菜鸟学习Nginx之HTTP会话

版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/xxb249/article/details/85071880

从本篇开始，介绍Nginx中HTTP相关。Nginx做为Web Server那么HTTP必然是重中之重。本篇不打算介绍太深入，本篇最主要内容是如何与事件驱动关联起来。

一、监听端口

Nginx默认监听端口是80端口，那么Nginx是如何将80端口的listening事件注册到事件驱动（epoll）中呢？这里简单回顾一下，具体内容可参考《菜鸟学习nginx之核心模块ngx_events_module》。

worker进程的进程主函数是ngx_worker_process_cycle。进入该函数会先调用函数ngx_worker_process_init进行初始化流程，当ngx_worker_process_init返回之后则进入无限循环即投入服务。那么在ngx_worker_process_init函数就会把listening socket加入到事件驱动epoll中。然而真正加入到epoll对象中是函数ngx_event_process_init函数，部分代码如下：

    /**
     * for each listening socket 
     * 循环遍历listening 将listen socket 与connection对象进行绑定
     */
    ls = cycle->listening.elts;
    for (i = 0; i < cycle->listening.nelts; i++)
    {

#if (NGX_HAVE_REUSEPORT)
        if (ls[i].reuseport && ls[i].worker != ngx_worker)
        {
            continue;
        }
#endif
        /* 返回可用连接对象 这里listening socket也会占用一个connection对象 */
        c = ngx_get_connection(ls[i].fd, cycle->log);

        if (c == NULL)
        {
            return NGX_ERROR;
        }

        c->type = ls[i].type;
        c->log = &ls[i].log;

        c->listening = &ls[i];
        ls[i].connection = c;

        rev = c->read;

        rev->log = c->log;
        rev->accept = 1; /* 表示当前读事件是accept事件 用于区别正常数据报文读取还是Accept事件读取 */

#if (NGX_HAVE_DEFERRED_ACCEPT)
        rev->deferred_accept = ls[i].deferred_accept;
#endif

        if (!(ngx_event_flags & NGX_USE_IOCP_EVENT))
        {
            if (ls[i].previous)
            {

                /*
                 * delete the old accept events that were bound to
                 * the old cycle read events array
                 */

                old = ls[i].previous->connection;

                if (ngx_del_event(old->read, NGX_READ_EVENT, NGX_CLOSE_EVENT) == NGX_ERROR)
                {
                    return NGX_ERROR;
                }

                old->fd = (ngx_socket_t)-1;
            }
        }

#if (NGX_WIN32)

        ...

#else

        rev->handler = (c->type == SOCK_STREAM) ? ngx_event_accept
                                                : ngx_event_recvmsg;

#if (NGX_HAVE_REUSEPORT)

        if (ls[i].reuseport)
        {
            if (ngx_add_event(rev, NGX_READ_EVENT, 0) == NGX_ERROR)
            {
                return NGX_ERROR;
            }

            continue;
        }

#endif

        if (ngx_use_accept_mutex)
        {
            continue;
        }

        //注册读事件 主要是用于listen监听
        if (ngx_add_event(rev, NGX_READ_EVENT, 0) == NGX_ERROR)
        {
            return NGX_ERROR;
        }

#endif
    }

循环遍历listenings数组，通过ngx_add_event函数，逐个将listening socket加入到epoll对象中且设置事件类型为READ事件。

至此Nginx完成了listening socket注册到epoll事件驱动中。当客户端发来请求时就会调用回调函数--ngx_event_accept。

二、Accept事件

注册listening socket到epoll的流程中有这样一行代码，用于设置回调函数，具体如下：

/* TCP连接的回调函数是ngx_event_accept，其他的回调函数则是ngx_event_recv_msg */
rev->handler = (c->type == SOCK_STREAM) ? ngx_event_accept
                                                : ngx_event_recvmsg;

这里使用的协议是TCP，所以当客户端的连接请求到来时，会触发回调函数ngx_event_accept函数执行，具体如何执行可参考《菜鸟学习nginx之事件模块epoll（1）》。

