最近要做实时H264 RTP传输,那么如何充分利用live555来做呢?
大家可以看到现有的H264VideoFileServerMediaSubsession中,sink使用了H264VideoRTPSink,source使用了H264VideoStreamFramer,然而这个连接是很复杂的,在这两个节点间要插入了很多其它的节点,其实际情况是这样的:ByteStreamFileSource-->H264VideoStreamParser-->H264VideoStreamFramer-->H264FUAFragmenter-->H264VideoRTPSink.哇!真的这么复杂吗?一点没错!
当然你可以不用理它们的来龙去脉,你只需自己实现一个source,能采集图像并进行h264编码的source(当然你可以用CPU也可以用DSP进行编码),然后用它替代ByteStreamFileSource,就成了,比如你这个source可以叫做H264ByteStreamSource.当然为了提高效率,采集和编码部分应放在另一个线程中执行.
然而,我还是很想了解H264VideoStreamParser到底是什么,Parser到底有什么用?它做了什么?它与H264VideoStreamFramer是如何配合的?它们之间有内存copy发生吗?先设想一个问题:H264VideoStreamFramer是什么角色?跟据H264VideoFileServerMediaSubsession的代码,H264VideoStreamFramer是真正代表source的,Sink所面对的Source就是它.但是它又连接了一个ByteStreamFileSource.look一下这部分代码:
1: FramedSource* H264VideoFileServerMediaSubsession::2: createNewStreamSource(unsigned /*clientSessionId*/,3: unsigned& estBitrate)
4: {5: estBitrate = 500; // kbps, estimate
6:7: // Create the video source:
8: ByteStreamFileSource* fileSource = ByteStreamFileSource::createNew(envir(),9: fFileName);10: if (fileSource == NULL)
11: return NULL;
12: fFileSize = fileSource->fileSize();13:14: // Create a framer for the Video Elementary Stream:
15: return H264VideoStreamFramer::createNew(envir(), fileSource);
16: }
ByteStreamFileSource是从文件取得数据的,它不管是到底什么媒体格式,它只是读文件.所以很明显H264VideoStreamFramer利用ByteStreamFileSource从文件取得数据,然后H264VideoStreamFramer再对数据进行分析.比如找出每个NALU,然后传给Sink.但是H264VideoStreamFramer没有自己去分析,而是利用了Parser,所以那一串中就多了一个H264VideoStreamParser.
H264VideoStreamParser拥有两个source指针,一个是FramedSource* fInputSource,另一个是H264VideoStreamFramer* fUsingSource.可以看出,H264VideoStreamParser把fInputSource和fUsingSource串了起来,那么fInputSource就是ByteStreamFileSource.
我们想像一下H264VideoStreamParser的所作所为:H264VideoStreamFramer把自己的缓冲(其实是sink的)传给H264VideoStreamParser,每当H264VideoStreamFramer要获取一个NALU时,就跟H264VideoStreamParser要,H264VideoStreamParser就从ByteStreamFileSource读一坨数据,然后进行分析,如果取得了一个NALU,就传给H264VideoStreamFramer.唉,H264VideoStreamFramer真是个不劳而获的坏家伙!
看一下实际的流程:
1: //Sink调用Source(H264VideoStreamFramer)的GetNextFrame()获取数据,
2: //H264VideoStreamFramer从MPEGVideoStreamFramer派生,所以下面的函数会被调用:
3: void MPEGVideoStreamFramer::doGetNextFrame()
4: {5: fParser->registerReadInterest(fTo, fMaxSize);6: continueReadProcessing();7: }8:9: void MPEGVideoStreamFramer::continueReadProcessing(void* clientData,10: unsigned char* /*ptr*/,11: unsigned /*size*/,12: struct timeval /*presentationTime*/)13: {14: MPEGVideoStreamFramer* framer = (MPEGVideoStreamFramer*) clientData;15: framer->continueReadProcessing();16: }
上两个是过渡,最终在这里执行:
1: void MPEGVideoStreamFramer::continueReadProcessing()
2: {3: //调用Parser的parser()分析出一个NALU.如果得到了一个NALU,则
4: //用afterGetting(this)返回给Sink.
5: unsigned acquiredFrameSize = fParser->parse();
6: if (acquiredFrameSize > 0)
7: {8: // We were able to acquire a frame from the input.
9: // It has already been copied to the reader's space.
10: fFrameSize = acquiredFrameSize;11: fNumTruncatedBytes = fParser->numTruncatedBytes();12:13: // "fPresentationTime" should have already been computed.
14:15: // Compute "fDurationInMicroseconds" now:
16: fDurationInMicroseconds =17: (fFrameRate == 0.0 || ((int) fPictureCount) < 0) ?
18: 0 : (unsigned) ((fPictureCount * 1000000) / fFrameRate);
19: fPictureCount = 0;20:21: // Call our own 'after getting' function. Because we're not a 'leaf'
22: // source, we can call this directly, without risking infinite recursion.
23: afterGetting(this);
24: }25: else
26: {27: //执行到此处并不代表parser()中没有取得数据!!
28: // We were unable to parse a complete frame from the input, because:
29: // - we had to read more data from the source stream, or
30: // - the source stream has ended.
31: }32: }
上面这个函数的else{}中的注释大家注意了没有?这里关连到一个很难搞懂的现象,后面会解释之.这里先看一下parser()函数是怎样取得数据并进行分析的.parser()中读新数据是由那些test4Bytes(),skipBytes()之类的函数引起的,它们都最终调用了ensureValidBytes1():
1: void StreamParser::ensureValidBytes1(unsigned numBytesNeeded)2: {3: // We need to read some more bytes from the input source.
