MPEG4Writer.cpp revision 05e65807d48c2c7b954af82eda0532a48948e96a
1/* 2 * Copyright (C) 2009 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 * See the License for the specific language governing permissions and 14 * limitations under the License. 15 */ 16 17//#define LOG_NDEBUG 0 18#define LOG_TAG "MPEG4Writer" 19#include <utils/Log.h> 20 21#include <arpa/inet.h> 22 23#include <pthread.h> 24#include <sys/prctl.h> 25#include <sys/resource.h> 26 27#include <media/stagefright/MPEG4Writer.h> 28#include <media/stagefright/MediaBuffer.h> 29#include <media/stagefright/MetaData.h> 30#include <media/stagefright/MediaDebug.h> 31#include <media/stagefright/MediaDefs.h> 32#include <media/stagefright/MediaErrors.h> 33#include <media/stagefright/MediaSource.h> 34#include <media/stagefright/Utils.h> 35#include <media/mediarecorder.h> 36#include <cutils/properties.h> 37#include <sys/types.h> 38#include <sys/stat.h> 39#include <fcntl.h> 40#include <unistd.h> 41 42#include "include/ESDS.h" 43 44namespace android { 45 46static const int64_t kMax32BitFileSize = 0x007fffffffLL; 47static const uint8_t kNalUnitTypeSeqParamSet = 0x07; 48static const uint8_t kNalUnitTypePicParamSet = 0x08; 49static const int64_t kInitialDelayTimeUs = 700000LL; 50 51// Using longer adjustment period to suppress fluctuations in 52// the audio encoding paths 53static const int64_t kVideoMediaTimeAdjustPeriodTimeUs = 600000000LL; // 10 minutes 54 55class MPEG4Writer::Track { 56public: 57 Track(MPEG4Writer *owner, const sp<MediaSource> &source, size_t trackId); 58 59 ~Track(); 60 61 status_t start(MetaData *params); 62 status_t stop(); 63 status_t pause(); 64 bool reachedEOS(); 65 66 int64_t getDurationUs() const; 67 int64_t getEstimatedTrackSizeBytes() const; 68 void writeTrackHeader(bool use32BitOffset = true); 69 void bufferChunk(int64_t timestampUs); 70 bool isAvc() const { return mIsAvc; } 71 bool isAudio() const { return mIsAudio; } 72 bool isMPEG4() const { return mIsMPEG4; } 73 void addChunkOffset(off64_t offset); 74 int32_t getTrackId() const { return mTrackId; } 75 status_t dump(int fd, const Vector<String16>& args) const; 76 77private: 78 MPEG4Writer *mOwner; 79 sp<MetaData> mMeta; 80 sp<MediaSource> mSource; 81 volatile bool mDone; 82 volatile bool mPaused; 83 volatile bool mResumed; 84 volatile bool mStarted; 85 bool mIsAvc; 86 bool mIsAudio; 87 bool mIsMPEG4; 88 int32_t mTrackId; 89 int64_t mTrackDurationUs; 90 int64_t mMaxChunkDurationUs; 91 92 // For realtime applications, we need to adjust the media clock 93 // for video track based on the audio media clock 94 bool mIsRealTimeRecording; 95 int64_t mMaxTimeStampUs; 96 int64_t mEstimatedTrackSizeBytes; 97 int64_t mMdatSizeBytes; 98 int32_t mTimeScale; 99 100 pthread_t mThread; 101 102 // mNumSamples is used to track how many samples in mSampleSizes List. 103 // This is to reduce the cost associated with mSampleSizes.size() call, 104 // since it is O(n). Ideally, the fix should be in List class. 105 size_t mNumSamples; 106 List<size_t> mSampleSizes; 107 bool mSamplesHaveSameSize; 108 109 List<MediaBuffer *> mChunkSamples; 110 111 size_t mNumStcoTableEntries; 112 List<off64_t> mChunkOffsets; 113 114 size_t mNumStscTableEntries; 115 struct StscTableEntry { 116 117 StscTableEntry(uint32_t chunk, uint32_t samples, uint32_t id) 118 : firstChunk(chunk), 119 samplesPerChunk(samples), 120 sampleDescriptionId(id) {} 121 122 uint32_t firstChunk; 123 uint32_t samplesPerChunk; 124 uint32_t sampleDescriptionId; 125 }; 126 List<StscTableEntry> mStscTableEntries; 127 128 size_t mNumStssTableEntries; 129 List<int32_t> mStssTableEntries; 130 131 struct SttsTableEntry { 132 133 SttsTableEntry(uint32_t count, uint32_t duration) 134 : sampleCount(count), sampleDuration(duration) {} 135 136 uint32_t sampleCount; 137 uint32_t sampleDuration; // time scale based 138 }; 139 size_t mNumSttsTableEntries; 140 List<SttsTableEntry> mSttsTableEntries; 141 142 struct CttsTableEntry { 143 CttsTableEntry(uint32_t count, int32_t timescaledDur) 144 : sampleCount(count), sampleDuration(timescaledDur) {} 145 146 uint32_t sampleCount; 147 int32_t sampleDuration; // time scale based 148 }; 149 bool mHasNegativeCttsDeltaDuration; 150 size_t mNumCttsTableEntries; 151 List<CttsTableEntry> mCttsTableEntries; 152 153 // Sequence parameter set or picture parameter set 154 struct AVCParamSet { 155 AVCParamSet(uint16_t length, const uint8_t *data) 156 : mLength(length), mData(data) {} 157 158 uint16_t mLength; 159 const uint8_t *mData; 160 }; 161 List<AVCParamSet> mSeqParamSets; 162 List<AVCParamSet> mPicParamSets; 163 uint8_t mProfileIdc; 164 uint8_t mProfileCompatible; 165 uint8_t mLevelIdc; 166 167 void *mCodecSpecificData; 168 size_t mCodecSpecificDataSize; 169 bool mGotAllCodecSpecificData; 170 bool mTrackingProgressStatus; 171 172 bool mReachedEOS; 173 int64_t mStartTimestampUs; 174 int64_t mStartTimeRealUs; 175 int64_t mFirstSampleTimeRealUs; 176 int64_t mPreviousTrackTimeUs; 177 int64_t mTrackEveryTimeDurationUs; 178 179 // Has the media time adjustment for video started? 