• Redis源码分析(二十二)--- networking网络协议传输


              上次我只分析了Redis网络部分的代码一部分,今天我把networking的代码实现部分也学习了一遍,netWorking的代码更多偏重的是Client客户端的操作。里面addReply()系列的方法操作是主要的部分。光光这个系列的方法,应该占据了一半的API的数量。我把API分成了3个部分:

    /* ------------ API ---------------------- */
    void *dupClientReplyValue(void *o)	/* 复制value一份 */
    int listMatchObjects(void *a, void *b) /* 比价2个obj是否相等 */
    robj *dupLastObjectIfNeeded(list *reply) /* 返回回复列表中最后一个元素对象 */
    void copyClientOutputBuffer(redisClient *dst, redisClient *src) /* 将源Client的输出buffer复制给目标Client */
    static void acceptCommonHandler(int fd, int flags) /* 网络连接后的调用方法 */
    void acceptTcpHandler(aeEventLoop *el, int fd, void *privdata, int mask)
    void acceptUnixHandler(aeEventLoop *el, int fd, void *privdata, int mask)
    void disconnectSlaves(void) /* 使server的slave失去连接 */
    void replicationHandleMasterDisconnection(void)
    void flushSlavesOutputBuffers(void) /* 从方法将会在freeMemoryIfNeeded(),释放内存空间函数,将存在内存中数据操作结果刷新到磁盘中 */
    int processEventsWhileBlocked(void)
    
    /* ------------- addReply API -----------------   */
    int _addReplyToBuffer(redisClient *c, char *s, size_t len) /* 往客户端缓冲区中添加内容 */
    void _addReplyObjectToList(redisClient *c, robj *o) /* robj添加到reply的列表中 */
    void _addReplySdsToList(redisClient *c, sds s) /* 在回复列表中添加Sds字符串对象 */
    void _addReplyStringToList(redisClient *c, char *s, size_t len) /* 在回复列表中添加字符串对象,参数中已经给定字符的长度 */
    void addReply(redisClient *c, robj *obj) /* 在redisClient的buffer中写入数据,数据存在obj->ptr的指针中 */
    void addReplySds(redisClient *c, sds s) /* 在回复中添加Sds字符串,下面的额addReply()系列方法原理基本类似 */
    void addReplyString(redisClient *c, char *s, size_t len)
    void addReplyErrorLength(redisClient *c, char *s, size_t len)
    void addReplyError(redisClient *c, char *err) /* 往Reply中添加error类的信息 */
    void addReplyErrorFormat(redisClient *c, const char *fmt, ...)
    void addReplyStatusLength(redisClient *c, char *s, size_t len)
    void addReplyStatus(redisClient *c, char *status)
    void addReplyStatusFormat(redisClient *c, const char *fmt, ...)
    void *addDeferredMultiBulkLength(redisClient *c) /* 在reply list 中添加一个空的obj对象 */
    void setDeferredMultiBulkLength(redisClient *c, void *node, long length)
    void addReplyDouble(redisClient *c, double d) /* 在bulk reply中添加一个double类型值,bulk的意思为大块的,bulk reply的意思为大数据量的回复 */
    void addReplyLongLongWithPrefix(redisClient *c, long long ll, char prefix)
    void addReplyLongLong(redisClient *c, long long ll)
    void addReplyMultiBulkLen(redisClient *c, long length)
    void addReplyBulkLen(redisClient *c, robj *obj) /* 添加bulk 大块的数据的长度 */
    void addReplyBulk(redisClient *c, robj *obj) /* 将一个obj的数据,拆分成大块数据的添加 */
    void addReplyBulkCBuffer(redisClient *c, void *p, size_t len)
    void addReplyBulkCString(redisClient *c, char *s)
    void addReplyBulkLongLong(redisClient *c, long long ll)
    
