Linux IPC实践(2) --匿名PIPE

管道概念

管道是Unix中最古老的进程间通信的形式,我们把从一个进程连接到另一个进程的一个数据流称为一个“管道”, 管道的本质是固定大小的内核缓冲区;

如:ps aux | grep httpd | awk '{print $2}'

管道限制

1)管道是半双工的，数据只能向一个方向流动；需要双方通信时，需要建立起两个管道;

2)匿名管道只能用于具有共同祖先的进程(如父进程与fork出的子进程)之间进行通信, 原因是pipe创建的是两个文件描述符, 不同进程直接无法直接获得;[通常，一个管道由一个进程创建，然后该进程调用fork，此后父子进程共享该管道]

匿名管道pipe

#include <unistd.h>
int pipe(int pipefd[2]);

创建一无名管道

参数

Pipefd:文件描述符数组,其中pipefd[0]表示读端,pipefd[1]表示写端

管道创建示意图

/**示例: 从子进程向父进程发送数据
管道示意图如上面第二副图
**/
int main()
{int fd[2];if (pipe(fd) == -1)err_exit("pipe error");pid_t pid = fork();if (pid == -1)err_exit("fork error");if (pid == 0)   //子进程: 向管道中写入数据{close(fd[0]);   //关闭读端string str("message from child process!");write(fd[1], str.c_str(), str.size());  //向写端fd[1]写入数据close(fd[1]);exit(EXIT_SUCCESS);}//父进程: 从管道中读出数据close(fd[1]);   //关闭写端char buf[BUFSIZ] = {0};read(fd[0], buf, sizeof(buf));close(fd[0]);cout << buf << endl;
}

/**示例: 用管道模拟: ls | wc -w的运行
1.子进程运行ls
2.父进程运行wc -w
3.通过管道, 将子进程的输出发送到wc的输入
**/
int main()
{int pipefd[2];if (pipe(pipefd) == -1)err_exit("pipe error");pid_t pid = fork();if (pid == -1)err_exit("fork error");if (pid == 0)   //子进程{close(pipefd[0]);   //关闭读端//使得STDOUT_FILENO也指向pipefd[1],亦即ls命令的输出将打印到管道中dup2(pipefd[1], STDOUT_FILENO); //此时可以关闭管道写端close(pipefd[1]);execlp("/bin/ls", "ls", NULL);//如果进程映像替换失败,则打印下面出错信息cerr << "child execlp error" << endl;;exit(EXIT_FAILURE);}//父进程close(pipefd[1]);   //关闭写端//使得STDIN_FILENO也指向pipefd[2],亦即wc命令将从管道中读取输入dup2(pipefd[0], STDIN_FILENO);close(pipefd[0]);execlp("/usr/bin/wc", "wc", "-w", NULL);cerr << "parent execlp error" << endl;exit(EXIT_FAILURE);
}

匿名管道读写规则

规则 1)管道空时

O_NONBLOCK disable：read调用阻塞，即进程暂停执行，一直等到有数据来到为止。

O_NONBLOCK enable：read调用返回-1，errno值为EAGAIN。

//验证
int main()
{int pipefd[2];if (pipe(pipefd) != 0)err_exit("pipe error");pid_t pid = fork();if (pid == -1)err_exit("fork error");if (pid == 0)   //In Child, Write pipe{sleep(10);close(pipefd[0]);   //Close Read pipestring str("I Can Write Pipe from Child!");write(pipefd[1],str.c_str(),str.size());    //Write to pipeclose(pipefd[1]);exit(EXIT_SUCCESS);}//In Parent, Read pipeclose(pipefd[1]);   //Close Write pipechar buf[1024] = {0};//Set Read pipefd UnBlock! 查看在下面四行语句注释的前后有什么区别
//    int flags = fcntl(pipefd[0],F_GETFL, 0);
//    flags |= O_NONBLOCK;
//    if (fcntl(pipefd[0],F_SETFL,flags) == -1)
//        err_exit("Set UnBlock error");int readCount = read(pipefd[0],buf,sizeof(buf));    //Read from pipeif (readCount < 0)//read立刻返回,不再等待子进程发送数据err_exit("read error");cout << "Read from pipe: " << buf << endl;close(pipefd[0]);
}

规则 2)管道满时

O_NONBLOCK disable： write调用阻塞，直到有进程读走数据

O_NONBLOCK enable：调用返回-1，errno值为EAGAIN

/** 验证规则2)
同时测试管道的容量
**/
int main()
{if (signal(SIGPIPE, handler) == SIG_ERR)err_exit("signal error");int pipefd[2];if (pipe(pipefd) != 0)err_exit("pipe error");// 将管道的写端设置成为非阻塞模式// 将下面三行注释之后查看效果int flags = fcntl(pipefd[1], F_GETFL, 0);if (fcntl(pipefd[1], F_SETFL, flags|O_NONBLOCK) == -1)err_exit("fcntl set error");int count = 0;while (true){if (write(pipefd[1], "A", 1) == -1){cerr << "write pipe error: " << strerror(errno) << endl;break;}++ count;}cout << "pipe size = " << count << endl;
}