180 bool mIsMediaTimeAdjustmentOn; 181 // The time stamp when previous media time adjustment period starts 182 int64_t mPrevMediaTimeAdjustTimestampUs; 183 // Number of vidoe frames whose time stamp may be adjusted 184 int64_t mMediaTimeAdjustNumFrames; 185 // The sample number when previous meida time adjustmnet period starts 186 int64_t mPrevMediaTimeAdjustSample; 187 // The total accumulated drift time within a period of 188 // kVideoMediaTimeAdjustPeriodTimeUs. 189 int64_t mTotalDriftTimeToAdjustUs; 190 // The total accumalated drift time since the start of the recording 191 // excluding the current time adjustment period 192 int64_t mPrevTotalAccumDriftTimeUs; 193 194 // Update the audio track's drift information. 195 void updateDriftTime(const sp<MetaData>& meta); 196 197 // Adjust the time stamp of the video track according to 198 // the drift time information from the audio track. 199 void adjustMediaTime(int64_t *timestampUs); 200 201 static void *ThreadWrapper(void *me); 202 status_t threadEntry(); 203 204 const uint8_t *parseParamSet( 205 const uint8_t *data, size_t length, int type, size_t *paramSetLen); 206 207 status_t makeAVCCodecSpecificData(const uint8_t *data, size_t size); 208 status_t copyAVCCodecSpecificData(const uint8_t *data, size_t size); 209 status_t parseAVCCodecSpecificData(const uint8_t *data, size_t size); 210 211 // Track authoring progress status 212 void trackProgressStatus(int64_t timeUs, status_t err = OK); 213 void initTrackingProgressStatus(MetaData *params); 214 215 void getCodecSpecificDataFromInputFormatIfPossible(); 216 217 // Determine the track time scale 218 // If it is an audio track, try to use the sampling rate as 219 // the time scale; however, if user chooses the overwrite 220 // value, the user-supplied time scale will be used. 221 void setTimeScale(); 222 223 // Simple validation on the codec specific data 224 status_t checkCodecSpecificData() const; 225 int32_t mRotation; 226 227 void updateTrackSizeEstimate(); 228 void addOneStscTableEntry(size_t chunkId, size_t sampleId); 229 void addOneStssTableEntry(size_t sampleId); 230 231 // Duration is time scale based 232 void addOneSttsTableEntry(size_t sampleCount, int32_t timescaledDur); 233 void addOneCttsTableEntry(size_t sampleCount, int32_t timescaledDur); 234 void sendTrackSummary(bool hasMultipleTracks); 235 236 // Write the boxes 237 void writeStcoBox(bool use32BitOffset); 238 void writeStscBox(); 239 void writeStszBox(); 240 void writeStssBox(); 241 void writeSttsBox(); 242 void writeCttsBox(); 243 void writeD263Box(); 244 void writePaspBox(); 245 void writeAvccBox(); 246 void writeUrlBox(); 247 void writeDrefBox(); 248 void writeDinfBox(); 249 void writeDamrBox(); 250 void writeMdhdBox(time_t now); 251 void writeSmhdBox(); 252 void writeVmhdBox(); 253 void writeHdlrBox(); 254 void writeTkhdBox(time_t now); 255 void writeMp4aEsdsBox(); 256 void writeMp4vEsdsBox(); 257 void writeAudioFourCCBox(); 258 void writeVideoFourCCBox(); 259 void writeStblBox(bool use32BitOffset); 260 261 Track(const Track &); 262 Track &operator=(const Track &); 263}; 264 265MPEG4Writer::MPEG4Writer(const char *filename) 266 : mFd(-1), 267 mInitCheck(NO_INIT), 268 mUse4ByteNalLength(true), 269 mUse32BitOffset(true), 270 mIsFileSizeLimitExplicitlyRequested(false), 271 mPaused(false), 272 mStarted(false), 273 mOffset(0), 274 mMdatOffset(0), 275 mEstimatedMoovBoxSize(0), 276 mInterleaveDurationUs(1000000), 277 mLatitudex10000(0), 278 mLongitudex10000(0), 279 mAreGeoTagsAvailable(false), 280 mStartTimeOffsetMs(-1) { 281 282 mFd = open(filename, O_CREAT | O_LARGEFILE | O_TRUNC | O_RDWR); 283 if (mFd >= 0) { 284 mInitCheck = OK; 285 } 286} 287 288MPEG4Writer::MPEG4Writer(int fd) 289 : mFd(dup(fd)), 290 mInitCheck(mFd < 0? NO_INIT: OK), 291 mUse4ByteNalLength(true), 292 mUse32BitOffset(true), 293 mIsFileSizeLimitExplicitlyRequested(false), 294 mPaused(false), 295 mStarted(false), 296 mOffset(0), 297 mMdatOffset(0), 298 mEstimatedMoovBoxSize(0), 299 mInterleaveDurationUs(1000000), 300 mLatitudex10000(0), 301 mLongitudex10000(0), 302 mAreGeoTagsAvailable(false), 303 mStartTimeOffsetMs(-1) { 304} 305 306MPEG4Writer::~MPEG4Writer() { 307 stop(); 308 309 while (!mTracks.empty()) { 310 List<Track *>::iterator it = mTracks.begin(); 311 delete *it; 312 (*it) = NULL; 313 mTracks.erase(it); 314 } 315 mTracks.clear(); 316} 317 318status_t MPEG4Writer::dump( 319 int fd, const Vector<String16>& args) { 320 const size_t SIZE = 256; 321 char buffer[SIZE]; 322 String8 result; 323 snprintf(buffer, SIZE, " MPEG4Writer %p\n", this); 324 result.append(buffer); 325 snprintf(buffer, SIZE, " mStarted: %s\n", mStarted? "true": "false"); 326 result.append(buffer); 327 ::write(fd, result.string(), result.size()); 328 for (List<Track *>::iterator it = mTracks.begin(); 329 it != mTracks.end(); ++it) { 330 (*it)->dump(fd, args); 331 } 332 return OK; 333} 334 335status_t MPEG4Writer::Track::dump( 336 int fd, const Vector<String16>& args) const { 337 const size_t SIZE = 256; 338 char buffer[SIZE]; 339 String8 result; 340 snprintf(buffer, SIZE, " %s track\n", mIsAudio? "Audio": "Video"); 341 result.append(buffer); 342 snprintf(buffer, SIZE, " reached EOS: %s\n", 343 mReachedEOS? "true": "false"); 344 result.append(buffer); 345 ::write(fd, result.string(), result.size()); 346 return OK; 347} 348 349status_t