2.1、流程图

从流程图中可以看出，这部分代码逻辑还是比较繁琐，下面我们分片进行分析解读。

2.2、超时处理 

/**
 * 处理Accept事件
 * @param ev 待读事件
 */
void
ngx_event_accept(ngx_event_t *ev)
{
    socklen_t          socklen;
    ngx_err_t          err;
    ngx_log_t         *log;
    ngx_uint_t         level;
    ngx_socket_t       s;
    ngx_event_t       *rev, *wev;
    ngx_sockaddr_t     sa;
    ngx_listening_t   *ls;
    ngx_connection_t  *c, *lc;
    ngx_event_conf_t  *ecf;
#if (NGX_HAVE_ACCEPT4)
    static ngx_uint_t  use_accept4 = 1;
#endif

    if (ev->timedout) {//表示超时 重新添加listen socket到事件驱动中
        if (ngx_enable_accept_events((ngx_cycle_t *) ngx_cycle) != NGX_OK) {
            return;
        }

        ev->timedout = 0;
    }
    //获取event配置项
    ecf = ngx_event_get_conf(ngx_cycle->conf_ctx, ngx_event_core_module);

    if (!(ngx_event_flags & NGX_USE_KQUEUE_EVENT)) {
        ev->available = ecf->multi_accept;
    }
    //从用户私有数据中获取当前事件所对应的connection以及listening对象
    lc = ev->data;
    ls = lc->listening;
    ev->ready = 0;

    ngx_log_debug2(NGX_LOG_DEBUG_EVENT, ev->log, 0,
                   "accept on %V, ready: %d", &ls->addr_text, ev->available);

这代码主要用于判断当前事件是否为超时事件。超时事件的产生条件是在定时器周期内没有新的请求到来，此时需要重新注册listening socket到epoll中。下面进入主流程do while循环。

2.3、调用accept接收用户请求

    do {
        socklen = sizeof(ngx_sockaddr_t);
        /* 接收客户端连接建立请求 */
#if (NGX_HAVE_ACCEPT4)
        if (use_accept4) {
            s = accept4(lc->fd, &sa.sockaddr, &socklen, SOCK_NONBLOCK);//默认创建的socket是非阻塞socket
        } else {
            s = accept(lc->fd, &sa.sockaddr, &socklen);
        }
#else
        s = accept(lc->fd, &sa.sockaddr, &socklen);
#endif
        /* 连接建立失败 */
        if (s == (ngx_socket_t) -1) {
            err = ngx_socket_errno;

            if (err == NGX_EAGAIN) {
                ngx_log_debug0(NGX_LOG_DEBUG_EVENT, ev->log, err,
                               "accept() not ready");
                return;
            }

            level = NGX_LOG_ALERT;

            if (err == NGX_ECONNABORTED) {
                level = NGX_LOG_ERR;

            } else if (err == NGX_EMFILE || err == NGX_ENFILE) {
                level = NGX_LOG_CRIT;
            }

#if (NGX_HAVE_ACCEPT4)
            ngx_log_error(level, ev->log, err,
                          use_accept4 ? "accept4() failed" : "accept() failed");

            if (use_accept4 && err == NGX_ENOSYS) {
                use_accept4 = 0;
                ngx_inherited_nonblocking = 0;//尝试用accept接口处理新请求
                continue;
            }
#else
            ngx_log_error(level, ev->log, err, "accept() failed");
#endif

            if (err == NGX_ECONNABORTED) {
                if (ngx_event_flags & NGX_USE_KQUEUE_EVENT) {
                    ev->available--;
                }

                if (ev->available) {
                    continue;
                }
            }

            if (err == NGX_EMFILE || err == NGX_ENFILE) {
                if (ngx_disable_accept_events((ngx_cycle_t *) ngx_cycle, 1)
                    != NGX_OK)
                {
                    return;
                }

                if (ngx_use_accept_mutex) {
                    if (ngx_accept_mutex_held) {
                        ngx_shmtx_unlock(&ngx_accept_mutex);
                        ngx_accept_mutex_held = 0;
                    }

                    ngx_accept_disabled = 1;

