最高效的进(线)程间通信机制: eventfd

一句话总结：用于进程/线程间通信，效率比pipe高

#include <sys/eventfd.h>
int eventfd(unsigned int initval, int flags);

常用的进程（线程）间通信机制有管道、信号量、消息队列、信号、共享内存、socket等等，其中主要作为进程（线程）间通知/等待的有管道pipe和socket。从Linux 2.6.27版本开始增加了eventfd，主要用于进程或者线程间的通信（如通知/等待机制的实现）。

eventfd()创建了一个"eventfd object"，能在用户态用做事件wait/notify机制，通过内核取唤醒用户态的事件。这个对象保存了一个内核维护的uint64_t类型的整型counter，这个counter初始值被参数initval指定，在Linux 2.6.26之前flags参数是没用的，必须指定为0。

initial：初始值

flags：

EFD_CLOEXEC： 调用exec后关闭

EFD_NONBLOCK：非阻塞模式

实例：

#include <unistd.h>#include <stdlib.h>#include <stdio.h>#include <stdint.h>#include <sys/eventfd.h>#define handle_error(msg) { perror(msg); exit(EXIT_FAILURE); }int main(int argc, char *argv[]){    //handle_error("handle_error test");    uint64_t writeCnt;    int efd = eventfd(0, EFD_NONBLOCK | EFD_CLOEXEC);    if (-1 == efd)        handle_error("eventfd");        int ret = fork();    if(ret == 0) //子进程    {        for (int j = 1; j < argc; j++) {            printf("Child writing %s to efd\n", argv[j]);            writeCnt = atoll(argv[j]); //把字符串转换成长长整型数  atoi atof atol            ssize_t res = write(efd, &writeCnt, sizeof(uint64_t));            if (res != sizeof(uint64_t))                handle_error("write efd");        }        printf("Child has completed write!\n");                exit(EXIT_SUCCESS);    }    else    //父进程    {        sleep(2);   //等待子进程执行完毕                ssize_t res = read(efd, &writeCnt, sizeof(uint64_t));        if (res != sizeof(uint64_t))            handle_error("read efd");        printf("Parent read %lu from efd\n",(uint64_t)writeCnt);        exit(EXIT_SUCCESS);    }  }

运行结果：

./eventfd 3 4 5

Child writing 3 to efd
Child writing 4 to efd
Child writing 5 to efd
Child has completed write!
Parent read 12 from end

结果12=3+4+5+初始值0

运行./eventfd

Child has completed write!
read efd: Resource temporarily unavailable