MPEG4Writer::addSource(const sp<MediaSource> &source) { 350 Mutex::Autolock l(mLock); 351 if (mStarted) { 352 LOGE("Attempt to add source AFTER recording is started"); 353 return UNKNOWN_ERROR; 354 } 355 Track *track = new Track(this, source, mTracks.size()); 356 mTracks.push_back(track); 357 358 return OK; 359} 360 361status_t MPEG4Writer::startTracks(MetaData *params) { 362 for (List<Track *>::iterator it = mTracks.begin(); 363 it != mTracks.end(); ++it) { 364 status_t err = (*it)->start(params); 365 366 if (err != OK) { 367 for (List<Track *>::iterator it2 = mTracks.begin(); 368 it2 != it; ++it2) { 369 (*it2)->stop(); 370 } 371 372 return err; 373 } 374 } 375 return OK; 376} 377 378int64_t MPEG4Writer::estimateMoovBoxSize(int32_t bitRate) { 379 // This implementation is highly experimental/heurisitic. 380 // 381 // Statistical analysis shows that metadata usually accounts 382 // for a small portion of the total file size, usually < 0.6%. 383 384 // The default MIN_MOOV_BOX_SIZE is set to 0.6% x 1MB / 2, 385 // where 1MB is the common file size limit for MMS application. 386 // The default MAX _MOOV_BOX_SIZE value is based on about 3 387 // minute video recording with a bit rate about 3 Mbps, because 388 // statistics also show that most of the video captured are going 389 // to be less than 3 minutes. 390 391 // If the estimation is wrong, we will pay the price of wasting 392 // some reserved space. This should not happen so often statistically. 393 static const int32_t factor = mUse32BitOffset? 1: 2; 394 static const int64_t MIN_MOOV_BOX_SIZE = 3 * 1024; // 3 KB 395 static const int64_t MAX_MOOV_BOX_SIZE = (180 * 3000000 * 6LL / 8000); 396 int64_t size = MIN_MOOV_BOX_SIZE; 397 398 // Max file size limit is set 399 if (mMaxFileSizeLimitBytes != 0 && mIsFileSizeLimitExplicitlyRequested) { 400 size = mMaxFileSizeLimitBytes * 6 / 1000; 401 } 402 403 // Max file duration limit is set 404 if (mMaxFileDurationLimitUs != 0) { 405 if (bitRate > 0) { 406 int64_t size2 = 407 ((mMaxFileDurationLimitUs * bitRate * 6) / 1000 / 8000000); 408 if (mMaxFileSizeLimitBytes != 0 && mIsFileSizeLimitExplicitlyRequested) { 409 // When both file size and duration limits are set, 410 // we use the smaller limit of the two. 411 if (size > size2) { 412 size = size2; 413 } 414 } else { 415 // Only max file duration limit is set 416 size = size2; 417 } 418 } 419 } 420 421 if (size < MIN_MOOV_BOX_SIZE) { 422 size = MIN_MOOV_BOX_SIZE; 423 } 424 425 // Any long duration recording will be probably end up with 426 // non-streamable mp4 file. 427 if (size > MAX_MOOV_BOX_SIZE) { 428 size = MAX_MOOV_BOX_SIZE; 429 } 430 431 LOGI("limits: %lld/%lld bytes/us, bit rate: %d bps and the estimated" 432 " moov size %lld bytes", 433 mMaxFileSizeLimitBytes, mMaxFileDurationLimitUs, bitRate, size); 434 return factor * size; 435} 436 437status_t MPEG4Writer::start(MetaData *param) { 438 if (mInitCheck != OK) { 439 return UNKNOWN_ERROR; 440 } 441 442 /* 443 * Check mMaxFileSizeLimitBytes at the beginning 444 * since mMaxFileSizeLimitBytes may be implicitly 445 * changed later for 32-bit file offset even if 446 * user does not ask to set it explicitly. 447 */ 448 if (mMaxFileSizeLimitBytes != 0) { 449 mIsFileSizeLimitExplicitlyRequested = true; 450 } 451 452 int32_t use64BitOffset; 453 if (param && 454 param->findInt32(kKey64BitFileOffset, &use64BitOffset) && 455 use64BitOffset) { 456 mUse32BitOffset = false; 457 } 458 459 if (mUse32BitOffset) { 460 // Implicit 32 bit file size limit 461 if (mMaxFileSizeLimitBytes == 0) { 462 mMaxFileSizeLimitBytes = kMax32BitFileSize; 463 } 464 465 // If file size is set to be larger than the 32 bit file 466 // size limit, treat it as an error. 467 if (mMaxFileSizeLimitBytes > kMax32BitFileSize) { 468 LOGW("32-bit file size limit (%lld bytes) too big. " 469 "It is changed to %lld bytes", 470 mMaxFileSizeLimitBytes, kMax32BitFileSize); 471 mMaxFileSizeLimitBytes = kMax32BitFileSize; 472 } 473 } 474 475 int32_t use2ByteNalLength; 476 if (param && 477 param->findInt32(kKey2ByteNalLength, &use2ByteNalLength) && 478 use2ByteNalLength) { 479 mUse4ByteNalLength = false; 480 } 481 482 mStartTimestampUs = -1; 483 484 if (mStarted) { 485 if (mPaused) { 486 mPaused = false; 487 return startTracks(param); 488 } 489 return OK; 490 } 491 492 if (!param || 493 !param->findInt32(kKeyTimeScale, &mTimeScale)) { 494 mTimeScale = 1000; 495 } 496 CHECK(mTimeScale > 0); 497 LOGV("movie time scale: %d", mTimeScale); 498 499 mStreamableFile = true; 500 mWriteMoovBoxToMemory = false; 501 mMoovBoxBuffer = NULL; 502 mMoovBoxBufferOffset = 0; 503 504 writeFtypBox(param); 505 506 mFreeBoxOffset = mOffset; 507 508 if (mEstimatedMoovBoxSize == 0) { 509 int32_t bitRate = -1; 510 if (param) { 511 param->findInt32(kKeyBitRate, &bitRate); 512 } 513 mEstimatedMoovBoxSize = estimateMoovBoxSize(bitRate); 514 } 515 CHECK(mEstimatedMoovBoxSize >= 8); 516 lseek64(mFd, mFreeBoxOffset, SEEK_SET); 517 writeInt32(mEstimatedMoovBoxSize); 518 write("free", 4); 519 520 mMdatOffset = mFreeBoxOffset + mEstimatedMoovBoxSize; 521 mOffset = mMdatOffset; 522 lseek64(mFd, mMdatOffset, SEEK_SET); 523 if (mUse32BitOffset) { 524 write("????mdat", 8); 525 } else { 526 