基础

假定是udp上接收包，端口 5060 ，如果是tcp是要加一些额外得代码，
可以使用jrtplib，或者自己写udpserver 放出udp socket 接收包，当然使用jrtplib等稍微简单一些。不过也一样失去了一些灵活性。

步骤

1 接收udp over rtp包
2 接收后分析ps包头，获取es流
3 如果不是ps则接收到包后根据rtp协议再解析rtp， 获取分片等信息组合成包
4 分析es流

code

#define FILE_STORAGE 0Service_RTP_Analyse::Service_RTP_Analyse()
{
}Service_RTP_Analyse::~Service_RTP_Analyse()
{
}
int Service_RTP_Analyse::analyse_rtp(uint8_t * rtp, size_t len)
{return 0;
}//注意这里只是适应h264
int Service_RTP_Analyse::analyse_h264es(uint8_t * pos, int len, uint8_t ** nalu, size_t &nalulen)
{uint8_t* p = pos + 4 + 2 + 2; //跳过4字节00 00 01 E0 和两个字节长度 以及两个字节的跳空int skip = *p;p = p + skip + 1; //p 指向　00 00 00 01 67 等字节上*nalu = p;while (!*(p++));uint8_t c1 = (uint8_t)(*p);unsigned char nal_type = c1 & 0x1f;nalulen = len - (int)(*nalu - pos); //因为　00 00 01 E0 以及两个代表长度的字节没有计算return nal_type;
}int Service_RTP_Analyse::analyseps0(uint8_t * ps, size_t len, int &type, int &length, uint8_t **next, size_t &nextlen)
{#define D(y,x) (*(y+x))if (D(ps, 0) == 0x00 && D(ps, 1) == 0x00 && D(ps, 2) == 0x01){uint8_t num = D(ps, 3);switch (num){case 0xBA:{int skip1 = D(ps, 13) & 0x07;uint8_t * nps = ps + 14 + skip1;length = 14;type = PS_PS;*next = nps;nextlen = len - (nps - ps);}return 0;case 0xBB://系统标题起始码字段case 0xBC://映射流标识字段 {char temp[2];temp[0] = D(ps, 5);temp[1] = D(ps, 4);length = *(unsigned short*)(&temp[0]) + 6; //00 00 01 xx 以及两个字节的长度type = PS_SYS;*next = ps + length;nextlen = len - (int)(*next - ps);}return 0;case 0xE0:{char temp[2];temp[0] = D(ps, 5);temp[1] = D(ps, 4);length = *(unsigned short*)(&temp[0]) + 6;//00 00 01 xx 以及两个字节的长度type = PS_VIDEO;*next = ps + length;nextlen = len -(int)(*next - ps);}return 0;case 0xBD:case 0xC0:{char temp[2];temp[0] = D(ps, 5);temp[1] = D(ps, 4);length = *(unsigned short*)(&temp[0]) + 6;//00 00 01 xx 以及两个字节的长度if (num == 0xBD)type = PS_PRIVATE;elsetype = PS_AUDIO;*next = ps + length;nextlen = len - (int)(*next - ps);}return 0;default:{length = 0;type = PS_UNKONWN;}return -1;}}else{length = 0;type = PS_UNKONWN;return -1;}
}/*
//该函数解一帧完整的ps，输入时一帧完整的ps帧
//rc 上下文环境
//pos ps流数据指针
//len 长度
*/
int Service_RTP_Analyse::analyseps(rtp_context *prc, uint8_t * pos, size_t len)
{int type = 0;int length = 0; //当前段长度uint8_t * ps = pos;size_t pslen = len;uint8_t * nextpos = NULL;size_t nextlen = 0;uint8_t * sps = NULL;int spslen = 0;uint8_t * pps = NULL;int ppslen = 0;uint8_t * se = NULL;int selen = 0;uint8_t *nalu = NULL;size_t nalulen = 0;while (1){int ret = analyseps0(ps, pslen, type, length, &nextpos, nextlen);if (ret == -1)return -1;switch (type) //处理当前段{case PS_PS:ps = nextpos;pslen = nextlen;break;case PS_SYS:ps = nextpos;pslen = nextlen;break;case PS_VIDEO:{int type = analyse_h264es(ps, length, &nalu, nalulen);switch (type){case 0x07:sps = nalu;spslen = (int)nalulen;break;case 0x08:pps = nalu;ppslen = (int)nalulen;break;case 0x06: //skipse = nalu;selen = (int)nalulen;break;case 0x05:case 0x01://skip,because except sps pps ,only one 0x05 key frame nalu break;default:break;}ps = nextpos;pslen = nextlen;}break;case PS_PRIVATE:ps = nextpos;pslen = nextlen;break;case PS_AUDIO:ps = nextpos;pslen = nextlen;break;}if (nextlen <= 0)break;}//顺序找到了流以后，开始移动//有关键帧，则移动sps和pps
