件权限(请参阅umask(2)和chmod(2))
* return: 成功返回 0,错误返回非 0

模拟使用场景:

1
2
3
4
5
6
7
* 模拟场景
* 1. 在 shell 端使用命令创建有名管道: mkfifo /tmp/named_pipe
* 2. 运行该接口
* 3. 在 shell 端写入数据到管道 :echo "hey, this is a message" > /tmp/named_pipe
* 4. 测试完毕,删除管道: rm /tmp/named_pipe
*
* 当然我们也可以自己启动一个任意进程向FIFO管道里面写数据

具体实现代码:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
int fifo_read()
{
int pipe;
char ch;
static char TmpCh;

/* we expect a named pipe has been created
* in /tmp/named_pipe . The command
* $ mkfifo /tmp/named_pipe
* is a friend to get to this point
*/

pipe = open("/tmp/named_pipe", O_RDONLY);
if (pipe < 0) {
printf("open fifo error, maybe you should create a fifo first !\n");
exit(1);
}

/* preparing to read from the pipe... */
printf("Waiting data from the pipe... \n");

/* reading one char a time from the pipe */
while (1) {
if (read(pipe, &ch, 1) < 0) {
printf("Read the pipe failed! \n");
exit(2);
}


if(ch != TmpCh)
{
printf("%c", ch);
}

TmpCh = ch;
}

/* leaving the pipe */
close(pipe);

return 0;
}

总结

上述主要对IPC做了一个大体的介绍,以及介绍了信号(signal)和管道两种IPC方式,这两者在Linux系统自带功能实现中用的较多。至于 PosixSystem V的进程间通信方式 消息队列信号共享内存,由于内容较多后面再单独介绍。