                } else {
                    ngx_add_timer(ev, ecf->accept_mutex_delay);
                }
            }

            return;
        }

根据不同的错误码，进行特殊处理，大部分场景最后都是return操作。

2.4、创建connection对象

当accept返回成功，则表新的tcp连接已经建立成功，Nginx需要对其进行封装成connection对象并对connection结构成员进行初始化操作，例如socket操作设置等，具体代码如下：

        /* 负数 负载均衡*/       
        ngx_accept_disabled = ngx_cycle->connection_n / 8
                              - ngx_cycle->free_connection_n;

        c = ngx_get_connection(s, ev->log);//获取新的connection对象

        if (c == NULL) {
            if (ngx_close_socket(s) == -1) {
                ngx_log_error(NGX_LOG_ALERT, ev->log, ngx_socket_errno,
                              ngx_close_socket_n " failed");
            }

            return;
        }

        c->type = SOCK_STREAM;

#if (NGX_STAT_STUB)
        (void) ngx_atomic_fetch_add(ngx_stat_active, 1);
#endif
        /* 创建连接级内存池 */
        c->pool = ngx_create_pool(ls->pool_size, ev->log);
        if (c->pool == NULL) {
            ngx_close_accepted_connection(c);
            return;
        }

        if (socklen > (socklen_t) sizeof(ngx_sockaddr_t)) {
            socklen = sizeof(ngx_sockaddr_t);
        }

        c->sockaddr = ngx_palloc(c->pool, socklen);
        if (c->sockaddr == NULL) {
            ngx_close_accepted_connection(c);
            return;
        }

        ngx_memcpy(c->sockaddr, &sa, socklen);

        log = ngx_palloc(c->pool, sizeof(ngx_log_t));
        if (log == NULL) {
            ngx_close_accepted_connection(c);
            return;
        }

        /* set a blocking mode for iocp and non-blocking mode for others */

        if (ngx_inherited_nonblocking) {
            if (ngx_event_flags & NGX_USE_IOCP_EVENT) {
                if (ngx_blocking(s) == -1) {
                    ngx_log_error(NGX_LOG_ALERT, ev->log, ngx_socket_errno,
                                  ngx_blocking_n " failed");
                    ngx_close_accepted_connection(c);
                    return;
                }
            }

        } else {
            if (!(ngx_event_flags & NGX_USE_IOCP_EVENT)) {
                if (ngx_nonblocking(s) == -1) {
                    ngx_log_error(NGX_LOG_ALERT, ev->log, ngx_socket_errno,
                                  ngx_nonblocking_n " failed");
                    ngx_close_accepted_connection(c);
                    return;
                }
            }
        }

        *log = ls->log;
        /* 初始化connection对象 参考ngx_linux_io */
        c->recv = ngx_recv;
        c->send = ngx_send;
        c->recv_chain = ngx_recv_chain;
        c->send_chain = ngx_send_chain;

        c->log = log;
        c->pool->log = log;

        c->socklen = socklen;
        c->listening = ls;
        c->local_sockaddr = ls->sockaddr;
        c->local_socklen = ls->socklen;

#if (NGX_HAVE_UNIX_DOMAIN)
        if (c->sockaddr->sa_family == AF_UNIX) {
            c->tcp_nopush = NGX_TCP_NOPUSH_DISABLED;
            c->tcp_nodelay = NGX_TCP_NODELAY_DISABLED;
#if (NGX_SOLARIS)
            /* Solaris's sendfilev() supports AF_NCA, AF_INET, and AF_INET6 */
            c->sendfile = 0;
#endif
        }
#endif

        rev = c->read;
        wev = c->write;

        wev->ready = 1;

        if (ngx_event_flags & NGX_USE_IOCP_EVENT) {
            rev->ready = 1;
        }

        if (ev->deferred_accept) {
            rev->ready = 1;
#if (NGX_HAVE_KQUEUE || NGX_HAVE_EPOLLRDHUP)
            rev->available = 1;
#endif
        }

        rev->log = log;
        wev->log = log;

