父进程无法读取来自 C 中 4 个不同管道的所有消息
Posted
技术标签:
【中文标题】父进程无法读取来自 C 中 4 个不同管道的所有消息【英文标题】：Parent process can't read all messages from 4 different pipes in C 【发布时间】：2015-06-30 04:45:30 【问题描述】：最后在这里查看并提出问题后，我能够编写这段代码，它生成 4 个子进程并通过 4 个不同的管道进行通信。 4 个子进程中的每一个都写入单个管道 (fd_log)，但是当父进程从管道中读取时，它只读取来自进程 A 的第一条消息，来自进程 C 的第一条消息，来自进程的所有消息进程B 没有来自进程D。这是代码：
int main(int argc, char *argv[])
printf("\nWritten by Nawar Youssef\n");
int i, x=0, fd_log[2], fd_A_B[2], fd_B_C[2], fd_B_D[2], pipe_size=500;
char ch, msg_from_A[pipe_size], msg_to_B[pipe_size], msg_to_log[pipe_size],
msg_to_C[pipe_size], msg_to_D[pipe_size], msg_B_C[pipe_size], msg_B_D[pipe_size],
msg_from_log[pipe_size], temp_char[pipe_size], msg_C_log[pipe_size], msg_D_log[pipe_size];
pipe(fd_log);
pipe(fd_A_B);
pipe(fd_B_C);
pipe(fd_B_D);
if (fork()==0) //child A
for (i=0; i < 10; i++)
x = (rand() % 2);
if (x == 1)
ch='C';
else
ch='D';
//write records to A-B pipe
close(fd_A_B[READ]);
sprintf(msg_to_B, "%c %d", ch, i);
write(fd_A_B[WRITE], msg_to_B, strlen(msg_to_B)+1);
//write records to log pipe
close(fd_log[READ]);
sprintf(msg_to_log, "A sent to B: %c %d", ch, i);
write(fd_log[WRITE], msg_to_log, strlen(msg_to_log)+1);
//for
close(fd_A_B[WRITE]);
close(fd_log[WRITE]);
_exit(1);
//if
if (fork()==0) //child B
//read A-B pipe
close(fd_A_B[WRITE]);
int n_bytes = read(fd_A_B[READ], msg_from_A, sizeof(msg_from_A));
for(i=0; i < pipe_size; i++)
if ( msg_from_A[i] == 'C')
//write the message from B to C pipe
sprintf(msg_to_C, "%c %c", msg_from_A[i], msg_from_A[i+2]);
//printf("--%s\n", msg_to_C);
close(fd_B_C[READ]);
write(fd_B_C[WRITE], msg_to_C, strlen(msg_to_C)+1);
//write C message to log pipe
close(fd_log[READ]);
sprintf(msg_to_log, "B sent to C: %s", msg_to_C);
write(fd_log[WRITE], msg_to_log, strlen(msg_to_log)+1);
sleep(1);
else if (msg_from_A[i] == 'D')
//write the message from B to D pipes
sprintf(msg_to_D, "%c %c", msg_from_A[i], msg_from_A[i+2]);
//printf("--%s\n", msg_to_D);
close(fd_B_D[READ]);
write(fd_B_D[WRITE], msg_to_D, strlen(msg_to_D)+1);
//write D message to log pipe
close(fd_log[READ]);
sprintf(msg_to_log, "B sent to D: %s", msg_to_D);
write(fd_log[WRITE], msg_to_log, strlen(msg_to_log)+1);
sleep(5);
else
continue;
//for
close(fd_B_C[WRITE]); close(fd_B_D[WRITE]); close(fd_log[WRITE]);
_exit(1); //process B
//if
if (fork()==0) //child C
//read from B-C pipe
close(fd_B_C[WRITE]);
int n_bytes = read(fd_B_C[READ], msg_B_C, sizeof(msg_B_C));
for (i=0; i < pipe_size; i++)
//write to log pipe
if (msg_B_C[i] == 'C')
sprintf(msg_C_log, "%c %c", msg_B_C[i], msg_B_C[i+2]);
close(fd_log[READ]);
sprintf(msg_to_log, "C sent to log: %s", msg_C_log);