#if FILE_STORAGEif (ar->fp == NULL)ar->fp = fopen("h264save.264", "wb");if (ar->fpps == NULL)ar->fpps = fopen("h264saveps.264", "wb");fwrite(pos, 1, len, ar->fpps);
#endifif (sps != NULL && pps != NULL){//获取完整的一帧//移动少量数据，达到意义上的零拷贝memmove(nalu - ppslen, pps, ppslen);memmove(nalu - ppslen - spslen, sps, spslen);sps = nalu - ppslen - spslen;pps = nalu - ppslen;if (prc->_callback != NULL){prc->_callback(prc->ip,prc->port,sps,ppslen + spslen + nalulen,KEY_FRAME,sps,spslen + ppslen,prc->pts++);}
#if FILE_STORAGEfwrite(sps, 1, spslen + ppslen + nalulen, ar->fp);
#endif}else //非关键帧{if (prc->_callback != NULL)prc->_callback(prc->ip, prc->port, nalu, nalulen, NOT_KEY_FRAME,NULL,0, prc->pts++);#if FILE_STORAGEfwrite(nalu, 1, nalulen, ar->fp);
#endif}return 0;
}
//ar 因为零拷贝的事情buffer可能没有值int Service_RTP_Analyse::insert(uint32_t ip, uint16_t port, void *pkt0)
{RTPPacket * pkt = (RTPPacket*)pkt0;//if (pkt->GetPayloadLength() > BUF_LENGTH) //这里还是要修改 钱波//	return -1;auto iter = _hash.find(ip);if (iter == _hash.end()) //没有找到该IP地址的存储{OutputDebugString(L"not found");uint8_t * ps = pkt->GetPayloadData();//分析ps流if (*ps == 0x00 && *(ps + 1) == 0x00 && *(ps + 2) == 0x01 && *(ps + 3) == 0xBA){analyse_struct as;as.in = pkt->GetPayloadData();as.inlen = pkt->GetPayloadLength();AnalyseH264Frame(&as);if (as.ret == 0) //非关键帧{OutputDebugString(L"第一RTP包非关键帧头部,丢弃\n");return -1;}rtp_context rc;rc.ip = ip;rc.port = port;rc._callback = _callback;if (pkt->HasMarker()){rc.blen = 0;analyseps(&rc, pkt->GetPayloadData(), pkt->GetPayloadLength());//send_func(&rc, ip, port, pkt->GetPayloadData(), pkt->GetPayloadLength(), 0);}else{ALLOC_MEM(rc);rc.blen= pkt->GetPayloadLength();memcpy(rc.buffer, pkt->GetPayloadData(), rc.blen);}//fixme:qianbo//_hash.emplace(pair<uint32_t, rtp_context>(ip, as0));_hash.insert(pair<uint32_t, rtp_context>(ip, rc));}else //分析普通rtp流{}}else //找到存储{//代码优化零拷贝rtp_context &rc = iter->second;size_t &len = rc.blen;if (pkt->HasMarker()) // 回调启用{if (len == 0) {//执行零拷贝操作analyseps(&rc, pkt->GetPayloadData(), pkt->GetPayloadLength());//send_func(&rc, ip, port, pkt->GetPayloadData(), pkt->GetPayloadLength(), 0);}else // len!=0{uint8_t * buf = iter->second.buffer;memcpy(buf + len, pkt->GetPayloadData(), pkt->GetPayloadLength());len += pkt->GetPayloadLength();analyseps(&rc, buf, len);//send_func(&rc, ip, port, buf, len, 0);}len = 0;}else{ALLOC_MEM(rc);uint8_t * buf = iter->second.buffer;memcpy(buf + len, pkt->GetPayloadData(), pkt->GetPayloadLength());len += pkt->GetPayloadLength();}}return 0;}// 功能：解码RTP H.264视频// 参数：1.RTP包缓冲地址 2.RTP包数据大小 3.H264输出地址 4.