4: // First, clarify how much data to ask for:
5: unsigned maxInputFrameSize = fInputSource->maxFrameSize();
6: if (maxInputFrameSize > numBytesNeeded)
7: numBytesNeeded = maxInputFrameSize;8:9: // First, check whether these new bytes would overflow the current
10: // bank. If so, start using a new bank now.
11: if (fCurParserIndex + numBytesNeeded > BANK_SIZE)
12: {13: // Swap banks, but save any still-needed bytes from the old bank:
14: unsigned numBytesToSave = fTotNumValidBytes - fSavedParserIndex;
15: unsigned char const* from = &curBank()[fSavedParserIndex];16:17: fCurBankNum = (fCurBankNum + 1) % 2;18: fCurBank = fBank[fCurBankNum];19: memmove(curBank(), from, numBytesToSave);20: fCurParserIndex = fCurParserIndex - fSavedParserIndex;21: fSavedParserIndex = 0;22: fTotNumValidBytes = numBytesToSave;23: }24:25: // ASSERT: fCurParserIndex + numBytesNeeded > fTotNumValidBytes
26: // && fCurParserIndex + numBytesNeeded <= BANK_SIZE
27: if (fCurParserIndex + numBytesNeeded > BANK_SIZE)
28: {29: // If this happens, it means that we have too much saved parser state.
30: // To fix this, increase BANK_SIZE as appropriate.
31: fInputSource->envir() << "StreamParser internal error ("
32: << fCurParserIndex << "+ " << numBytesNeeded << " > "33: << BANK_SIZE << ")\n";
34: fInputSource->envir().internalError();35: }36:37: // Try to read as many new bytes as will fit in the current bank:
38: unsigned maxNumBytesToRead = BANK_SIZE - fTotNumValidBytes;
39: fInputSource->getNextFrame(&curBank()[fTotNumValidBytes], maxNumBytesToRead,40: afterGettingBytes, this, onInputClosure, this);41:42: throw NO_MORE_BUFFERED_INPUT;
43: }
可以看到一个奇怪的现象:这个函数没有返回值,但最终抛出了一个异常,而且只要执行这个函数,就会抛出这个异常.还是先分析一下这个函数做了什么吧:首先判断自己的缓冲区是否能容纳所需的数据量,如果实在不能,也只能提示一下,最后从ByteStreamFileSource获取一坨数据.curBack()返回的就是Parser自己的缓冲.而afterGettingBytes这个回调函数是H264VideoStreamFramer传入的,所以获取数据之后会执行H264VideoStreamFramer的函数,中转几下后,最终执行的就是上面的void MPEGVideoStreamFramer::continueReadProcessing().哇,看到了一个问题:Parser()中嵌套执行Parser()!而第二次执行Parser()完成后,返回到ensureValidBytes1(),然后由于抛出异常而退出,退出到哪里了呢?退回到上次调用的Parser()中了,因为Parser()中写了try{}catch{}.catch{}中的代码如下:
1: catch (int /*e*/)2: {3: #ifdef DEBUG4: fprintf(stderr, "H264VideoStreamParser::parse() EXCEPTION (This is normal behavior - *not* an error)\n");
5: #endif6: return 0; // the parsing got interrupted7: }
可见parser()此时返回0,parser()返回0就执行到MPEGVideoStreamFramer::continueReadProcessing()中的else{}部分了,回去看看吧,其实啥也没做.也就是说,第一次调用parser()时,它只是从ByteStreamFileSource获取数据,那么这个parser()获取数据后什么也不做,但实际上对NALU分析和处里在这次Parser()的调用中已经完成了,不是在它本身完成的,而是在它引起了parser()的嵌套调用中完成.好迷糊,理顺一下过程就知道了:
sink要获取数据,执行到MPEGVideoStreamFramer::continueReadProcessing(),MPEGVideoStreamFramer::continueReadProcessing调用parser(),parser()要使用数据时发现没有,于是ensureValidBytes1()被调用来从ByteStreamFileSource获取数据,取得数据后MPEGVideoStreamFramer::afterGettingBytes()被调用,并中转到MPEGVideoStreamFramer::continueReadProcessing(),MPEGVideoStreamFramer::continueReadProcessing()被嵌套调用!,MPEGVideoStreamFramer::continueReadProcessing()中又会调用parser(),此时parser()要使用数据时发现有数据了,所以就进行分析,分析出一个NALU后,返回到MPEGVideoStreamFramer::continueReadProcessing(),MPEGVideoStreamFramer::continueReadProcessing()会调用afterGetting(this)把数据返回给sink.sink处理完数据后返回到MPEGVideoStreamFramer::continueReadProcessing(),MPEGVideoStreamFramer::continueReadProcessing()再返回到ensureValidBytes1(),ensureValidBytes1()抛出异常返回到第一次被调用的parser()的catch{}中,parser()返回到第一次调用的MPEGVideoStreamFramer::continueReadProcessing()中,MPEGVideoStreamFramer::continueReadProcessing()发现parser()没有取得NALU,于是啥也不做,返回到sink中,sink会继续通过source->getNextFrame()->MPEGVideoStreamFramer::continueReadProcessing()...这样再次获取NALU.
可以看到,parser中是有自己的缓冲的,而且其大小是固定的:#define BANK_SIZE 150000
你自己写Source时,每次输出的是一帧数据,包含多个NALU,所以你只要确定你的一帧不超过150000字节,你就可以放心的往fTo中copy,如果你的帧太大,就改这个宏吧.