| 1 | /********** |
| 2 | This library is free software; you can redistribute it and/or modify it under |
| 3 | the terms of the GNU Lesser General Public License as published by the |
| 4 | Free Software Foundation; either version 3 of the License, or (at your |
| 5 | option) any later version. (See <http://www.gnu.org/copyleft/lesser.html>.) |
| 6 | |
| 7 | This library is distributed in the hope that it will be useful, but WITHOUT |
| 8 | ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS |
| 9 | FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for |
| 10 | more details. |
| 11 | |
| 12 | You should have received a copy of the GNU Lesser General Public License |
| 13 | along with this library; if not, write to the Free Software Foundation, Inc., |
| 14 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 15 | **********/ |
| 16 | // "liveMedia" |
| 17 | // Copyright (c) 1996-2020 Live Networks, Inc. All rights reserved. |
| 18 | // RTP sink for H.264 or H.265 video |
| 19 | // Implementation |
| 20 | |
| 21 | #include "H264or5VideoRTPSink.hh" |
| 22 | #include "H264or5VideoStreamFramer.hh" |
| 23 | |
| 24 | ////////// H264or5Fragmenter definition ////////// |
| 25 | |
| 26 | // Because of the ideosyncracies of the H.264 RTP payload format, we implement |
| 27 | // "H264or5VideoRTPSink" using a separate "H264or5Fragmenter" class that delivers, |
| 28 | // to the "H264or5VideoRTPSink", only fragments that will fit within an outgoing |
| 29 | // RTP packet. I.e., we implement fragmentation in this separate "H264or5Fragmenter" |
| 30 | // class, rather than in "H264or5VideoRTPSink". |
| 31 | // (Note: This class should be used only by "H264or5VideoRTPSink", or a subclass.) |
| 32 | |
| 33 | class H264or5Fragmenter: public FramedFilter { |
| 34 | public: |
| 35 | H264or5Fragmenter(int hNumber, UsageEnvironment& env, FramedSource* inputSource, |
| 36 | unsigned inputBufferMax, unsigned maxOutputPacketSize); |
| 37 | virtual ~H264or5Fragmenter(); |
| 38 | |
| 39 | Boolean lastFragmentCompletedNALUnit() const { return fLastFragmentCompletedNALUnit; } |
| 40 | |
| 41 | private: // redefined virtual functions: |
| 42 | virtual void doGetNextFrame(); |
| 43 | virtual void doStopGettingFrames(); |
| 44 | |
| 45 | private: |
| 46 | static void afterGettingFrame(void* clientData, unsigned frameSize, |
| 47 | unsigned numTruncatedBytes, |
| 48 | struct timeval presentationTime, |
| 49 | unsigned durationInMicroseconds); |
| 50 | void afterGettingFrame1(unsigned frameSize, |
| 51 | unsigned numTruncatedBytes, |
| 52 | struct timeval presentationTime, |
| 53 | unsigned durationInMicroseconds); |
| 54 | void reset(); |
| 55 | |
| 56 | private: |
| 57 | int fHNumber; |
| 58 | unsigned fInputBufferSize; |
| 59 | unsigned fMaxOutputPacketSize; |
| 60 | unsigned char* fInputBuffer; |
| 61 | unsigned fNumValidDataBytes; |
| 62 | unsigned fCurDataOffset; |
| 63 | unsigned fSaveNumTruncatedBytes; |
| 64 | Boolean fLastFragmentCompletedNALUnit; |
| 65 | }; |
| 66 | |
| 67 | |
| 68 | ////////// H264or5VideoRTPSink implementation ////////// |
| 69 | |
| 70 | H264or5VideoRTPSink |
| 71 | ::H264or5VideoRTPSink(int hNumber, |
| 72 | UsageEnvironment& env, Groupsock* RTPgs, unsigned char rtpPayloadFormat, |
| 73 | u_int8_t const* vps, unsigned vpsSize, |
| 74 | u_int8_t const* sps, unsigned spsSize, |
| 75 | u_int8_t const* pps, unsigned ppsSize) |
