RTMPdump（libRTMP）源代码分析 6 建立一个流媒体连接（NetStream部分 1）

//创建流intRTMP_ConnectStream(RTMP *r, int seekTime){  RTMPPacket packet = { 0 };  /* seekTime was already set by SetupStream / SetupURL.   * This is only needed by ReconnectStream.   */  if (seekTime > 0)    r->Link.seekTime = seekTime;  r->m_mediaChannel = 0;  while (!r->m_bPlaying && RTMP_IsConnected(r) && RTMP_ReadPacket(r, &packet))    {      if (RTMPPacket_IsReady(&packet)) {   if (!packet.m_nBodySize)     continue;   if ((packet.m_packetType == RTMP_PACKET_TYPE_AUDIO) ||       (packet.m_packetType == RTMP_PACKET_TYPE_VIDEO) ||       (packet.m_packetType == RTMP_PACKET_TYPE_INFO))     {       RTMP_Log(RTMP_LOGWARNING, "Received FLV packet before play()! Ignoring.");       RTMPPacket_Free(&packet);       continue;     }   //处理Packet！   //----------------   r->dlg->AppendCInfo("建立网络流：处理收到的数据。开始处理收到的数据");   //-----------------------------   RTMP_ClientPacket(r, &packet);   //----------------   r->dlg->AppendCInfo("建立网络流：处理收到的数据。处理完毕，清除数据。");   //-----------------------------   RTMPPacket_Free(&packet); }    }  return r->m_bPlaying;}

乍一看，这个函数的代码量好像挺少的，实际上不然，其复杂度还是挺高的。我觉得比RTMP_Connect()要复杂不少。

其关键就在于这个While()循环。首先，循环的三个条件都满足，就能进行循环。只有出错或者建立网络流（NetStream）的步骤完成后，才能跳出循环。

在这个函数中有两个函数尤为重要：

RTMP_ReadPacket()

RTMP_ClientPacket()

第一个函数的作用是读取通过Socket接收下来的消息（Message）包，但是不做任何处理。第二个函数则是处理消息（Message），并做出响应。这两个函数结合，就可以完成接收消息然后响应消息的步骤。

