[LinuxServerCodes] 网络聊天室使用共享内存和多进程处理(服务器端)

使用了统一事件源管理 信号和IO
详细实现请见注释
特点:
多个进程共享同一个内存进行读写,效率快,
而且往不同区域写入是通过客户端下标 1, 2, 3 控制的(没有覆盖问题)
而且读数据不会破坏数据(没有竟态条件) 所以此临界区不需要加锁,效率高

#include <sys/socket.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <assert.h>
#include <stdio.h>
#include <unistd.h>
#include <errno.h>
#include <string.h>
#include <fcntl.h>
#include <stdlib.h>
#include <sys/epoll.h>
#include <signal.h>
#include <sys/wait.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <fcntl.h>

#define USER_LIMIT 5
#define BUFFER_SIZE 1024
#define FD_LIMIT 65535
#define MAX_EVENT_NUMBER 1024
#define PROCESS_LIMIT 65536
/*
 * 父进程工作: epoll_wait等待1. 到来的新连接,建立子进程epoll_Wait新连接.2 处理信号,3.子进程接收到数据通知父进程进行群发的信息
 * 子进程工作: epoll_wait 等待1. 新的连接上有新的数据,写入共享内存,并且通知父进程进行群发,2.接收到父进程的群发消息,发送给connfd. (父进程仅仅通知是第几个客户端),子进程通过共享内存的偏移量获得数据
 *   
 */ 

struct client_data
{
    sockaddr_in address;
    int connfd;
    pid_t pid;
    int pipefd[2];
};

static const char* shm_name = "/my_shm";
int sig_pipefd[2];
int epollfd;
int listenfd;
int shmfd;
char* share_mem = 0;
//客户端数据数据
client_data* users = 0;
// 子进程退出的时候,通过这个数组[pid]找到相应的客户端数据, pid = waitpid  然后在主进程清空对应的客户端
int* sub_process = 0;
int user_count = 0;//当前数组的长度,表示当前有多少客户端
bool stop_child = false;

int setnonblocking( int fd )
{
    int old_option = fcntl( fd, F_GETFL );
    int new_option = old_option | O_NONBLOCK;
    fcntl( fd, F_SETFL, new_option );
    return old_option;
}

void addfd( int epollfd, int fd )
{
    epoll_event event;
    event.data.fd = fd;
    event.events = EPOLLIN | EPOLLET;
    epoll_ctl( epollfd, EPOLL_CTL_ADD, fd, &event );
    setnonblocking( fd );
}

void sig_handler( int sig )
{
    int save_errno = errno;
    int msg = sig;
    send( sig_pipefd[1], ( char* )&msg, 1, 0 );
    errno = save_errno;
}

void addsig( int sig, void(*handler)(int), bool restart = true )
{
    struct sigaction sa;
    memset( &sa, '\0', sizeof( sa ) );
    sa.sa_handler = handler;
    if( restart )
    {
        sa.sa_flags |= SA_RESTART;
    }
    sigfillset( &sa.sa_mask );
    assert( sigaction( sig, &sa, NULL ) != -1 );
}

void del_resource()
{
    close( sig_pipefd[0] );
    close( sig_pipefd[1] );
    close( listenfd );
    close( epollfd );
    shm_unlink( shm_name );
    delete [] users;
    delete [] sub_process;
}

void child_term_handler( int sig )
{
    stop_child = true;
}

int run_child( int idx, client_data* users, char* share_mem )
{
//idx 是为客户端分配的下标,也就是user_count
    epoll_event events[ MAX_EVENT_NUMBER ];
    int child_epollfd = epoll_create( 5 );
    assert( child_epollfd != -1 );
    // connfd 是父进程通过listen获得的文件描述符 
    int connfd = users[idx].connfd;
    addfd( child_epollfd, connfd );
    int pipefd = users[idx].pipefd[1];
    addfd( child_epollfd, pipefd );
    int ret;
    addsig( SIGTERM, child_term_handler, false );

    while( !stop_child )
    {
        int number = epoll_wait( child_epollfd, events, MAX_EVENT_NUMBER, -1 );
        if ( ( number < 0 ) && ( errno != EINTR ) )
        {
            printf( "epoll failure\n" );
            break;
        }