        /*
         * TODO: MT: - ngx_atomic_fetch_add()
         *             or protection by critical section or light mutex
         *
         * TODO: MP: - allocated in a shared memory
         *           - ngx_atomic_fetch_add()
         *             or protection by critical section or light mutex
         */
        c->number = ngx_atomic_fetch_add(ngx_connection_counter, 1);

#if (NGX_STAT_STUB)
        (void) ngx_atomic_fetch_add(ngx_stat_handled, 1);
#endif

        if (ls->addr_ntop) {
            c->addr_text.data = ngx_pnalloc(c->pool, ls->addr_text_max_len);
            if (c->addr_text.data == NULL) {
                ngx_close_accepted_connection(c);
                return;
            }

            c->addr_text.len = ngx_sock_ntop(c->sockaddr, c->socklen,
                                             c->addr_text.data,
                                             ls->addr_text_max_len, 0);
            if (c->addr_text.len == 0) {
                ngx_close_accepted_connection(c);
                return;
            }
        }

2.5、HTTP框架初始化connection对象

下面这段代码最主要功能是对connection对象初始化流程，如下：

        /**
         * ngx_add_conn不空且没有设置NGX_USE_EPOLL_EVENT标志位 
         * epoll模型不会进入此分支
         */
        if (ngx_add_conn && (ngx_event_flags & NGX_USE_EPOLL_EVENT) == 0) {
            if (ngx_add_conn(c) == NGX_ERROR) {
                ngx_close_accepted_connection(c);
                return;
            }
        }

        log->data = NULL;
        log->handler = NULL;
        
        /*
         * 此回调函数用于处理新的连接 handler赋值由HTTP框架设置
         * ngx_http_init_connection 这个函数主要功能是将当前socket注册到事件驱动中
         */
        ls->handler(c);

        if (ngx_event_flags & NGX_USE_KQUEUE_EVENT) {
            ev->available--;
        }

    } while (ev->available);

三、注册事件驱动

上面章节，Nginx是如何处理Accept事件，但是有一点疑问是accept函数返回的socket是如何加入到epoll中呢？listening socket是如何再次加入到epoll中呢（每次accept事件之后需要重新加入）?

3.1、新socket加入epoll事件驱动

在上一节中有这一行代码：

        /*
         * 此回调函数用于处理新的连接 handler赋值由HTTP框架设置
         * ngx_http_init_connection 这个函数主要功能是将当前socket注册到事件驱动中
         */
        ls->handler(c);

该代码是HTTP框架用于初始化连接 ，回调函数是ngx_http_init_connection，该函数用于将客户端与服务端通信的socket加入到事件驱动中：

/**
 * 初始化http连接
 * @param c TCP连接
 */
void ngx_http_init_connection(ngx_connection_t *c)
{
    ngx_uint_t i;
    ngx_event_t *rev;
    struct sockaddr_in *sin;
    ngx_http_port_t *port;
    ngx_http_in_addr_t *addr;
    ngx_http_log_ctx_t *ctx;
    ngx_http_connection_t *hc;
#if (NGX_HAVE_INET6)
    struct sockaddr_in6 *sin6;
    ngx_http_in6_addr_t *addr6;
#endif
    /* 在内存池中申请http connection对象 */
    hc = ngx_pcalloc(c->pool, sizeof(ngx_http_connection_t));
    if (hc == NULL)
    {
        ngx_http_close_connection(c);
        return;
    }

    c->data = hc;//save 

    /* find the server configuration for the address:port */

    port = c->listening->servers;
    /* 获取监听地址 如果有多个地址 则循环遍历 */
    if (port->naddrs > 1)
    {

        /*
         * there are several addresses on this port and one of them
         * is an "*:port" wildcard so getsockname() in ngx_http_server_addr()
         * is required to determine a server address
         */

        if (ngx_connection_local_sockaddr(c, NULL, 0) != NGX_OK)
        {
            ngx_http_close_connection(c);
            return;
        }

        switch (c->local_sockaddr->sa_family)
        {

#if (NGX_HAVE_INET6)
        case AF_INET6:
            sin6 = (struct sockaddr_in6 *)c->local_sockaddr;

            addr6 = port->addrs;