write("\x00\x00\x00\x01mdat????????", 16); 527 } 528 529 status_t err = startWriterThread(); 530 if (err != OK) { 531 return err; 532 } 533 534 err = startTracks(param); 535 if (err != OK) { 536 return err; 537 } 538 539 mStarted = true; 540 return OK; 541} 542 543bool MPEG4Writer::use32BitFileOffset() const { 544 return mUse32BitOffset; 545} 546 547status_t MPEG4Writer::pause() { 548 if (mInitCheck != OK) { 549 return OK; 550 } 551 mPaused = true; 552 status_t err = OK; 553 for (List<Track *>::iterator it = mTracks.begin(); 554 it != mTracks.end(); ++it) { 555 status_t status = (*it)->pause(); 556 if (status != OK) { 557 err = status; 558 } 559 } 560 return err; 561} 562 563void MPEG4Writer::stopWriterThread() { 564 LOGD("Stopping writer thread"); 565 566 { 567 Mutex::Autolock autolock(mLock); 568 569 mDone = true; 570 mChunkReadyCondition.signal(); 571 } 572 573 void *dummy; 574 pthread_join(mThread, &dummy); 575 LOGD("Writer thread stopped"); 576} 577 578/* 579 * MP4 file standard defines a composition matrix: 580 * | a b u | 581 * | c d v | 582 * | x y w | 583 * 584 * the element in the matrix is stored in the following 585 * order: {a, b, u, c, d, v, x, y, w}, 586 * where a, b, c, d, x, and y is in 16.16 format, while 587 * u, v and w is in 2.30 format. 588 */ 589void MPEG4Writer::writeCompositionMatrix(int degrees) { 590 LOGV("writeCompositionMatrix"); 591 uint32_t a = 0x00010000; 592 uint32_t b = 0; 593 uint32_t c = 0; 594 uint32_t d = 0x00010000; 595 switch (degrees) { 596 case 0: 597 break; 598 case 90: 599 a = 0; 600 b = 0x00010000; 601 c = 0xFFFF0000; 602 d = 0; 603 break; 604 case 180: 605 a = 0xFFFF0000; 606 d = 0xFFFF0000; 607 break; 608 case 270: 609 a = 0; 610 b = 0xFFFF0000; 611 c = 0x00010000; 612 d = 0; 613 break; 614 default: 615 CHECK(!"Should never reach this unknown rotation"); 616 break; 617 } 618 619 writeInt32(a); // a 620 writeInt32(b); // b 621 writeInt32(0); // u 622 writeInt32(c); // c 623 writeInt32(d); // d 624 writeInt32(0); // v 625 writeInt32(0); // x 626 writeInt32(0); // y 627 writeInt32(0x40000000); // w 628} 629 630 631status_t MPEG4Writer::stop() { 632 if (mInitCheck != OK) { 633 return OK; 634 } 635 636 status_t err = OK; 637 int64_t maxDurationUs = 0; 638 int64_t minDurationUs = 0x7fffffffffffffffLL; 639 for (List<Track *>::iterator it = mTracks.begin(); 640 it != mTracks.end(); ++it) { 641 status_t status = (*it)->stop(); 642 if (err == OK && status != OK) { 643 err = status; 644 } 645 646 int64_t durationUs = (*it)->getDurationUs(); 647 if (durationUs > maxDurationUs) { 648 maxDurationUs = durationUs; 649 } 650 if (durationUs < minDurationUs) { 651 minDurationUs = durationUs; 652 } 653 } 654 655 if (mTracks.size() > 1) { 656 LOGD("Duration from tracks range is [%lld, %lld] us", 657 minDurationUs, maxDurationUs); 658 } 659 660 stopWriterThread(); 661 662 // Do not write out movie header on error. 663 if (err != OK) { 664 close(mFd); 665 mFd = -1; 666 mInitCheck = NO_INIT; 667 mStarted = false; 668 return err; 669 } 670 671 // Fix up the size of the 'mdat' chunk. 672 if (mUse32BitOffset) { 673 lseek64(mFd, mMdatOffset, SEEK_SET); 674 int32_t size = htonl(static_cast<int32_t>(mOffset - mMdatOffset)); 675 ::write(mFd, &size, 4); 676 } else { 677 lseek64(mFd, mMdatOffset + 8, SEEK_SET); 678 int64_t size = mOffset - mMdatOffset; 679 size = hton64(size); 680 ::write(mFd, &size, 8); 681 } 682 lseek64(mFd, mOffset, SEEK_SET); 683 684 const off64_t moovOffset = mOffset; 685 mWriteMoovBoxToMemory = true; 686 mMoovBoxBuffer = (uint8_t *) malloc(mEstimatedMoovBoxSize); 687 mMoovBoxBufferOffset = 0; 688 CHECK(mMoovBoxBuffer != NULL); 689 writeMoovBox(maxDurationUs); 690 691 mWriteMoovBoxToMemory = false; 692 if (mStreamableFile) { 693 CHECK(mMoovBoxBufferOffset + 8 <= mEstimatedMoovBoxSize); 694 695 // Moov box 696 lseek64(mFd, mFreeBoxOffset, SEEK_SET); 697 mOffset = mFreeBoxOffset; 698 write(mMoovBoxBuffer, 1, mMoovBoxBufferOffset); 699 700 // Free box 701 lseek64(mFd, mOffset, SEEK_SET); 702 writeInt32(mEstimatedMoovBoxSize - mMoovBoxBufferOffset); 703 write("free", 4); 704 705 // Free temp memory 706 free(mMoovBoxBuffer); 707 mMoovBoxBuffer = NULL; 708 mMoovBoxBufferOffset = 0; 709 } else { 710 LOGI("The mp4 file will not be streamable."); 711 } 712 713 CHECK(mBoxes.empty()); 714 715 close(mFd); 716 mFd = -1; 717 mInitCheck = NO_INIT; 718 mStarted = false; 719 720 return err; 721} 722 723void MPEG4Writer::writeMvhdBox(int64_t durationUs) { 724 time_t now = time(NULL); 725 beginBox("mvhd"); 726 writeInt32(0); // version=0, flags=0 727 writeInt32(now); // creation time 728 writeInt32(now); // modification time 729 writeInt32(mTimeScale); // mvhd timescale 730 int32_t duration = (durationUs * mTimeScale + 5E5) / 1E6; 731 writeInt32(duration); 732 writeInt32(0x10000); // rate: 1.0 733 writeInt16(0x100); // volume 734 writeInt16(0); // reserved 735 writeInt32(0); // reserved 736 writeInt32(0); // reserved 737 writeCompositionMatrix(0); // matrix 738 writeInt32(0); // predefined 739 writeInt32(0); // predefined 740 writeInt32(0); // predefined 741 writeInt32(0); // predefined 742 writeInt32(0); // predefined 743 writeInt32(0); // predefined 744 writeInt32(mTracks.size() + 