| 76 | : VideoRTPSink(env, RTPgs, rtpPayloadFormat, 90000, hNumber == 264 ? "H264" : "H265" ), |
| 77 | fHNumber(hNumber), fOurFragmenter(NULL), fFmtpSDPLine(NULL) { |
| 78 | if (vps != NULL) { |
| 79 | fVPSSize = vpsSize; |
| 80 | fVPS = new u_int8_t[fVPSSize]; |
| 81 | memmove(fVPS, vps, fVPSSize); |
| 82 | } else { |
| 83 | fVPSSize = 0; |
| 84 | fVPS = NULL; |
| 85 | } |
| 86 | if (sps != NULL) { |
| 87 | fSPSSize = spsSize; |
| 88 | fSPS = new u_int8_t[fSPSSize]; |
| 89 | memmove(fSPS, sps, fSPSSize); |
| 90 | } else { |
| 91 | fSPSSize = 0; |
| 92 | fSPS = NULL; |
| 93 | } |
| 94 | if (pps != NULL) { |
| 95 | fPPSSize = ppsSize; |
| 96 | fPPS = new u_int8_t[fPPSSize]; |
| 97 | memmove(fPPS, pps, fPPSSize); |
| 98 | } else { |
| 99 | fPPSSize = 0; |
| 100 | fPPS = NULL; |
| 101 | } |
| 102 | } |
| 103 | |
| 104 | H264or5VideoRTPSink::~H264or5VideoRTPSink() { |
| 105 | fSource = fOurFragmenter; // hack: in case "fSource" had gotten set to NULL before we were called |
| 106 | delete[] fFmtpSDPLine; |
| 107 | delete[] fVPS; delete[] fSPS; delete[] fPPS; |
| 108 | stopPlaying(); // call this now, because we won't have our 'fragmenter' when the base class destructor calls it later. |
| 109 | |
| 110 | // Close our 'fragmenter' as well: |
| 111 | Medium::close(fOurFragmenter); |
| 112 | fSource = NULL; // for the base class destructor, which gets called next |
| 113 | } |
| 114 | |
| 115 | Boolean H264or5VideoRTPSink::continuePlaying() { |
| 116 | // First, check whether we have a 'fragmenter' class set up yet. |
| 117 | // If not, create it now: |
| 118 | if (fOurFragmenter == NULL) { |
| 119 | fOurFragmenter = new H264or5Fragmenter(fHNumber, envir(), fSource, OutPacketBuffer::maxSize, |
| 120 | ourMaxPacketSize() - 12/*RTP hdr size*/); |
| 121 | } else { |
| 122 | fOurFragmenter->reassignInputSource(fSource); |
| 123 | } |
| 124 | fSource = fOurFragmenter; |
| 125 | |
| 126 | // Then call the parent class's implementation: |
| 127 | return MultiFramedRTPSink::continuePlaying(); |
| 128 | } |
| 129 | |
| 130 | void H264or5VideoRTPSink::doSpecialFrameHandling(unsigned /*fragmentationOffset*/, |
| 131 | unsigned char* /*frameStart*/, |
| 132 | unsigned /*numBytesInFrame*/, |
| 133 | struct timeval framePresentationTime, |
| 134 | unsigned /*numRemainingBytes*/) { |
| 135 | // Set the RTP 'M' (marker) bit iff |
| 136 | // 1/ The most recently delivered fragment was the end of (or the only fragment of) an NAL unit, and |
| 137 | // 2/ This NAL unit was the last NAL unit of an 'access unit' (i.e. video frame). |
| 138 | if (fOurFragmenter != NULL) { |
| 139 | H264or5VideoStreamFramer* framerSource |
| 140 | = (H264or5VideoStreamFramer*)(fOurFragmenter->inputSource()); |
| 141 | // This relies on our fragmenter's source being a "H264or5VideoStreamFramer". |
| 142 | if (((H264or5Fragmenter*)fOurFragmenter)->lastFragmentCompletedNALUnit() |
| 143 | && framerSource != NULL && framerSource->pictureEndMarker()) { |
| 144 | setMarkerBit(); |
| 145 | framerSource->pictureEndMarker() = False; |
| 146 | } |
| 147 | } |
| 148 | |
| 149 | setTimestamp(framePresentationTime); |
| 150 | } |
| 151 | |
| 152 | Boolean H264or5VideoRTPSink |
| 153 | ::frameCanAppearAfterPacketStart(unsigned char const* /*frameStart*/, |
| 154 | unsigned /*numBytesInFrame*/) const { |
| 155 | return False; |
| 156 | } |
| 157 | |
| 158 | |
| 159 | ////////// H264or5Fragmenter implementation ////////// |
| 160 | |