        for ( int i = 0; i < number; i++ )
        {
            int sockfd = events[i].data.fd;
            if( ( sockfd == connfd ) && ( events[i].events & EPOLLIN ) )
            {
                memset( share_mem + idx*BUFFER_SIZE, '\0', BUFFER_SIZE );
                ret = recv( connfd, share_mem + idx*BUFFER_SIZE, BUFFER_SIZE-1, 0 );
                if( ret < 0 )
                {
                    if( errno != EAGAIN )
                    {
                        stop_child = true;
                    }
                }
                else if( ret == 0 )
                {
                    stop_child = true;
                }
                else
                {
                  // 向父进程发送读取成功的信息
                    send( pipefd, ( char* )&idx, sizeof( idx ), 0 );
                }
            }
            // 群发消息,获得消息的地址
            else if( ( sockfd == pipefd ) && ( events[i].events & EPOLLIN ) )
            {
                int client = 0;
                ret = recv( sockfd, ( char* )&client, sizeof( client ), 0 );
                if( ret < 0 )
                {
                    if( errno != EAGAIN )
                    {
                        stop_child = true;
                    }
                }
                else if( ret == 0 )
                {
                    stop_child = true;
                }
                else
                {
                  // 从古拿到获取client的编号,就找到了在共享内存中的偏移量,然后将内容发送给connfd
                    send( connfd, share_mem + client * BUFFER_SIZE, BUFFER_SIZE, 0 );
                }
            }
            else
            {
                continue;
            }
        }
    }

    close( connfd );
    close( pipefd );
    close( child_epollfd );
    return 0;
}

int main( int argc, char* argv[] )
{
    if( argc <= 2 )
    {
        printf( "usage: %s ip_address port_number\n", basename( argv[0] ) );
        return 1;
    }
    const char* ip = argv[1];
    int port = atoi( argv[2] );

    int ret = 0;
    struct sockaddr_in address;
    bzero( &address, sizeof( address ) );
    address.sin_family = AF_INET;
    inet_pton( AF_INET, ip, &address.sin_addr );
    address.sin_port = htons( port );

    listenfd = socket( PF_INET, SOCK_STREAM, 0 );
    assert( listenfd >= 0 );

    ret = bind( listenfd, ( struct sockaddr* )&address, sizeof( address ) );
    assert( ret != -1 );

    ret = listen( listenfd, 5 );
    assert( ret != -1 );

    user_count = 0;
    users = new client_data [ USER_LIMIT+1 ];
    sub_process = new int [ PROCESS_LIMIT ];
    for( int i = 0; i < PROCESS_LIMIT; ++i )
    {
        sub_process[i] = -1;
    }

    epoll_event events[ MAX_EVENT_NUMBER ];
    epollfd = epoll_create( 5 );
    assert( epollfd != -1 );
    addfd( epollfd, listenfd );

    ret = socketpair( PF_UNIX, SOCK_STREAM, 0, sig_pipefd );
    assert( ret != -1 );
    setnonblocking( sig_pipefd[1] );
    addfd( epollfd, sig_pipefd[0] );

    addsig( SIGCHLD, sig_handler );
    addsig( SIGTERM, sig_handler );
    addsig( SIGINT, sig_handler );
    addsig( SIGPIPE, SIG_IGN );
    bool stop_server = false;
    bool terminate = false;

    shmfd = shm_open( shm_name, O_CREAT | O_RDWR, 0666 );
    assert( shmfd != -1 );
    ret = ftruncate( shmfd, USER_LIMIT * BUFFER_SIZE ); 
    assert( ret != -1 );

    share_mem = (char*)mmap( NULL, USER_LIMIT * BUFFER_SIZE, PROT_READ | PROT_WRITE, MAP_SHARED, shmfd, 0 );
    assert( share_mem != MAP_FAILED );
    close( shmfd );

    while( !stop_server )
    {
        int number = epoll_wait( epollfd, events, MAX_EVENT_NUMBER, -1 );
        if ( ( number < 0 ) && ( errno != EINTR ) )
        {
            printf( "epoll failure\n" );
            break;
        }