1); // nextTrackID 745 endBox(); // mvhd 746} 747 748void MPEG4Writer::writeMoovBox(int64_t durationUs) { 749 beginBox("moov"); 750 writeMvhdBox(durationUs); 751 if (mAreGeoTagsAvailable) { 752 writeUdtaBox(); 753 } 754 int32_t id = 1; 755 for (List<Track *>::iterator it = mTracks.begin(); 756 it != mTracks.end(); ++it, ++id) { 757 (*it)->writeTrackHeader(mUse32BitOffset); 758 } 759 endBox(); // moov 760} 761 762void MPEG4Writer::writeFtypBox(MetaData *param) { 763 beginBox("ftyp"); 764 765 int32_t fileType; 766 if (param && param->findInt32(kKeyFileType, &fileType) && 767 fileType != OUTPUT_FORMAT_MPEG_4) { 768 writeFourcc("3gp4"); 769 } else { 770 writeFourcc("isom"); 771 } 772 773 writeInt32(0); 774 writeFourcc("isom"); 775 writeFourcc("3gp4"); 776 endBox(); 777} 778 779static bool isTestModeEnabled() { 780#if (PROPERTY_VALUE_MAX < 5) 781#error "PROPERTY_VALUE_MAX must be at least 5" 782#endif 783 784 // Test mode is enabled only if rw.media.record.test system 785 // property is enabled. 786 char value[PROPERTY_VALUE_MAX]; 787 if (property_get("rw.media.record.test", value, NULL) && 788 (!strcasecmp(value, "true") || !strcasecmp(value, "1"))) { 789 return true; 790 } 791 return false; 792} 793 794void MPEG4Writer::sendSessionSummary() { 795 // Send session summary only if test mode is enabled 796 if (!isTestModeEnabled()) { 797 return; 798 } 799 800 for (List<ChunkInfo>::iterator it = mChunkInfos.begin(); 801 it != mChunkInfos.end(); ++it) { 802 int trackNum = it->mTrack->getTrackId() << 28; 803 notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 804 trackNum | MEDIA_RECORDER_TRACK_INTER_CHUNK_TIME_MS, 805 it->mMaxInterChunkDurUs); 806 } 807} 808 809status_t MPEG4Writer::setInterleaveDuration(uint32_t durationUs) { 810 mInterleaveDurationUs = durationUs; 811 return OK; 812} 813 814void MPEG4Writer::lock() { 815 mLock.lock(); 816} 817 818void MPEG4Writer::unlock() { 819 mLock.unlock(); 820} 821 822off64_t MPEG4Writer::addSample_l(MediaBuffer *buffer) { 823 off64_t old_offset = mOffset; 824 825 ::write(mFd, 826 (const uint8_t *)buffer->data() + buffer->range_offset(), 827 buffer->range_length()); 828 829 mOffset += buffer->range_length(); 830 831 return old_offset; 832} 833 834static void StripStartcode(MediaBuffer *buffer) { 835 if (buffer->range_length() < 4) { 836 return; 837 } 838 839 const uint8_t *ptr = 840 (const uint8_t *)buffer->data() + buffer->range_offset(); 841 842 if (!memcmp(ptr, "\x00\x00\x00\x01", 4)) { 843 buffer->set_range( 844 buffer->range_offset() + 4, buffer->range_length() - 4); 845 } 846} 847 848off64_t MPEG4Writer::addLengthPrefixedSample_l(MediaBuffer *buffer) { 849 off64_t old_offset = mOffset; 850 851 size_t length = buffer->range_length(); 852 853 if (mUse4ByteNalLength) { 854 uint8_t x = length >> 24; 855 ::write(mFd, &x, 1); 856 x = (length >> 16) & 0xff; 857 ::write(mFd, &x, 1); 858 x = (length >> 8) & 0xff; 859 ::write(mFd, &x, 1); 860 x = length & 0xff; 861 ::write(mFd, &x, 1); 862 863 ::write(mFd, 864 (const uint8_t *)buffer->data() + buffer->range_offset(), 865 length); 866 867 mOffset += length + 4; 868 } else { 869 CHECK(length < 65536); 870 871 uint8_t x = length >> 8; 872 ::write(mFd, &x, 1); 873 x = length & 0xff; 874 ::write(mFd, &x, 1); 875 ::write(mFd, (const uint8_t *)buffer->data() + buffer->range_offset(), length); 876 mOffset += length + 2; 877 } 878 879 return old_offset; 880} 881 882size_t MPEG4Writer::write( 883 const void *ptr, size_t size, size_t nmemb) { 884 885 const size_t bytes = size * nmemb; 886 if (mWriteMoovBoxToMemory) { 887 // This happens only when we write the moov box at the end of 888 // recording, not for each output video/audio frame we receive. 889 off64_t moovBoxSize = 8 + mMoovBoxBufferOffset + bytes; 890 if (moovBoxSize > mEstimatedMoovBoxSize) { 891 for (List<off64_t>::iterator it = mBoxes.begin(); 892 it != mBoxes.end(); ++it) { 893 (*it) += mOffset; 894 } 895 lseek64(mFd, mOffset, SEEK_SET); 896 ::write(mFd, mMoovBoxBuffer, mMoovBoxBufferOffset); 897 ::write(mFd, ptr, size * nmemb); 898 mOffset += (bytes + mMoovBoxBufferOffset); 899 free(mMoovBoxBuffer); 900 mMoovBoxBuffer = NULL; 901 mMoovBoxBufferOffset = 0; 902 mWriteMoovBoxToMemory = false; 903 mStreamableFile = false; 904 } else { 905 memcpy(mMoovBoxBuffer + mMoovBoxBufferOffset, ptr, bytes); 906 mMoovBoxBufferOffset += bytes; 907 } 908 } else { 909 ::write(mFd, ptr, size * nmemb); 910 mOffset += bytes; 911 } 912 return bytes; 913} 914 915void MPEG4Writer::beginBox(const char *fourcc) { 916 CHECK_EQ(strlen(fourcc), 4); 917 918 mBoxes.push_back(mWriteMoovBoxToMemory? 