输出数据大小// 返回：true:表示一帧结束  false:FU-A分片未结束或帧未结束bool  Service_RTP_Analyse::decode_rtp_h264(void *bufIn, int len, void **pBufOut, int *pOutLen){*pOutLen = 0;if (len  <  RTP_HEADLEN){return   false;}uint8_t *src = (uint8_t*)bufIn + RTP_HEADLEN;//第1字节uint8_t  head_1 = *src;//第2字节uint8_t  head_2 = *(src + 1);//FU indicator的类型域，uint8_t  nal = head_1 & 0x1f;//FU header的前三位，当前是分包的1开始 2中间 3结束uint8_t  flag = head_2 & 0xe0;uint8_t  nal_fua = (head_1 & 0xe0) | (head_2 & 0x1f); // FU_A nalbool  bFinishFrame = false;if (nal == 0x1c) // 判断NAL的类型为0x1c=28，说明是FU-A分片{ // fu-aif (flag == 0x80) // 开始{*pBufOut = src - 3;*((int *)(*pBufOut)) = 0x01000000; // *((char *)(*pBufOut) + 4) = nal_fua;*pOutLen = len - RTP_HEADLEN + 3;}else if (flag == 0x40) // 结束{*pBufOut = src + 2;*pOutLen = len - RTP_HEADLEN - 2;}else // 中间{*pBufOut = src + 2;*pOutLen = len - RTP_HEADLEN - 2;}}else // 单包数据{*pBufOut = src - 4;*((int *)(*pBufOut)) = 0x01000000; // 大模式会有问题*pOutLen = len - RTP_HEADLEN + 4;}uint8_t *  tmp = (unsigned  char*)bufIn;bFinishFrame = (tmp[1] & 0x80) ? true : false;return  bFinishFrame;}//功能：解RTP AAC音频包，声道和采样频率必须知道。//参数：1.RTP包缓冲地址 2.RTP包数据大小 3.H264输出地址 4.输出数据大小//返回：true:表示一帧结束  false:帧未结束 一般AAC音频包比较小，没有分片。bool Service_RTP_Analyse::decode_rtp_aac(void * bufIn, int recvLen, void** pBufOut, int* pOutLen){unsigned char*  bufRecv = (unsigned char*)bufIn;//char strFileName[20];unsigned char ADTS[] = { 0xFF, 0xF1, 0x00, 0x00, 0x00, 0x00, 0xFC };int audioSamprate = 32000;//音频采样率int audioChannel = 2;//音频声道 1或2int audioBit = 16;//16位 固定switch (audioSamprate){case  16000:ADTS[2] = 0x60;break;case  32000:ADTS[2] = 0x54;break;case  44100:ADTS[2] = 0x50;break;case  48000:ADTS[2] = 0x4C;break;case  96000:ADTS[2] = 0x40;break;default:break;}ADTS[3] = (audioChannel == 2) ? 0x80 : 0x40;int len = recvLen - 16 + 7;len <<= 5;//8bit * 2 - 11 = 5(headerSize 11bit)len |= 0x1F;//5 bit    1            ADTS[4] = len >> 8;ADTS[5] = len & 0xFF;*pBufOut = (char*)bufIn + 16 - 7;memcpy(*pBufOut, ADTS, sizeof(ADTS));*pOutLen = recvLen - 16 + 7;unsigned char* bufTmp = (unsigned char*)bufIn;bool bFinishFrame = false;if (bufTmp[1] & 0x80){//DebugTrace::D("Marker");bFinishFrame = true;}else{bFinishFrame = false;}return true;}int Service_RTP_Analyse::insert_rtp(uint32_t ip, uint16_t port, void * pkt_)
{RTPPacket * pkt = (RTPPacket*)pkt_;uint8_t *buf = pkt->GetPayloadData();size_t len   = pkt->GetPayloadLength();uint8_t * out_head = NULL;int       out_len = 0;uint64_t key = ((uint64_t)ip << 32) | port;//std::string key = itoa()uint32_t rssrc  = pkt->GetSSRC();uint32_t rstamp = pkt->GetTimestamp();auto iter = _hash.find(key);if (iter == _hash.end()){rtp_context s;ALLOC_MEM(s);_hash.insert(pair<uint64_t, rtp_context>(key,s));s.start_timestamp = rstamp;decode_rtp_h264(buf, len,(void**)&out_head, &out_len);if (s.blen - s.pos > out_len){memcpy(s.buffer + s.pos, out_head, out_len);s.pos += out_len;}}else{decode_rtp_h264(buf, len, (void**)&out_head, &out_len);rtp_context &s = iter->second;if (s.blen - s.pos > out_len){memcpy(s.buffer + s.pos, out_head, out_len);s.pos += out_len;}}return 0;
}