下面来开一下RTMP_ReadPacket()：

//读取收下来的ChunkintRTMP_ReadPacket(RTMP *r, RTMPPacket *packet){ //packet 存读取完后的的数据 //Chunk Header最大值18  uint8_t hbuf[RTMP_MAX_HEADER_SIZE] = { 0 }; //header 指向的是从Socket中收下来的数据  char *header = (char *)hbuf;  int nSize, hSize, nToRead, nChunk;  int didAlloc = FALSE;  RTMP_Log(RTMP_LOGDEBUG2, "%s: fd=%d", __FUNCTION__, r->m_sb.sb_socket);  //收下来的数据存入hbuf  if (ReadN(r, (char *)hbuf, 1) == 0)    {      RTMP_Log(RTMP_LOGERROR, "%s, failed to read RTMP packet header", __FUNCTION__);      return FALSE;    }  //块类型fmt  packet->m_headerType = (hbuf[0] & 0xc0) >> 6;  //块流ID（2-63）  packet->m_nChannel = (hbuf[0] & 0x3f);  header++;  //块流ID第1字节为0时，块流ID占2个字节  if (packet->m_nChannel == 0)    {      if (ReadN(r, (char *)&hbuf[1], 1) != 1) {   RTMP_Log(RTMP_LOGERROR, "%s, failed to read RTMP packet header 2nd byte",       __FUNCTION__);   return FALSE; }   //计算块流ID（64-319）      packet->m_nChannel = hbuf[1];      packet->m_nChannel += 64;      header++;    }  //块流ID第1字节为0时，块流ID占3个字节  else if (packet->m_nChannel == 1)    {      int tmp;      if (ReadN(r, (char *)&hbuf[1], 2) != 2) {   RTMP_Log(RTMP_LOGERROR, "%s, failed to read RTMP packet header 3nd byte",       __FUNCTION__);   return FALSE; }      tmp = (hbuf[2] << 8) + hbuf[1];   //计算块流ID（64-65599）      packet->m_nChannel = tmp + 64;      RTMP_Log(RTMP_LOGDEBUG, "%s, m_nChannel: %0x", __FUNCTION__, packet->m_nChannel);      header += 2;    }  //ChunkHeader的大小（4种）  nSize = packetSize[packet->m_headerType];  if (nSize == RTMP_LARGE_HEADER_SIZE) /* if we get a full header the timestamp is absolute */    packet->m_hasAbsTimestamp = TRUE; //11字节的完整ChunkMsgHeader的TimeStamp是绝对值  else if (nSize < RTMP_LARGE_HEADER_SIZE)    {    /* using values from the last message of this channel */      if (r->m_vecChannelsIn[packet->m_nChannel]) memcpy(packet, r->m_vecChannelsIn[packet->m_nChannel],        sizeof(RTMPPacket));    }  nSize--;  if (nSize > 0 && ReadN(r, header, nSize) != nSize)    {      RTMP_Log(RTMP_LOGERROR, "%s, failed to read RTMP packet header. type: %x",   __FUNCTION__, (unsigned int)hbuf[0]);      return FALSE;    }  hSize = nSize + (header - (char *)hbuf);  if (nSize >= 3)    { //TimeStamp(注意 BigEndian to SmallEndian)(11，7，3字节首部都有)      packet->m_nTimeStamp = AMF_DecodeInt24(header);      /*RTMP_Log(RTMP_LOGDEBUG, "%s, reading RTMP packet chunk on channel %x, headersz %i, timestamp %i, abs timestamp %i", __FUNCTION__, packet.m_nChannel, nSize, packet.m_nTimeStamp, packet.m_hasAbsTimestamp); */ //消息长度(11，7字节首部都有)      if (nSize >= 6) {   packet->m_nBodySize = AMF_DecodeInt24(header + 3);   packet->m_nBytesRead = 0;   RTMPPacket_Free(packet); //(11，7字节首部都有)   if (nSize > 6)     {    //Msg type ID       packet->m_packetType = header[6];    //Msg Stream ID       if (nSize == 11)  packet->m_nInfoField2 = DecodeInt32LE(header + 7);     } }   //Extend TimeStamp      if (packet->m_nTimeStamp == 0xffffff) {   if (ReadN(r, header + nSize, 4) != 4)     {       RTMP_Log(RTMP_LOGERROR, "%s, failed to read extended timestamp",    __FUNCTION__);       return FALSE;     }   packet->m_nTimeStamp = AMF_DecodeInt32(header + nSize);   hSize += 4; }    }  RTMP_LogHexString(RTMP_LOGDEBUG2, (uint8_t *)hbuf, hSize);  if (packet->m_nBodySize > 0 && packet->m_body == NULL)    {      if (!RTMPPacket_Alloc(packet, packet->m_nBodySize)) {   RTMP_Log(RTMP_LOGDEBUG, "%s, failed to allocate packet", __FUNCTION__);   return FALSE; }      didAlloc = TRUE;      packet->m_headerType = (hbuf[0] & 0xc0) >> 6;    }  nToRead = packet->m_nBodySize - packet->m_nBytesRead;  nChunk = r->m_inChunkSize;  if (nToRead < nChunk)    nChunk = nToRead;  /* Does the caller want the raw chunk? */  if (packet->m_chunk)    {      packet->m_chunk->c_headerSize = hSize;      memcpy(packet->m_chunk->c_header, hbuf, hSize);      packet->m_chunk->c_chunk = packet->m_body + packet->m_nBytesRead;      packet->m_chunk->c_chunkSize = nChunk;    }  if (ReadN(r, packet->m_body + packet->m_nBytesRead, nChunk) != nChunk)    {      RTMP_Log(RTMP_LOGERROR, "%s, failed to read RTMP packet body. len: %lu",   __FUNCTION__, packet->m_nBodySize);      return FALSE;    }  RTMP_LogHexString(RTMP_LOGDEBUG2, (uint8_t *)packet->m_body + packet->m_nBytesRead, nChunk);  packet->m_nBytesRead += nChunk;  /* keep the packet as ref for other packets on this channel */  if (!r->m_vecChannelsIn[packet->m_nChannel])    r->m_vecChannelsIn[packet->m_nChannel] = (RTMPPacket *) malloc(sizeof(RTMPPacket));  memcpy(r->m_vecChannelsIn[packet->m_nChannel], packet, sizeof(RTMPPacket));  //读取完毕  if (RTMPPacket_IsReady(packet))    {      /* make packet's timestamp absolute */      if (!packet->m_hasAbsTimestamp) packet->m_nTimeStamp += r->m_channelTimestamp[packet->m_nChannel]; /* timestamps seem to be always relative!! */      r->m_channelTimestamp[packet->m_nChannel] = packet->m_nTimeStamp;      /* reset the data from the stored packet. we keep the header since we may use it later if a new packet for this channel */      /* arrives and requests to re-use some info (small packet header) */      r->m_vecChannelsIn[packet->m_nChannel]->m_body = NULL;      r->m_vecChannelsIn[packet->m_nChannel]->m_nBytesRead = 0;      r->m_vecChannelsIn[packet->m_nChannel]->m_hasAbsTimestamp = FALSE; /* can only be false if we reuse header */    }  else    {      packet->m_body = NULL; /* so it won't be erased on free */    }  return TRUE;}

在这里要注意的是，接收下来的实际上是块（Chunk）而不是消息（Message），因为消息（Message）在网络上传播的时候，实际上要分割成块（Chunk）。

这里解析的就是块（Chunk）

可参考：RTMP规范简单分析

具体的解析代码我就不多说了，直接参考RTMP协议规范就可以了，一个字节一个字节的解析就OK了。

rtmpdump源代码（Linux）：http://download.csdn.net/detail/leixiaohua1020/6376561

rtmpdump源代码（VC 2005 工程）：http://download.csdn.net/detail/leixiaohua1020/6563163

再分享一下我老师大神的人工智能教程吧。零基础！通俗易懂！风趣幽默！还带黄段子！希望你也加入到我们人工智能的队伍中来！https://blog.csdn.net/jiangjunshow

RTMPdump（libRTMP） 源代码分析 6 建立一个流媒体连接 （NetStream部分 1）

函数调用结构图

详细分析

猜你喜欢

RTMPdump（libRTMP）源代码分析 6 建立一个流媒体连接（NetStream部分 1）