919 mMoovBoxBufferOffset: mOffset); 920 921 writeInt32(0); 922 writeFourcc(fourcc); 923} 924 925void MPEG4Writer::endBox() { 926 CHECK(!mBoxes.empty()); 927 928 off64_t offset = *--mBoxes.end(); 929 mBoxes.erase(--mBoxes.end()); 930 931 if (mWriteMoovBoxToMemory) { 932 int32_t x = htonl(mMoovBoxBufferOffset - offset); 933 memcpy(mMoovBoxBuffer + offset, &x, 4); 934 } else { 935 lseek64(mFd, offset, SEEK_SET); 936 writeInt32(mOffset - offset); 937 mOffset -= 4; 938 lseek64(mFd, mOffset, SEEK_SET); 939 } 940} 941 942void MPEG4Writer::writeInt8(int8_t x) { 943 write(&x, 1, 1); 944} 945 946void MPEG4Writer::writeInt16(int16_t x) { 947 x = htons(x); 948 write(&x, 1, 2); 949} 950 951void MPEG4Writer::writeInt32(int32_t x) { 952 x = htonl(x); 953 write(&x, 1, 4); 954} 955 956void MPEG4Writer::writeInt64(int64_t x) { 957 x = hton64(x); 958 write(&x, 1, 8); 959} 960 961void MPEG4Writer::writeCString(const char *s) { 962 size_t n = strlen(s); 963 write(s, 1, n + 1); 964} 965 966void MPEG4Writer::writeFourcc(const char *s) { 967 CHECK_EQ(strlen(s), 4); 968 write(s, 1, 4); 969} 970 971 972// Written in +/-DD.DDDD format 973void MPEG4Writer::writeLatitude(int degreex10000) { 974 bool isNegative = (degreex10000 < 0); 975 char sign = isNegative? '-': '+'; 976 977 // Handle the whole part 978 char str[9]; 979 int wholePart = degreex10000 / 10000; 980 if (wholePart == 0) { 981 snprintf(str, 5, "%c%.2d.", sign, wholePart); 982 } else { 983 snprintf(str, 5, "%+.2d.", wholePart); 984 } 985 986 // Handle the fractional part 987 int fractionalPart = degreex10000 - (wholePart * 10000); 988 if (fractionalPart < 0) { 989 fractionalPart = -fractionalPart; 990 } 991 snprintf(&str[4], 5, "%.4d", fractionalPart); 992 993 // Do not write the null terminator 994 write(str, 1, 8); 995} 996 997// Written in +/- DDD.DDDD format 998void MPEG4Writer::writeLongitude(int degreex10000) { 999 bool isNegative = (degreex10000 < 0); 1000 char sign = isNegative? '-': '+'; 1001 1002 // Handle the whole part 1003 char str[10]; 1004 int wholePart = degreex10000 / 10000; 1005 if (wholePart == 0) { 1006 snprintf(str, 6, "%c%.3d.", sign, wholePart); 1007 } else { 1008 snprintf(str, 6, "%+.3d.", wholePart); 1009 } 1010 1011 // Handle the fractional part 1012 int fractionalPart = degreex10000 - (wholePart * 10000); 1013 if (fractionalPart < 0) { 1014 fractionalPart = -fractionalPart; 1015 } 1016 snprintf(&str[5], 5, "%.4d", fractionalPart); 1017 1018 // Do not write the null terminator 1019 write(str, 1, 9); 1020} 1021 1022/* 1023 * Geodata is stored according to ISO-6709 standard. 1024 * latitudex10000 is latitude in degrees times 10000, and 1025 * longitudex10000 is longitude in degrees times 10000. 1026 * The range for the latitude is in [-90, +90], and 1027 * The range for the longitude is in [-180, +180] 1028 */ 1029status_t MPEG4Writer::setGeoData(int latitudex10000, int longitudex10000) { 1030 // Is latitude or longitude out of range? 1031 if (latitudex10000 < -900000 || latitudex10000 > 900000 || 1032 longitudex10000 < -1800000 || longitudex10000 > 1800000) { 1033 return BAD_VALUE; 1034 } 1035 1036 mLatitudex10000 = latitudex10000; 1037 mLongitudex10000 = longitudex10000; 1038 mAreGeoTagsAvailable = true; 1039 return OK; 1040} 1041 1042void MPEG4Writer::write(const void *data, size_t size) { 1043 write(data, 1, size); 1044} 1045 1046bool MPEG4Writer::isFileStreamable() const { 1047 return mStreamableFile; 1048} 1049 1050bool MPEG4Writer::exceedsFileSizeLimit() { 1051 // No limit 1052 if (mMaxFileSizeLimitBytes == 0) { 1053 return false; 1054 } 1055 1056 int64_t nTotalBytesEstimate = static_cast<int64_t>(mEstimatedMoovBoxSize); 1057 for (List<Track *>::iterator it = mTracks.begin(); 1058 it != mTracks.end(); ++it) { 1059 nTotalBytesEstimate += (*it)->getEstimatedTrackSizeBytes(); 1060 } 1061 1062 // Be conservative in the estimate: do not exceed 95% of 1063 // the target file limit. For small target file size limit, though, 1064 // this will not help. 1065 return (nTotalBytesEstimate >= (95 * mMaxFileSizeLimitBytes) / 100); 1066} 1067 1068bool MPEG4Writer::exceedsFileDurationLimit() { 1069 // No limit 1070 if (mMaxFileDurationLimitUs == 0) { 1071 return false; 1072 } 1073 1074 for (List<Track *>::iterator it = mTracks.begin(); 1075 it != mTracks.end(); ++it) { 1076 if ((*it)->getDurationUs() >= mMaxFileDurationLimitUs) { 1077 return true; 1078 } 1079 } 1080 return false; 1081} 1082 1083bool MPEG4Writer::reachedEOS() { 1084 bool allDone = true; 1085 for (List<Track *>::iterator it = mTracks.begin(); 1086 it != mTracks.end(); ++it) { 1087 if (!(*it)->reachedEOS()) { 1088 allDone = false; 1089 break; 1090 } 1091 } 1092 1093 return allDone; 1094} 1095 1096void MPEG4Writer::setStartTimestampUs(int64_t timeUs) { 1097 LOGI("setStartTimestampUs: %lld", timeUs); 1098 CHECK(timeUs >= 0); 1099 Mutex::Autolock autoLock(mLock); 1100 if (mStartTimestampUs < 0 || mStartTimestampUs > timeUs) { 1101 mStartTimestampUs = timeUs; 1102 LOGI("Earliest track starting time: %lld", mStartTimestampUs); 1103 } 1104} 1105 1106int64_t MPEG4Writer::getStartTimestampUs() { 1107 Mutex::Autolock autoLock(mLock); 1108 return mStartTimestampUs; 1109} 1110 1111size_t MPEG4Writer::numTracks() { 1112 Mutex::Autolock autolock(mLock); 1113 return mTracks.size(); 1114} 1115 1116//////////////////////////////////////////////////////////////////////////////// 1117 1118MPEG4Writer::Track::Track( 1119 MPEG4Writer *owner, const sp<MediaSource> &source, size_t trackId) 1120 : mOwner(owner), 1121 mMeta(source->getFormat()), 1122 mSource(source), 1123 mDone(false), 1124 mPaused(false), 1125 mResumed(false), 1126 mStarted(false), 1127 mTrackId(trackId), 1128 mTrackDurationUs(0), 1129 mEstimatedTrackSizeBytes(0), 1130 mSamplesHaveSameSize(true), 1131 mCodecSpecificData(NULL), 1132 mCodecSpecificDataSize(0), 1133 mGotAllCodecSpecificData(false), 1134 mReachedEOS(false), 1135 mRotation(0) { 1136 getCodecSpecificDataFromInputFormatIfPossible(); 1137 1138 const char *mime; 1139 mMeta->findCString(kKeyMIMEType, &mime); 1140 mIsAvc = !strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_AVC); 1141 mIsAudio = !strncasecmp(mime, "audio/", 6); 1142 mIsMPEG4 = !strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_MPEG4) || 1143 !strcasecmp(mime, MEDIA_MIMETYPE_AUDIO_AAC); 1144 1145 setTimeScale(); 1146} 1147 1148void MPEG4Writer::Track::updateTrackSizeEstimate() { 1149 1150 int64_t stcoBoxSizeBytes = mOwner->use32BitFileOffset() 1151 ? mNumStcoTableEntries * 4 1152 : mNumStcoTableEntries * 8; 1153 1154 int64_t stszBoxSizeBytes = mSamplesHaveSameSize? 