    /* ------------- Client API -----------------   */	
    redisClient *createClient(int fd) /* 创建redisClient客户端,1.建立连接,2.设置数据库,3.属性设置 */
    int prepareClientToWrite(redisClient *c) /* 此方法将会被调用于Client准备接受新数据之前调用,在fileEvent为客户端设定writer的handler处理事件 */
    static void freeClientArgv(redisClient *c)
    void freeClient(redisClient *c) /* 释放freeClient,要分为Master和Slave2种情况作不同的处理 */
    void freeClientAsync(redisClient *c)
    void freeClientsInAsyncFreeQueue(void) /* 异步的free客户端 */
    void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) /* 将Client中的reply数据存入文件中 */
    void resetClient(redisClient *c)
    int processInlineBuffer(redisClient *c) /* 处理redis Client的内链的buffer,就是c->querybuf */
    static void setProtocolError(redisClient *c, int pos)
    int processMultibulkBuffer(redisClient *c) /* 处理大块的buffer */
    void processInputBuffer(redisClient *c) /* 处理redisClient的查询buffer */
    void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask) /* 从Client获取查询query语句 */
    void getClientsMaxBuffers(unsigned long *longest_output_list,
                              unsigned long *biggest_input_buffer) /* 获取Client中输入buffer和输出buffer的最大长度值 */
    void formatPeerId(char *peerid, size_t peerid_len, char *ip, int port) /* 格式化ip,port端口号的输出,ip:port */
    int genClientPeerId(redisClient *client, char *peerid, size_t peerid_len) /* 获取Client客户端的ip,port地址信息 */
    char *getClientPeerId(redisClient *c) /* 获取c->peerid客户端的地址信息 */
    sds catClientInfoString(sds s, redisClient *client) /* 格式化的输出客户端的属性信息,直接返回一个拼接好的字符串 */
    sds getAllClientsInfoString(void) /* 获取所有Client客户端的属性信息,并连接成一个总的字符串并输出 */
    void clientCommand(redisClient *c) /* 执行客户端的命令的作法 */
    void rewriteClientCommandVector(redisClient *c, int argc, ...) /* 重写客户端的命令集合,旧的命令集合的应用计数减1,新的Command  Vector的命令集合增1 */
    void rewriteClientCommandArgument(redisClient *c, int i, robj *newval) /* 重写Client中的第i个参数 */
    unsigned long getClientOutputBufferMemoryUsage(redisClient *c) /* 获取Client中已经用去的输出buffer的大小 */
    int getClientType(redisClient *c)
    int getClientTypeByName(char *name) /* Client中的名字的3种类型,normal,slave,pubsub */
    char *getClientTypeName(int class)
    int checkClientOutputBufferLimits(redisClient *c) /* 判断Clint的输出缓冲区的已经占用大小是否超过软限制或是硬限制 */
    void asyncCloseClientOnOutputBufferLimitReached(redisClient *c) /* 异步的关闭Client,如果缓冲区中的软限制或是硬限制已经到达的时候,缓冲区超出限制的结果会导致释放不安全, */
    我们从最简单的_addReplyToBuffer在缓冲区中添加回复数据开始说起,因为后面的各种addReply的方法都或多或少的调用了和这个歌方法。

    /* -----------------------------------------------------------------------------
     * Low level functions to add more data to output buffers.
     * -------------------------------------------------------------------------- */
    /* 往客户端缓冲区中添加内容 */
    int _addReplyToBuffer(redisClient *c, char *s, size_t len) {
        size_t available = sizeof(c->buf)-c->bufpos;
    
        if (c->flags & REDIS_CLOSE_AFTER_REPLY) return REDIS_OK;
    
        /* If there already are entries in the reply list, we cannot
         * add anything more to the static buffer. */
         //如果当前的reply已经存在内容,则操作出错
        if (listLength(c->reply) > 0) return REDIS_ERR;
    
        /* Check that the buffer has enough space available for this string. */
        if (len > available) return REDIS_ERR;
    
        memcpy(c->buf+c->bufpos,s,len);
        c->bufpos+=len;
        return REDIS_OK;
    }
    最直接影响的一句话,就是memcpy(c->buf+c->bufpos,s,len);所以内容是加到c->buf中的,这也就是客户端的输出buffer,添加操作还有另外一种形式是添加对象类型:

    /* robj添加到reply的列表中 */
    void _addReplyObjectToList(redisClient *c, robj *o) {
        robj *tail;
    
        if (c->flags & REDIS_CLOSE_AFTER_REPLY) return;
    
        if (listLength(c->reply) == 0) {
            incrRefCount(o);
            //在回复列表汇总添加robj内容
            listAddNodeTail(c->reply,o);
            c->reply_bytes += zmalloc_size_sds(o->ptr);
        } else {
            tail = listNodeValue(listLast(c->reply));
    