            /* the last address is "*" */

            for (i = 0; i < port->naddrs - 1; i++)
            {
                if (ngx_memcmp(&addr6[i].addr6, &sin6->sin6_addr, 16) == 0)
                {
                    break;
                }
            }

            hc->addr_conf = &addr6[i].conf;

            break;
#endif

        default: /* AF_INET */
            sin = (struct sockaddr_in *)c->local_sockaddr;

            addr = port->addrs;

            /* the last address is "*" */

            for (i = 0; i < port->naddrs - 1; i++)
            {
                if (addr[i].addr == sin->sin_addr.s_addr)
                {
                    break;
                }
            }

            hc->addr_conf = &addr[i].conf;

            break;
        }
    }
    else
    {

        switch (c->local_sockaddr->sa_family)
        {

#if (NGX_HAVE_INET6)
        case AF_INET6:
            addr6 = port->addrs;
            hc->addr_conf = &addr6[0].conf;
            break;
#endif

        default: /* AF_INET */
            addr = port->addrs;
            hc->addr_conf = &addr[0].conf;
            break;
        }
    }

    /* the default server configuration for the address:port */
    hc->conf_ctx = hc->addr_conf->default_server->ctx;

    ctx = ngx_palloc(c->pool, sizeof(ngx_http_log_ctx_t));
    if (ctx == NULL)
    {
        ngx_http_close_connection(c);
        return;
    }

    ctx->connection = c;
    ctx->request = NULL;
    ctx->current_request = NULL;

    c->log->connection = c->number;
    c->log->handler = ngx_http_log_error;
    c->log->data = ctx;
    c->log->action = "waiting for request";

    c->log_error = NGX_ERROR_INFO;
    /* 设置读写事件handler回调函数 */
    rev = c->read;
    rev->handler = ngx_http_wait_request_handler; /* 读事件回调函数 */
    c->write->handler = ngx_http_empty_handler; /* 写事件回调函数 再未收到client请求不会主动发送数据 */

#if (NGX_HTTP_V2)
    if (hc->addr_conf->http2)
    {
        rev->handler = ngx_http_v2_init;
    }
#endif

#if (NGX_HTTP_SSL)
    {
        ngx_http_ssl_srv_conf_t *sscf;

        sscf = ngx_http_get_module_srv_conf(hc->conf_ctx, ngx_http_ssl_module);

        if (sscf->enable || hc->addr_conf->ssl)
        {

            c->log->action = "SSL handshaking";

            if (hc->addr_conf->ssl && sscf->ssl.ctx == NULL)
            {
                ngx_log_error(NGX_LOG_ERR, c->log, 0,
                              "no \"ssl_certificate\" is defined "
                              "in server listening on SSL port");
                ngx_http_close_connection(c);
                return;
            }

            hc->ssl = 1;

            rev->handler = ngx_http_ssl_handshake;
        }
    }
#endif

    if (hc->addr_conf->proxy_protocol)
    {
        hc->proxy_protocol = 1;
        c->log->action = "reading PROXY protocol";
    }

    if (rev->ready)
    {
        /* the deferred accept(), iocp */

        if (ngx_use_accept_mutex)
        {
            ngx_post_event(rev, &ngx_posted_events);
            return;
        }

        rev->handler(rev);
        return;
    }
    /*
     * 将事件添加到定时器和epoll事件驱动中
     * 回调函数都是ngx_http_wait_request_handler
     */
    ngx_add_timer(rev, c->listening->post_accept_timeout);
    ngx_reusable_connection(c, 1);
    if (ngx_handle_read_event(rev, 0) != NGX_OK)
    {
        ngx_http_close_connection(c);
        return;
    }
}

通过上述代码可知：

1、读事件处理函数为ngx_http_wait_request_handler，写事件处理函数为ngx_http_empty_handler。

2、将当前socket加入到epoll以及定时器中。

四、总结

至此，HTTP会话流程已经建立完成。HTTP会话主要是初始化connection对象以及将通信socket加入到事件驱动中即可。