4: (mNumSamples * 4); 1155 1156 mEstimatedTrackSizeBytes = mMdatSizeBytes; // media data size 1157 if (!mOwner->isFileStreamable()) { 1158 // Reserved free space is not large enough to hold 1159 // all meta data and thus wasted. 1160 mEstimatedTrackSizeBytes += mNumStscTableEntries * 12 + // stsc box size 1161 mNumStssTableEntries * 4 + // stss box size 1162 mNumSttsTableEntries * 8 + // stts box size 1163 mNumCttsTableEntries * 8 + // ctts box size 1164 stcoBoxSizeBytes + // stco box size 1165 stszBoxSizeBytes; // stsz box size 1166 } 1167} 1168 1169void MPEG4Writer::Track::addOneStscTableEntry( 1170 size_t chunkId, size_t sampleId) { 1171 1172 StscTableEntry stscEntry(chunkId, sampleId, 1); 1173 mStscTableEntries.push_back(stscEntry); 1174 ++mNumStscTableEntries; 1175} 1176 1177void MPEG4Writer::Track::addOneStssTableEntry(size_t sampleId) { 1178 mStssTableEntries.push_back(sampleId); 1179 ++mNumStssTableEntries; 1180} 1181 1182void MPEG4Writer::Track::addOneSttsTableEntry( 1183 size_t sampleCount, int32_t duration) { 1184 1185 SttsTableEntry sttsEntry(sampleCount, duration); 1186 mSttsTableEntries.push_back(sttsEntry); 1187 ++mNumSttsTableEntries; 1188} 1189 1190void MPEG4Writer::Track::addOneCttsTableEntry( 1191 size_t sampleCount, int32_t duration) { 1192 1193 if (mIsAudio) { 1194 return; 1195 } 1196 if (duration < 0 && !mHasNegativeCttsDeltaDuration) { 1197 mHasNegativeCttsDeltaDuration = true; 1198 } 1199 CttsTableEntry cttsEntry(sampleCount, duration); 1200 mCttsTableEntries.push_back(cttsEntry); 1201 ++mNumCttsTableEntries; 1202} 1203 1204void MPEG4Writer::Track::addChunkOffset(off64_t offset) { 1205 ++mNumStcoTableEntries; 1206 mChunkOffsets.push_back(offset); 1207} 1208 1209void MPEG4Writer::Track::setTimeScale() { 1210 LOGV("setTimeScale"); 1211 // Default time scale 1212 mTimeScale = 90000; 1213 1214 if (mIsAudio) { 1215 // Use the sampling rate as the default time scale for audio track. 1216 int32_t sampleRate; 1217 bool success = mMeta->findInt32(kKeySampleRate, &sampleRate); 1218 CHECK(success); 1219 mTimeScale = sampleRate; 1220 } 1221 1222 // If someone would like to overwrite the timescale, use user-supplied value. 1223 int32_t timeScale; 1224 if (mMeta->findInt32(kKeyTimeScale, &timeScale)) { 1225 mTimeScale = timeScale; 1226 } 1227 1228 CHECK(mTimeScale > 0); 1229} 1230 1231void MPEG4Writer::Track::getCodecSpecificDataFromInputFormatIfPossible() { 1232 const char *mime; 1233 CHECK(mMeta->findCString(kKeyMIMEType, &mime)); 1234 1235 if (!strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_AVC)) { 1236 uint32_t type; 1237 const void *data; 1238 size_t size; 1239 if (mMeta->findData(kKeyAVCC, &type, &data, &size)) { 1240 mCodecSpecificData = malloc(size); 1241 mCodecSpecificDataSize = size; 1242 memcpy(mCodecSpecificData, data, size); 1243 mGotAllCodecSpecificData = true; 1244 } 1245 } else if (!strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_MPEG4) 1246 || !strcasecmp(mime, MEDIA_MIMETYPE_AUDIO_AAC)) { 1247 uint32_t type; 1248 const void *data; 1249 size_t size; 1250 if (mMeta->findData(kKeyESDS, &type, &data, &size)) { 1251 ESDS esds(data, size); 1252 if (esds.getCodecSpecificInfo(&data, &size) == OK) { 1253 mCodecSpecificData = malloc(size); 1254 mCodecSpecificDataSize = size; 1255 memcpy(mCodecSpecificData, data, size); 1256 mGotAllCodecSpecificData = true; 1257 } 1258 } 1259 } 1260} 1261 1262MPEG4Writer::Track::~Track() { 1263 stop(); 1264 1265 if (mCodecSpecificData != NULL) { 1266 free(mCodecSpecificData); 1267 mCodecSpecificData = NULL; 1268 } 1269} 1270 1271void MPEG4Writer::Track::initTrackingProgressStatus(MetaData *params) { 1272 LOGV("initTrackingProgressStatus"); 1273 mPreviousTrackTimeUs = -1; 1274 mTrackingProgressStatus = false; 1275 mTrackEveryTimeDurationUs = 0; 1276 { 1277 int64_t timeUs; 1278 if (params && params->findInt64(kKeyTrackTimeStatus, &timeUs)) { 1279 LOGV("Receive request to track progress status for every %lld us", timeUs); 1280 mTrackEveryTimeDurationUs = timeUs; 1281 mTrackingProgressStatus = true; 1282 } 1283 } 1284} 1285 1286// static 1287void *MPEG4Writer::ThreadWrapper(void *me) { 1288 LOGV("ThreadWrapper: %p", me); 1289 MPEG4Writer *writer = static_cast<MPEG4Writer *>(me); 1290 writer->threadFunc(); 1291 return NULL; 1292} 1293 1294void MPEG4Writer::bufferChunk(const Chunk& chunk) { 1295 LOGV("bufferChunk: %p", chunk.mTrack); 1296 Mutex::Autolock autolock(mLock); 1297 CHECK_EQ(mDone, false); 