在写epoll回显服务器代码之前，可以先看看上一篇文章：select poll epoll三者之间的比较。最近在继续学习网络编程中的服务端编程中，了解到很多网游服务器是在IOMP（IO完成端口）框架下写的，但是这种方式只能在 Windows 下使用，奇了怪了，这么好的东西为什么不在Linux下也实现一套呢？这个问题我继续学习IOMP再来谈一谈！
epoll是linux下高并发服务器的完美方案，因为是基于事件触发的，所以比select快的不只是一个数量级。单线程epoll，触发量可达到15000，但是加上业务后，因为大多数业务都与数据库打交道，所以就会存在阻塞的情况，这个时候就必须用多线程来提速。业务在线程池内，这里要加锁才行。测试结果2300个/s
在Linux下，一个高效率一点的多线程并发服务器可以使用epoll来实现，我记得有一个库 libevent 好像就是在 epoll 基础上实现的。
epoll的两种工作方式：
1. LT
LT(level triggered)是 epoll 缺省的工作方式，并且同时支持 block 和 no-block socket.
在这种做法中，内核告诉你一个文件描述符是否就绪了，然后你可以对这个就绪的 fd 进行 IO 操作。如果你不作任何操作，内核还是会继续通知你的，所以，这种模式编程出错误可能性要小一点。传统的 select/poll 都是这种模型的代表。
2. ET
ET (edge-triggered)是高速工作方式，只支持no-block socket，它效率要比LT更高。
ET与LT的区别在于，当 一个新的事件到来时，ET模式下当然可以从 epoll_wait 调用中获取到这个事件，可是如果这次没有把这个事件对应的套接字缓冲区处理完，在这个套接字中没有新的事件再次到来时，在ET模式下是无法再次从epoll_wait调用中获取这个事件的。
而LT模式正好相反，只要一个事件对应的套接字缓冲区还有数据，就总能从epoll_wait中获取这个事件。
因此，LT模式下开发基于epoll的应用要简单些，不太容易出错。而在ET模式下事件发生时，如果没有彻底地将缓冲区数据处理完，则会导致缓冲区中的用户请求得不到响应。
好，废话不多说，为了方便以后使用基于 epoll 的服务器，把代码贴在这里，方便以后查找！
/************************************************************************* > File Name: epoll_echosrv.c > Author: huabo > Mail: bohua1@126.com > Created Time: Sun 05 Oct 2014 08:27:06 PM HKT ************************************************************************/ #include<stdio.h> #include <stdlib.h> #include <errno.h> #include <unistd.h> #include <fcntl.h> #include <sys/types.h> #include <netinet/in.h> #include <sys/socket.h> #include <string.h> #include <arpa/inet.h> #include <sys/epoll.h> #include <pthread.h> #define MAXSIZE 5000 #define MAXLINE 10 #define OPEN_MAX 10 #define LISTENQ 20 #define INFTIM 1000 #define ERR_EXIT(m) do { perror(m); exit(EXIT_FAILURE); }while(0) //record the socket descriptor which need to read or write. struct task { int fd; struct task *next; }; struct user_data { int fd; unsigned int n_size; char line[MAXLINE]; }; void * readtask(void *args); void * writetask(void *args); struct epoll_event ev, events[20]; int epfd; pthread_mutex_t mutex; pthread_cond_t cond1; struct task *readhead = NULL, *readtail = NULL, *writehead = NULL; void setnonblocking(int sockfd) { int opts; opts = fcntl(sockfd, F_GETFL); if(opts < 0) { ERR_EXIT("fcntl"); } opts = (opts | O_NONBLOCK); opts = fcntl(sockfd, F_SETFL, opts); if(opts < 0) { ERR_EXIT("fcntl"); } } int main() { int listenfd; //read and write thread ID pthread_t tid1, tid2; struct task *new_task = NULL; struct user_data *rdata = NULL; pthread_mutex_init(&mutex, NULL); pthread_cond_init(&cond1, NULL); //initial id of read thread and write thread pthread_create(&tid1, NULL, readtask, NULL); pthread_create(&tid2, NULL, writetask, NULL); if( (listenfd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP)) < 0) { ERR_EXIT("socket"); } setnonblocking(listenfd); struct sockaddr_in servaddr; memset(&servaddr, 0, sizeof(servaddr)); servaddr.sin_family = AF_INET; servaddr.sin_port = htons(8189); servaddr.sin_addr.s_addr = htonl(INADDR_ANY); //reuse address int on = 1; if(setsockopt(listenfd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on))) { ERR_EXIT("setsockopt"); } //bind port if(bind(listenfd, (struct sockaddr*)&servaddr, sizeof(servaddr)) < 0) { ERR_EXIT("bind"); } //listen if(listen(listenfd, SOMAXCONN) < 0) { ERR_EXIT("listen"); } struct sockaddr_in peeraddr; socklen_t peerlen = sizeof(peeraddr); int connfd; int sockfd; //epoll server epfd = epoll_create(256); //step 1: create descriptor for events struct epoll_event ev, events[5000]; ev.data.fd = listenfd; ev.events = EPOLLIN|EPOLLET; epoll_ctl(epfd, EPOLL_CTL_ADD, listenfd, &ev); //step 2: add descriptor and events int nfds; int i; while(1) { nfds = epoll_wait(epfd, events, 20, -1); for(i = 0; i < nfds; ++i) { if(events[i].data.fd == listenfd) { printf("listen = %d ", events[i].data.fd); connfd = accept(listenfd, (struct sockaddr *)(&peeraddr), &peerlen); if(connfd < 0) { ERR_EXIT("accept"); } //set non-block when use epoll setnonblocking(connfd); char *str = inet_ntoa(peeraddr.sin_addr); printf("connect from >> %s %d ", str, connfd); //add new file descriptor ev.data.fd = connfd; ev.events = EPOLLIN|EPOLLET; epoll_ctl(epfd, EPOLL_CTL_ADD, connfd, &ev); //reset the listen descriptor ev.data.fd = listenfd; ev.events = EPOLLIN|EPOLLET; epoll_ctl(epfd, EPOLL_CTL_MOD, listenfd, &ev); } else if(events[i].events & EPOLLIN) //there is data to read. { if((sockfd = events[i].data.fd) <= 0) { printf("file descriptor error "); continue; } new_task = (struct task*)malloc(sizeof(struct task)); new_task->fd = sockfd; new_task->next = NULL; //add to the read task queue. pthread_mutex_lock(&mutex); if(readhead == NULL) { readhead = new_task; readtail = new_task; } else { readtail->next = new_task; readtail = new_task; } //wake up the thread wait for cond1 pthread_cond_broadcast(&cond1); pthread_mutex_unlock(&mutex); } else if(events[i].events & EPOLLOUT) //there is data to write { //something needed to be done here. rdata = (struct user_data*)events[i].data.ptr; sockfd = rdata->fd; write(sockfd, rdata->line, rdata->n_size); free(rdata); ev.data.fd = sockfd; ev.events = EPOLLIN|EPOLLET; epoll_ctl(epfd, EPOLL_CTL_MOD, sockfd, &ev); } else if( (events[i].events & EPOLLHUP) || (events[i].events & EPOLLERR) || !(events[i].events & EPOLLIN)) { //An error has occured on this fd, or the socket is not ready for reading(why were we notified then?) fprintf(stderr, "epoll error "); close(events[i].data.fd); continue; } } } return 0; } static int count111 = 0; static time_t oldtime = 0, nowtime = 0; void* readtask(void *args) { printf("read task begin "); int fd = -1; unsigned int n; struct user_data *data = NULL; while(1) { pthread_mutex_lock(&mutex); while(readhead == NULL) pthread_cond_wait(&cond1, &mutex); fd = readhead->fd; struct task *tmp = readhead; readhead = readhead->next; free(tmp); pthread_mutex_unlock(&mutex); data = (struct user_data*)malloc(sizeof(struct user_data)); data->fd = fd; char recvBuf[1024] = {0}; int ret = 999; int rs = 1; while(rs) { ret = recv(fd, recvBuf, 1024, 0); if(ret < 0) { if(errno == EAGAIN) { printf("EAGAIN "); break; } else { printf("recv error! "); close(fd); break; } } else if(ret == 0) { printf("client close. "); rs = 0; } if(ret == sizeof(recvBuf)) { //need to recv again. rs = 1; write(fd, recvBuf, ret); fputs(recvBuf, stdout); } else { write(fd, recvBuf, ret); fputs(recvBuf, stdout); rs = 0; } } if(ret > 0) { data->n_size = n; count111++; struct tm *today; time_t ltime; time(&nowtime); if(nowtime != oldtime) { printf("%d ", count111); oldtime = nowtime; count111 = 0; } char buf[1000] = {0}; sprintf(buf, "HTTP/1.0 200 OK Content-type: text/plain %s", "Hello world! "); send(fd, buf, strlen(buf), 0); close(fd); } } } void * writetask(void *args) { }