| 161 | H264or5Fragmenter::H264or5Fragmenter(int hNumber, |
| 162 | UsageEnvironment& env, FramedSource* inputSource, |
| 163 | unsigned inputBufferMax, unsigned maxOutputPacketSize) |
| 164 | : FramedFilter(env, inputSource), |
| 165 | fHNumber(hNumber), |
| 166 | fInputBufferSize(inputBufferMax+1), fMaxOutputPacketSize(maxOutputPacketSize) { |
| 167 | fInputBuffer = new unsigned char[fInputBufferSize]; |
| 168 | reset(); |
| 169 | } |
| 170 | |
| 171 | H264or5Fragmenter::~H264or5Fragmenter() { |
| 172 | delete[] fInputBuffer; |
| 173 | detachInputSource(); // so that the subsequent ~FramedFilter() doesn't delete it |
| 174 | } |
| 175 | |
| 176 | void H264or5Fragmenter::doGetNextFrame() { |
| 177 | if (fNumValidDataBytes == 1) { |
| 178 | // We have no NAL unit data currently in the buffer. Read a new one: |
| 179 | fInputSource->getNextFrame(&fInputBuffer[1], fInputBufferSize - 1, |
| 180 | afterGettingFrame, this, |
| 181 | FramedSource::handleClosure, this); |
| 182 | } else { |
| 183 | // We have NAL unit data in the buffer. There are three cases to consider: |
| 184 | // 1. There is a new NAL unit in the buffer, and it's small enough to deliver |
| 185 | // to the RTP sink (as is). |
| 186 | // 2. There is a new NAL unit in the buffer, but it's too large to deliver to |
| 187 | // the RTP sink in its entirety. Deliver the first fragment of this data, |
| 188 | // as a FU packet, with one extra preceding header byte (for the "FU header"). |
| 189 | // 3. There is a NAL unit in the buffer, and we've already delivered some |
| 190 | // fragment(s) of this. Deliver the next fragment of this data, |
| 191 | // as a FU packet, with two (H.264) or three (H.265) extra preceding header bytes |
| 192 | // (for the "NAL header" and the "FU header"). |
| 193 | |
| 194 | if (fMaxSize < fMaxOutputPacketSize) { // shouldn't happen |
| 195 | envir() << "H264or5Fragmenter::doGetNextFrame(): fMaxSize (" |
| 196 | << fMaxSize << ") is smaller than expected\n" ; |
| 197 | } else { |
| 198 | fMaxSize = fMaxOutputPacketSize; |
| 199 | } |
| 200 | |
| 201 | fLastFragmentCompletedNALUnit = True; // by default |
| 202 | if (fCurDataOffset == 1) { // case 1 or 2 |
| 203 | if (fNumValidDataBytes - 1 <= fMaxSize) { // case 1 |
| 204 | memmove(fTo, &fInputBuffer[1], fNumValidDataBytes - 1); |
| 205 | fFrameSize = fNumValidDataBytes - 1; |
| 206 | fCurDataOffset = fNumValidDataBytes; |
| 207 | } else { // case 2 |
| 208 | // We need to send the NAL unit data as FU packets. Deliver the first |
| 209 | // packet now. Note that we add "NAL header" and "FU header" bytes to the front |
| 210 | // of the packet (overwriting the existing "NAL header"). |
| 211 | if (fHNumber == 264) { |
| 212 | fInputBuffer[0] = (fInputBuffer[1] & 0xE0) | 28; // FU indicator |
| 213 | fInputBuffer[1] = 0x80 | (fInputBuffer[1] & 0x1F); // FU header (with S bit) |
| 214 | } else { // 265 |
| 215 | u_int8_t nal_unit_type = (fInputBuffer[1]&0x7E)>>1; |
| 216 | fInputBuffer[0] = (fInputBuffer[1] & 0x81) | (49<<1); // Payload header (1st byte) |
| 217 | fInputBuffer[1] = fInputBuffer[2]; // Payload header (2nd byte) |
| 218 | fInputBuffer[2] = 0x80 | nal_unit_type; // FU header (with S bit) |
| 219 | } |
| 220 | memmove(fTo, fInputBuffer, fMaxSize); |
| 221 | fFrameSize = fMaxSize; |
| 222 | fCurDataOffset += fMaxSize - 1; |
| 223 | fLastFragmentCompletedNALUnit = False; |
| 224 | } |
| 225 | } else { // case 3 |