write(fd_log[WRITE], msg_to_log, strlen(msg_to_log)+1);
else
continue;
_exit(1); //process C
if (fork()==0) //child D
//read from fd_B_D
close(fd_B_D[WRITE]);
int n_bytes = read(fd_B_D[READ], msg_B_D, sizeof(msg_B_D));
for (i=0; i < pipe_size; i++)
//write to log pipe
if (msg_B_D[i] == 'D')
sprintf(msg_D_log, "%c %c", msg_B_D[i], msg_B_C[i+2]);
close(fd_log[READ]);
sprintf(msg_to_log, "D sent to log: %s", msg_D_log);
write(fd_log[WRITE], msg_to_log, strlen(msg_to_log)+1);
_exit(1);
//if
//parent
close(fd_log[WRITE]);
int n_bytes;
while( (n_bytes = read(fd_log[READ], msg_from_log, sizeof(msg_from_log)-1)) > 0)
printf("Log pipe reads (%d) bytes: %s\n", n_bytes, msg_from_log);
sleep(1);
close(fd_log[READ]);
return (0);
当我查看输出时，我知道所有消息都已读取（在输出中查看返回的字节数 read）。所以看起来问题在于显示消息。我用更简单的代码遇到了类似的问题，我通过使pipe_size 更大来解决它。但在这里它不起作用。
这是（#）之间的输出数是每次while循环读取的字节数：
由纳瓦尔·优素福撰写
日志管道读取 (187) 个字节：A 发送到 B：C 0
日志管道读取 (36) 字节：C 发送到日志：C 0
日志管道读取 (17) 个字节：B 发送到 C：C 2
日志管道读取 (35) 个字节：B 发送到 D：D 3
日志管道读取 (17) 个字节：B 发送到 D：D 4
日志管道读取 (17) 个字节：B 发送到 D：D 5
日志管道读取 (17) 个字节：B 发送到 D：D 6
日志管道读取 (17) 个字节：B 发送到 D：D 7
日志管道读取 (17) 个字节：B 发送到 C：C 8
日志管道读取 (17) 个字节：B 发送到 C：C 9
【问题讨论】：
【参考方案1】：代码有点混乱，因为您刚刚决定将所有内容都放在main() 中。一些辅助功能会很有用，但我不是来判断的。
几个更重要的错误：
没有错误处理。不是一个。read(2)、write(2)、close(2)、pipe(2) 和 fork(2) 都可以返回错误。你从不检查，你应该。如果你这样做了，你会看到close(2) 在大多数情况下返回EBADF，因为你在一个循环中一遍又一遍地用相同的文件描述符调用它。例如，查看子 A 的代码。它关闭循环内的两个管道。一旦循环到达第二次迭代，close(2) 就会开始返回错误，因为您已经关闭了该描述符。
您也永远不会关闭父级中的管道写入通道。这将使read(2) 中的其他子进程处于活动状态并被阻塞，因为管道（父级）中仍然有一个可以（但不会）写入的活动写入器。您必须关闭父级中的管道写入端。在分叉将使用它们的子代之前立即在父代中创建管道总是一个好主意，并在分叉后立即关闭父代中未使用的通道。子进程也是如此。
每个孩子都应该读取循环内的管道，类似于父母对日志所做的事情。否则，您将只阅读一次并丢弃同时写入的其他消息。您必须继续阅读，直到管道的写入通道被另一端关闭（这发生在 read(2) 返回 0 时）。
i+= 3，因为每次迭代消耗 3 个字节。
else continue; 在循环结束时完全没用。
请编译您的代码并启用警告。你会看到你有一个从未使用过的变量temp_char。
孩子 D 应该从msg_B_D 而不是msg_B_C 中读取数字。我相信这是一个复制/粘贴错误。
您不能通过管道发送空终止符。您正在这样做，因此，当您将其打印到输出时，您不会看到发送到日志的每条消息 - 请记住，printf() 只打印一个字符串，直到它看到一个终止的空字符。相反，我建议您将消息发送到日志并准备好换行符，以便父级打印它而无需担心。
这是解决了所有这些错误的代码。我注释掉了 sleep 以快速获得输出：
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <errno.h>
#include <unistd.h>
#include <fcntl.h>
#include <time.h>
#define READ 0
#define WRITE 1
int main(int argc, char *argv[])
printf("\nWritten by Nawar Youssef\n");
int i, x=0, fd_log[2], fd_A_B[2], fd_B_C[2], fd_B_D[2], pipe_size=500;
char ch, msg_from_A[pipe_size], msg_to_B[pipe_size], msg_to_log[pipe_size],
msg_to_C[pipe_size], msg_to_D[pipe_size], msg_B_C[pipe_size], msg_B_D[pipe_size],
msg_from_log[pipe_size], msg_C_log[pipe_size], msg_D_log[pipe_size];
if (pipe(fd_log) < 0)
fprintf(stderr, "%d: pipe(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (pipe(fd_A_B) < 0)
fprintf(stderr, "%d: pipe(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