        for ( int i = 0; i < number; i++ )
        {
            int sockfd = events[i].data.fd;
            if( sockfd == listenfd )
            {
                struct sockaddr_in client_address;
                socklen_t client_addrlength = sizeof( client_address );
                int connfd = accept( listenfd, ( struct sockaddr* )&client_address, &client_addrlength );
                if ( connfd < 0 )
                {
                    printf( "errno is: %d\n", errno );
                    continue;
                }
                if( user_count >= USER_LIMIT )
                {
                    const char* info = "too many users\n";
                    printf( "%s", info );
                    send( connfd, info, strlen( info ), 0 );
                    close( connfd );
                    continue;
                }
                users[user_count].address = client_address;
                users[user_count].connfd = connfd;
                // pipefd 变成socketpair
                ret = socketpair( PF_UNIX, SOCK_STREAM, 0, users[user_count].pipefd );
                assert( ret != -1 );
                pid_t pid = fork();
                //fork wrong
                if( pid < 0 )
                {
                    close( connfd );
                    continue;
                }
                // child process
                else if( pid == 0 )
                {
                    close( epollfd );
                    close( listenfd );
                    close( users[user_count].pipefd[0] );
                    close( sig_pipefd[0] );
                    close( sig_pipefd[1] );
                    // 首次user_count = 0
                    run_child( user_count, users, share_mem );
                    munmap( (void*)share_mem,  USER_LIMIT * BUFFER_SIZE );
                    exit( 0 );
                }
                else
                {
                    close( connfd );
                    close( users[user_count].pipefd[1] );
                    //注册 新连接的有数据的事件
                    addfd( epollfd, users[user_count].pipefd[0] );
                    // users[] 
                    users[user_count].pid = pid;
                    // 该子进程的 通过[pid] 找到连接的客户 首次是0
                    sub_process[pid] = user_count;
                    user_count++;
                }
            }
            else if( ( sockfd == sig_pipefd[0] ) && ( events[i].events & EPOLLIN ) )
            {
                int sig;
                char signals[1024];
                ret = recv( sig_pipefd[0], signals, sizeof( signals ), 0 );
                if( ret == -1 )
                {
                    continue;
                }
                else if( ret == 0 )
                {
                    continue;
                }
                else
                {
                    for( int i = 0; i < ret; ++i )
                    {
                      // 遍历数组来获得触发的信号
                        switch( signals[i] )
                        {
                            case SIGCHLD:
                            {
                                pid_t pid;
                                int stat;
                                while ( ( pid = waitpid( -1, &stat, WNOHANG ) ) > 0 )
                                {
                                  //通过子进程数组就可以找到删除的客户端编号/
                                    int del_user = sub_process[pid];
                                    //解除删除的 进程数组对应关系
                                    sub_process[pid] = -1;
                                    if( ( del_user < 0 ) || ( del_user > USER_LIMIT ) )
                                    {
                                        printf( "the deleted user was not change\n" );
                                        continue;
                                    }
                                    epoll_ctl( epollfd, EPOLL_CTL_DEL, users[del_user].pipefd[0], 0 );
                                    close( users[del_user].pipefd[0] );
                                        //用最后一个客户端填补 删除的客户端(保证数组下标连续)
                                    users[del_user] = users[--user_count];
                                    // del_user 已经更新 为 最后一个客户端的 中的pid, 此pid应该指向 del_user
                                    sub_process[users[del_user].pid] = del_user;
                                    printf( "child %d exit, now we have %d users\n", del_user, user_count ); 

                                }
                                if( terminate && user_count == 0 )
                                {
                                    stop_server = true;
                                }
                                break;
                            }
                            case SIGTERM:
                            case SIGINT:
                            {
                                printf( "kill all the clild now\n" );
                                //addsig( SIGTERM, SIG_IGN );
                                //addsig( SIGINT, SIG_IGN );
                                if( user_count == 0 )
                                {
                                    stop_server = true;
                                    break;
                                }
                                for( int i = 0; i < user_count; ++i )
                                {
                                    int pid = users[i].pid;
                                    kill( pid, SIGTERM );
                                }
                                terminate = true;
                                break;
                            }
                            default:
                            {
                                break;
                            }
                        }
                    }
                }
            }
            //  子进程通知父进程读取了数据,父进程就要开始向其他进程报告 该子进程的下标,通知其他进程来读取数据
            else if( events[i].events & EPOLLIN )
            {
                int child = 0;
                ret = recv( sockfd, ( char* )&child, sizeof( child ), 0 );
                printf( "read data from child accross pipe\n" );
                if( ret == -1 )
                {
                    continue;
                }
                else if( ret == 0 )
                {
                    continue;
                }
                else
                {
                  // 因为数据是共享的,所以只需要通知其他客户端 读取的位置
                    for( int j = 0; j < user_count; ++j )
                    {
                      // 每个客户端的piepefd[0] 都被epoll_wait注册过了,因为每个0 都是向服务器发送数据到来的消息
                        if( users[j].pipefd[0] != sockfd )
                        {
                            printf( "send data to child accross pipe\n" );
                            send( users[j].pipefd[0], ( char* )&child, sizeof( child ), 0 );
                        }
                    }
                }
            }
        }
    }
    del_resource();
    return 0;
}