| 226 | // We are sending this NAL unit data as FU packets. We've already sent the |
| 227 | // first packet (fragment). Now, send the next fragment. Note that we add |
| 228 | // "NAL header" and "FU header" bytes to the front. (We reuse these bytes that |
| 229 | // we already sent for the first fragment, but clear the S bit, and add the E |
| 230 | // bit if this is the last fragment.) |
| 231 | unsigned ; |
| 232 | if (fHNumber == 264) { |
| 233 | fInputBuffer[fCurDataOffset-2] = fInputBuffer[0]; // FU indicator |
| 234 | fInputBuffer[fCurDataOffset-1] = fInputBuffer[1]&~0x80; // FU header (no S bit) |
| 235 | numExtraHeaderBytes = 2; |
| 236 | } else { // 265 |
| 237 | fInputBuffer[fCurDataOffset-3] = fInputBuffer[0]; // Payload header (1st byte) |
| 238 | fInputBuffer[fCurDataOffset-2] = fInputBuffer[1]; // Payload header (2nd byte) |
| 239 | fInputBuffer[fCurDataOffset-1] = fInputBuffer[2]&~0x80; // FU header (no S bit) |
| 240 | numExtraHeaderBytes = 3; |
| 241 | } |
| 242 | unsigned numBytesToSend = numExtraHeaderBytes + (fNumValidDataBytes - fCurDataOffset); |
| 243 | if (numBytesToSend > fMaxSize) { |
| 244 | // We can't send all of the remaining data this time: |
| 245 | numBytesToSend = fMaxSize; |
| 246 | fLastFragmentCompletedNALUnit = False; |
| 247 | } else { |
| 248 | // This is the last fragment: |
| 249 | fInputBuffer[fCurDataOffset-1] |= 0x40; // set the E bit in the FU header |
| 250 | fNumTruncatedBytes = fSaveNumTruncatedBytes; |
| 251 | } |
| 252 | memmove(fTo, &fInputBuffer[fCurDataOffset-numExtraHeaderBytes], numBytesToSend); |
| 253 | fFrameSize = numBytesToSend; |
| 254 | fCurDataOffset += numBytesToSend - numExtraHeaderBytes; |
| 255 | } |
| 256 | |
| 257 | if (fCurDataOffset >= fNumValidDataBytes) { |
| 258 | // We're done with this data. Reset the pointers for receiving new data: |
| 259 | fNumValidDataBytes = fCurDataOffset = 1; |
| 260 | } |
| 261 | |
| 262 | // Complete delivery to the client: |
| 263 | FramedSource::afterGetting(this); |
| 264 | } |
| 265 | } |
| 266 | |
| 267 | void H264or5Fragmenter::doStopGettingFrames() { |
| 268 | // Make sure that we don't have any stale data fragments lying around, should we later resume: |
| 269 | reset(); |
| 270 | FramedFilter::doStopGettingFrames(); |
| 271 | } |
| 272 | |
| 273 | void H264or5Fragmenter::afterGettingFrame(void* clientData, unsigned frameSize, |
| 274 | unsigned numTruncatedBytes, |
| 275 | struct timeval presentationTime, |
| 276 | unsigned durationInMicroseconds) { |
| 277 | H264or5Fragmenter* fragmenter = (H264or5Fragmenter*)clientData; |
| 278 | fragmenter->afterGettingFrame1(frameSize, numTruncatedBytes, presentationTime, |
| 279 | durationInMicroseconds); |
| 280 | } |
| 281 | |
| 282 | void H264or5Fragmenter::afterGettingFrame1(unsigned frameSize, |
| 283 | unsigned numTruncatedBytes, |
| 284 | struct timeval presentationTime, |
| 285 | unsigned durationInMicroseconds) { |
| 286 | fNumValidDataBytes += frameSize; |
| 287 | fSaveNumTruncatedBytes = numTruncatedBytes; |
| 288 | fPresentationTime = presentationTime; |
| 289 | fDurationInMicroseconds = durationInMicroseconds; |
| 290 | |
| 291 | // Deliver data to the client: |
| 292 | doGetNextFrame(); |
| 293 | } |
| 294 | |
| 295 | void H264or5Fragmenter::reset() { |
| 296 | fNumValidDataBytes = fCurDataOffset = 1; |
| 297 | fSaveNumTruncatedBytes = 0; |
| 298 | fLastFragmentCompletedNALUnit = True; |
| 299 | } |
| 300 | |