pid_t pid;
if ((pid = fork()) < 0)
fprintf(stderr, "%d: fork(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (pid == 0)
//child A
for (i = 0; i < 10; i++)
x = (rand() % 2);
if (x == 1)
ch = 'C';
else
ch = 'D';
sprintf(msg_to_B, "%c %d", ch, i);
size_t to_write = strlen(msg_to_B);
if (write(fd_A_B[WRITE], msg_to_B, to_write) < 0)
fprintf(stderr, "%d: write(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
sprintf(msg_to_log, "A sent to B: %c %d\n", ch, i);
to_write = strlen(msg_to_log);
if (write(fd_log[WRITE], msg_to_log, to_write) < 0)
fprintf(stderr, "%d: write(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
exit(EXIT_SUCCESS);
if (close(fd_A_B[WRITE]) < 0)
fprintf(stderr, "%d: close(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (pipe(fd_B_C) < 0)
fprintf(stderr, "%d: pipe(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (pipe(fd_B_D) < 0)
fprintf(stderr, "%d: pipe(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if ((pid = fork()) < 0)
fprintf(stderr, "%d: fork(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (pid == 0)
// Child B
int n_bytes;
while ((n_bytes = read(fd_A_B[READ], msg_from_A, sizeof(msg_from_A)-1)) > 0)
size_t to_write;
for (i = 0; i+2 < n_bytes; i += 3)
if (msg_from_A[i] == 'C')
sprintf(msg_to_C, "%c %c", msg_from_A[i], msg_from_A[i+2]);
to_write = strlen(msg_to_C);
if (write(fd_B_C[WRITE], msg_to_C, to_write) != to_write)
fprintf(stderr, "%d: write(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
sprintf(msg_to_log, "B sent to C: %s\n", msg_to_C);
to_write = strlen(msg_to_log);
if (write(fd_log[WRITE], msg_to_log, to_write) < 0)
fprintf(stderr, "%d: write(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
//sleep(1);
else if (msg_from_A[i] == 'D')
sprintf(msg_to_D, "%c %c", msg_from_A[i], msg_from_A[i+2]);
to_write = strlen(msg_to_D);
if (write(fd_B_D[WRITE], msg_to_D, to_write) != to_write)
fprintf(stderr, "%d: write(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
sprintf(msg_to_log, "B sent to D: %s\n", msg_to_D);
to_write = strlen(msg_to_log);
if (write(fd_log[WRITE], msg_to_log, to_write) < 0)
fprintf(stderr, "%d: write(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
//sleep(5);
if (n_bytes < 0)
fprintf(stderr, "%d: read(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
exit(EXIT_SUCCESS);
if ((pid = fork()) < 0)
fprintf(stderr, "%d: fork(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (close(fd_B_C[WRITE]) < 0)
fprintf(stderr, "%d: close(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (close(fd_B_D[WRITE]) < 0)
fprintf(stderr, "%d: close(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (pid == 0)