3)如果所有管道写端对应的文件描述符被关闭，则read返回0

//验证规则3)
int main()
{int pipefd[2];if (pipe(pipefd) != 0)err_exit("pipe error");pid_t pid = fork();if (pid == -1)err_exit("fork error");else if (pid == 0){close(pipefd[1]);exit(EXIT_SUCCESS);}close(pipefd[1]);sleep(2);char buf[2];if (read(pipefd[0], buf, sizeof(buf)) == 0)cout << "sure" << endl;
}

4)如果所有管道读端对应的文件描述符被关闭，则write操作会产生信号SIGPIPE

//验证规则4)
int main()
{if (signal(SIGPIPE, handler) == SIG_ERR)err_exit("signal error");int pipefd[2];if (pipe(pipefd) != 0)err_exit("pipe error");pid_t pid = fork();if (pid == -1)err_exit("fork error");else if (pid == 0){close(pipefd[0]);exit(EXIT_SUCCESS);}close(pipefd[0]);sleep(2);char test;if (write(pipefd[1], &test, sizeof(test)) < 0)err_exit("write error");
}

Linux PIPE特征

1)当要写入的数据量不大于PIPE_BUF时，Linux将保证写入的原子性。

2)当要写入的数据量大于PIPE_BUF时，Linux将不再保证写入的原子性。

man说明:

POSIX.1-2001 says that write(2)s of less than PIPE_BUF bytes must be atomic:

the output data is written to the pipe as a contiguous sequence.

Writes of more than PIPE_BUF bytes may be nonatomic:

the kernel may interleave the data with data written by other processes.

POSIX.1-2001 requires PIPE_BUF to be at least 512 bytes.

(On Linux, PIPE_BUF is 4096 bytes. 在Linux当中, PIPE_BUF为4字节).

The precise semantics depend on whether the file descriptor is non-blocking(O_NONBLOCK),

whether there are multiple writers to the pipe, and on n, the number of bytes to be written:

O_NONBLOCK disabled(阻塞), n <= PIPE_BUF

All n bytes are written atomically; write(2) may block if there is not room for n bytes to be written immediately

O_NONBLOCK enabled(非阻塞), n <= PIPE_BUF

If there is room to write n bytes to the pipe, then write(2) succeeds immediately, writing all n bytes;

otherwise write(2) fails, with errno set to EAGAIN(注意: 如果空间不足以写入数据, 则一个字节也不写入, 直接出错返回).

O_NONBLOCK disabled, n > PIPE_BUF

The write is nonatomic: the data given to write(2) may be interleaved with write(2)s by other process;

the write(2) blocks until n bytes have been written.

O_NONBLOCK enabled, n > PIPE_BUF

If the pipe is full, then write(2) fails, with errno set to EAGAIN(此时也是没有一个字符写入管道).

Otherwise, from 1 to n bytes may be written (i.e., a "partial write" may occur;

the caller should check the return value from write(2) to see how many bytes were actually written),

and these bytes may be interleaved with writes by other processes.

/** 验证:
已知管道的PIPE_BUF为4K, 我们启动两个进程A, B向管道中各自写入68K的内容, 然后我们以4K为一组, 查看管道最后一个字节的内容, 多运行该程序几次, 就会发现这68K的数据会有交叉写入的情况
**/
int main()
{const int TEST_BUF = 68 * 1024; //设置写入的数据量为68Kchar bufA[TEST_BUF];char bufB[TEST_BUF];memset(bufA, 'A', sizeof(bufA));memset(bufB, 'B', sizeof(bufB));int pipefd[2];if (pipe(pipefd) != 0)err_exit("pipe error");pid_t pid;if ((pid = fork()) == -1)err_exit("first fork error");else if (pid == 0)  //第一个子进程A, 向管道写入bufA{close(pipefd[0]);int writeBytes = write(pipefd[1], bufA, sizeof(bufA));cout << "A Process " << getpid() << ", write "<< writeBytes << " bytes to pipe" << endl;exit(EXIT_SUCCESS);}if ((pid = fork()) == -1)err_exit("second fork error");else if (pid == 0)  //第二个子进程B, 向管道写入bufB{close(pipefd[0]);int writeBytes = write(pipefd[1], bufB, sizeof(bufB));cout << "B Process " << getpid() << ", write "<< writeBytes << " bytes to pipe" << endl;exit(EXIT_SUCCESS);}// 父进程close(pipefd[1]);sleep(2);   //等待两个子进程写完char buf[4 * 1024]; //申请一个4K的bufint fd = open("save.txt", O_WRONLY|O_TRUNC|O_CREAT, 0666);if (fd == -1)err_exit("file open error");while (true){int readBytes = read(pipefd[0], buf, sizeof(buf));if (readBytes == 0)break;if (write(fd, buf, readBytes) == -1)err_exit("write file error");cout << "Parent Process " << getpid() << " read " << readBytes<< " bytes from pipe, buf[4095] = " << buf[4095] << endl;}
}

附-管道容量查询

man 7 pipe

注意: 管道的容量不一定就等于PIPE_BUF, 如在Ubuntu中, 管道容量为64K, 而PIPE_BUF为4K.