            /* Append to this object when possible. */
            if (tail->ptr != NULL &&
                sdslen(tail->ptr)+sdslen(o->ptr) <= REDIS_REPLY_CHUNK_BYTES)
            {
                c->reply_bytes -= zmalloc_size_sds(tail->ptr);
                tail = dupLastObjectIfNeeded(c->reply);
                tail->ptr = sdscatlen(tail->ptr,o->ptr,sdslen(o->ptr));
                c->reply_bytes += zmalloc_size_sds(tail->ptr);
            } else {
                incrRefCount(o);
                listAddNodeTail(c->reply,o);
                c->reply_bytes += zmalloc_size_sds(o->ptr);
            }
        }
        asyncCloseClientOnOutputBufferLimitReached(c);
    }
    把robj对象加载reply列表中,并且改变reply的byte大小,最后还调用了一个asyncCloseClientOnOutputBufferLimitReached(c);方法,这个方法我是在这个文件的最底部找到的,一开始还真不知道什么意思,作用就是当添加完数据后,当客户端的输出缓冲的大小超出限制时,会被异步关闭:

    /* Asynchronously close a client if soft or hard limit is reached on the
     * output buffer size. The caller can check if the client will be closed
     * checking if the client REDIS_CLOSE_ASAP flag is set.
     *
     * Note: we need to close the client asynchronously because this function is
     * called from contexts where the client can't be freed safely, i.e. from the
     * lower level functions pushing data inside the client output buffers. */
    /* 异步的关闭Client,如果缓冲区中的软限制或是硬限制已经到达的时候,缓冲区超出限制的结果会导致释放不安全, */
    void asyncCloseClientOnOutputBufferLimitReached(redisClient *c) {
        redisAssert(c->reply_bytes < ULONG_MAX-(1024*64));
        if (c->reply_bytes == 0 || c->flags & REDIS_CLOSE_ASAP) return;
        if (checkClientOutputBufferLimits(c)) {
            sds client = catClientInfoString(sdsempty(),c);
    
            freeClientAsync(c);
            redisLog(REDIS_WARNING,"Client %s scheduled to be closed ASAP for overcoming of output buffer limits.", client);
            sdsfree(client);
        }
    }
    
    在addReply方法调用的时候,有时是需要一个前提的,我说的是在写数据事件发生的时候,你得先对写的文件创建一个监听事件:

    /* 在回复中添加Sds字符串 */
    void addReplySds(redisClient *c, sds s) {
    	//在调用添加操作之前,都要先执行prepareClientToWrite(c),设置文件事件的写事件
        if (prepareClientToWrite(c) != REDIS_OK) {
            /* The caller expects the sds to be free'd. */
            sdsfree(s);
            return;
        }
        if (_addReplyToBuffer(c,s,sdslen(s)) == REDIS_OK) {
            sdsfree(s);
        } else {
            /* This method free's the sds when it is no longer needed. */
            _addReplySdsToList(c,s);
        }
    }
    在这个prepareClientToWrite()里面是干嘛的呢?

    /* This function is called every time we are going to transmit new data
     * to the client. The behavior is the following:
     *
     * If the client should receive new data (normal clients will) the function
     * returns REDIS_OK, and make sure to install the write handler in our event
     * loop so that when the socket is writable new data gets written.
     *
     * If the client should not receive new data, because it is a fake client,
     * a master, a slave not yet online, or because the setup of the write handler
     * failed, the function returns REDIS_ERR.
     *
     * Typically gets called every time a reply is built, before adding more
     * data to the clients output buffers. If the function returns REDIS_ERR no
     * data should be appended to the output buffers. */
    /* 此方法将会被调用于Client准备接受新数据之前调用,在fileEvent为客户端设定writer的handler处理事件 */
    int prepareClientToWrite(redisClient *c) {
        if (c->flags & REDIS_LUA_CLIENT) return REDIS_OK;
        if ((c->flags & REDIS_MASTER) &&
            !(c->flags & REDIS_MASTER_FORCE_REPLY)) return REDIS_ERR;
        if (c->fd <= 0) return REDIS_ERR; /* Fake client */
        if (c->bufpos == 0 && listLength(c->reply) == 0 &&
            (c->replstate == REDIS_REPL_NONE ||
             c->replstate == REDIS_REPL_ONLINE) &&
            //在这里创建写的文件事件
            aeCreateFileEvent(server.el, c->fd, AE_WRITABLE,
            sendReplyToClient, c) == AE_ERR) return REDIS_ERR;
        return REDIS_OK;
    }
    