// Child C
int n_bytes;
while ((n_bytes = read(fd_B_C[READ], msg_B_C, sizeof(msg_B_C))) > 0)
for (i = 0; i+2 < n_bytes; i += 3)
if (msg_B_C[i] == 'C')
sprintf(msg_C_log, "%c %c", msg_B_C[i], msg_B_C[i+2]);
sprintf(msg_to_log, "C sent to log: %s\n", msg_C_log);
size_t to_write = strlen(msg_to_log);
if (write(fd_log[WRITE], msg_to_log, to_write) != to_write)
fprintf(stderr, "%d: write(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (n_bytes < 0)
fprintf(stderr, "%d: read(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
exit(EXIT_SUCCESS);
if ((pid = fork()) < 0)
fprintf(stderr, "%d: fork(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (pid == 0)
// Child D
int n_bytes;
while ((n_bytes = read(fd_B_D[READ], msg_B_D, sizeof(msg_B_D))) > 0)
for (i = 0; i+2 < n_bytes; i += 3)
if (msg_B_D[i] == 'D')
sprintf(msg_D_log, "%c %c", msg_B_D[i], msg_B_D[i+2]);
sprintf(msg_to_log, "D sent to log: %s\n", msg_D_log);
size_t to_write = strlen(msg_to_log);
if (write(fd_log[WRITE], msg_to_log, to_write) != to_write)
fprintf(stderr, "%d: write(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
if (n_bytes < 0)
fprintf(stderr, "%d: read(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
exit(EXIT_SUCCESS);
// Parent
if (close(fd_log[WRITE]) < 0)
fprintf(stderr, "%d: close(): %s\n", __LINE__, strerror(errno));
exit(EXIT_FAILURE);
int n_bytes;
while ((n_bytes = read(fd_log[READ], msg_from_log, sizeof(msg_from_log)-1)) > 0)
msg_from_log[n_bytes] = '\0';
printf("Log pipe reads (%d) bytes:\n%s", n_bytes, msg_from_log);
//sleep(1);
return 0;
还有很大的改进空间。您不需要这么多字符数组变量 - 请记住，内存不是在父进程和子进程之间共享的。在写入中间消息之前，您可能只有 1 或 2 个缓冲区来打印它们。但它似乎正在工作：
filipe@filipe-Kubuntu:~/dev$ ./a.out
Written by Nawar Youssef
Log pipe reads (170) bytes:
A sent to B: C 0
A sent to B: D 1
A sent to B: C 2
A sent to B: C 3
A sent to B: C 4
A sent to B: C 5
A sent to B: D 6
A sent to B: D 7
A sent to B: C 8
A sent to B: C 9
Log pipe reads (170) bytes:
B sent to C: C 0
B sent to D: D 1
B sent to C: C 2
B sent to C: C 3
B sent to C: C 4
B sent to C: C 5
B sent to D: D 6
B sent to D: D 7
B sent to C: C 8
B sent to C: C 9
Log pipe reads (57) bytes:
D sent to log: D 1
D sent to log: D 6
D sent to log: D 7
Log pipe reads (133) bytes:
C sent to log: C 0
C sent to log: C 2
C sent to log: C 3
C sent to log: C 4
C sent to log: C 5
C sent to log: C 8
C sent to log: C 9
filipe@filipe-Kubuntu:~/dev$
【讨论】：
Goncalves 这是我在这个网站上看到的最有帮助的答案。我还没有纠正错误，但只是看着答案我知道我的错误在哪里。这是我第一次用 C 编程（除了学习基础知识）并使用管道。你帮了我很多。 @NawarYoussef 很高兴我能帮上忙！作为第一次使用 C 的用户尝试管道和叉子是非常勇敢的。我知道一开始可能会很困惑，但请随时发布更多问题，我很乐意为您提供帮助。以上是关于父进程无法读取来自 C 中 4 个不同管道的所有消息的主要内容，如果未能解决你的问题，请参考以下文章