    在addReply的方法里提到了一个addReplyBulk类型方法,Bulk的中文意思为大块的,说明addReplyBulk添加的都是一些比较大块的数据,找一个方法看看:

    /* Add a Redis Object as a bulk reply */
    /* 将一个obj的数据,拆分成大块数据的添加 */
    void addReplyBulk(redisClient *c, robj *obj) {
    	//reply添加长度
        addReplyBulkLen(c,obj);
        //reply添加对象
        addReply(c,obj);
        addReply(c,shared.crlf);
    }
    将原本一个robj的数据拆分成可3个普通的addReply的方法调用。就变成了数据量变大了的数据。大数据的回复一个比较不好的地方是到时解析的时候或者是Data的复制的时候会比较耗时。在networking的方法里还提供了freeClient()的操作:

    /* 释放freeClient,要分为Master和Slave2种情况作不同的处理 */
    void freeClient(redisClient *c) {
        listNode *ln;
    
        /* If this is marked as current client unset it */
        if (server.current_client == c) server.current_client = NULL;
    
        /* If it is our master that's beging disconnected we should make sure
         * to cache the state to try a partial resynchronization later.
         *
         * Note that before doing this we make sure that the client is not in
         * some unexpected state, by checking its flags. */
        if (server.master && c->flags & REDIS_MASTER) {
            redisLog(REDIS_WARNING,"Connection with master lost.");
            if (!(c->flags & (REDIS_CLOSE_AFTER_REPLY|
                              REDIS_CLOSE_ASAP|
                              REDIS_BLOCKED|
                              REDIS_UNBLOCKED)))
            {
            	//如果是Master客户端,需要做缓存Client的处理,可以迅速重新启用
                replicationCacheMaster(c);
                return;
            }
        }
    
    ...后面代码略去了

       当Client中的输出buffer数据渐渐变多了的时候就要准备持久化到磁盘文件了,要调用下面这个方法了,

    /* Helper function used by freeMemoryIfNeeded() in order to flush slave
     * output buffers without returning control to the event loop. */
    /* 从方法将会在freeMemoryIfNeeded(),释放内存空间函数,将存在内存中数据操作结果刷新到磁盘中 */
    void flushSlavesOutputBuffers(void) {
        listIter li;
        listNode *ln;
    
        listRewind(server.slaves,&li);
        while((ln = listNext(&li))) {
            redisClient *slave = listNodeValue(ln);
            int events;
    
            events = aeGetFileEvents(server.el,slave->fd);
            if (events & AE_WRITABLE &&
                slave->replstate == REDIS_REPL_ONLINE &&
                listLength(slave->reply))
            {
            	//在这里调用了write的方法
                sendReplyToClient(server.el,slave->fd,slave,0);
            }
        }
    }
    
    这个方法的核心调用又在sendReplyToClient()方法,就是把Client的reply内容和buf内容存入文件。以上就是我的理解了,代码量有点大,的确看的我头有点大。
  • 相关阅读:
    不指定虚拟路径的前提下通过http访问pdf、图片等文件
    Java Service Wrapper将java程序设置为服务
    C# 上传excel文档解析出里面数据
    如何同时启动多个Tomcat服务
    struts2的action的知识点和利用action向页面注入值的操作
    IOS6 字体高亮显示
    微软安全新闻聚焦-双周刊第三十五期
    工厂三兄弟之工厂方法模式(四)
    清空文件夹里面的所有文件和文件夹
    OOP设计模式[JAVA]——03职责链模式
  • 原文地址:https://www.cnblogs.com/bianqi/p/12184224.html
Copyright © 2020-2023  润新知