1298 1299 for (List<ChunkInfo>::iterator it = mChunkInfos.begin(); 1300 it != mChunkInfos.end(); ++it) { 1301 1302 if (chunk.mTrack == it->mTrack) { // Found owner 1303 it->mChunks.push_back(chunk); 1304 mChunkReadyCondition.signal(); 1305 return; 1306 } 1307 } 1308 1309 CHECK("Received a chunk for a unknown track" == 0); 1310} 1311 1312void MPEG4Writer::writeChunkToFile(Chunk* chunk) { 1313 LOGV("writeChunkToFile: %lld from %s track", 1314 chunk.mTimestampUs, chunk.mTrack->isAudio()? "audio": "video"); 1315 1316 int32_t isFirstSample = true; 1317 while (!chunk->mSamples.empty()) { 1318 List<MediaBuffer *>::iterator it = chunk->mSamples.begin(); 1319 1320 off64_t offset = chunk->mTrack->isAvc() 1321 ? addLengthPrefixedSample_l(*it) 1322 : addSample_l(*it); 1323 1324 if (isFirstSample) { 1325 chunk->mTrack->addChunkOffset(offset); 1326 isFirstSample = false; 1327 } 1328 1329 (*it)->release(); 1330 (*it) = NULL; 1331 chunk->mSamples.erase(it); 1332 } 1333 chunk->mSamples.clear(); 1334} 1335 1336void MPEG4Writer::writeAllChunks() { 1337 LOGV("writeAllChunks"); 1338 size_t outstandingChunks = 0; 1339 Chunk chunk; 1340 while (findChunkToWrite(&chunk)) { 1341 writeChunkToFile(&chunk); 1342 ++outstandingChunks; 1343 } 1344 1345 sendSessionSummary(); 1346 1347 mChunkInfos.clear(); 1348 LOGD("%d chunks are written in the last batch", outstandingChunks); 1349} 1350 1351bool MPEG4Writer::findChunkToWrite(Chunk *chunk) { 1352 LOGV("findChunkToWrite"); 1353 1354 int64_t minTimestampUs = 0x7FFFFFFFFFFFFFFFLL; 1355 Track *track = NULL; 1356 for (List<ChunkInfo>::iterator it = mChunkInfos.begin(); 1357 it != mChunkInfos.end(); ++it) { 1358 if (!it->mChunks.empty()) { 1359 List<Chunk>::iterator chunkIt = it->mChunks.begin(); 1360 if (chunkIt->mTimeStampUs < minTimestampUs) { 1361 minTimestampUs = chunkIt->mTimeStampUs; 1362 track = it->mTrack; 1363 } 1364 } 1365 } 1366 1367 if (track == NULL) { 1368 LOGV("Nothing to be written after all"); 1369 return false; 1370 } 1371 1372 if (mIsFirstChunk) { 1373 mIsFirstChunk = false; 1374 } 1375 1376 for (List<ChunkInfo>::iterator it = mChunkInfos.begin(); 1377 it != mChunkInfos.end(); ++it) { 1378 if (it->mTrack == track) { 1379 *chunk = *(it->mChunks.begin()); 1380 it->mChunks.erase(it->mChunks.begin()); 1381 CHECK_EQ(chunk->mTrack, track); 1382 1383 int64_t interChunkTimeUs = 1384 chunk->mTimeStampUs - it->mPrevChunkTimestampUs; 1385 if (interChunkTimeUs > it->mPrevChunkTimestampUs) { 1386 it->mMaxInterChunkDurUs = interChunkTimeUs; 1387 } 1388 1389 return true; 1390 } 1391 } 1392 1393 return false; 1394} 1395 1396void MPEG4Writer::threadFunc() { 1397 LOGV("threadFunc"); 1398 1399 prctl(PR_SET_NAME, (unsigned long)"MPEG4Writer", 0, 0, 0); 1400 1401 Mutex::Autolock autoLock(mLock); 1402 while (!mDone) { 1403 Chunk chunk; 1404 bool chunkFound = false; 1405 1406 while (!mDone && !(chunkFound = findChunkToWrite(&chunk))) { 1407 mChunkReadyCondition.wait(mLock); 1408 } 1409 1410 // Actual write without holding the lock in order to 1411 // reduce the blocking time for media track threads. 1412 if (chunkFound) { 1413 mLock.unlock(); 1414 writeChunkToFile(&chunk); 1415 mLock.lock(); 1416 } 1417 } 1418 1419 writeAllChunks(); 1420} 1421 1422status_t MPEG4Writer::startWriterThread() { 1423 LOGV("startWriterThread"); 1424 1425 mDone = false; 1426 mIsFirstChunk = true; 1427 mDriftTimeUs = 0; 1428 for (List<Track *>::iterator it = mTracks.begin(); 1429 it != mTracks.end(); ++it) { 1430 ChunkInfo info; 1431 info.mTrack = *it; 1432 info.mPrevChunkTimestampUs = 0; 1433 info.mMaxInterChunkDurUs = 0; 1434 mChunkInfos.push_back(info); 1435 } 1436 1437 pthread_attr_t attr; 1438 pthread_attr_init(&attr); 1439 pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE); 1440 pthread_create(&mThread, &attr, ThreadWrapper, this); 1441 pthread_attr_destroy(&attr); 1442 return OK; 1443} 1444 1445 1446status_t MPEG4Writer::Track::start(MetaData *params) { 1447 if (!mDone && mPaused) { 1448 mPaused = false; 1449 mResumed = true; 1450 return OK; 1451 } 1452 1453 int64_t startTimeUs; 1454 if (params == NULL || !params->findInt64(kKeyTime, &startTimeUs)) { 1455 startTimeUs = 0; 1456 } 1457 mStartTimeRealUs = startTimeUs; 1458 1459 int32_t rotationDegrees; 1460 if (!mIsAudio && params && params->findInt32(kKeyRotation, &rotationDegrees)) { 1461 mRotation = rotationDegrees; 1462 } 1463 1464 mIsRealTimeRecording = true; 1465 { 1466 int32_t isNotRealTime; 1467 if (params && params->findInt32(kKeyNotRealTime, &isNotRealTime)) { 1468 mIsRealTimeRecording = (isNotRealTime == 0); 1469 } 1470 } 1471 1472 initTrackingProgressStatus(params); 1473 1474 sp<MetaData> meta = new MetaData; 1475 if (mIsRealTimeRecording && mOwner->numTracks() > 1) { 1476 /* 1477 * This extra delay of accepting incoming audio/video signals 1478 * helps to align a/v start time at the beginning of a recording 1479 * session, and it also helps eliminate the "recording" sound for 1480 * camcorder applications. 1481 * 1482 * If client does not set the start time offset, we fall back to 1483 * use the default initial delay value. 1484 */ 1485 int64_t startTimeOffsetUs = mOwner->getStartTimeOffsetMs() * 1000LL; 1486 if (startTimeOffsetUs < 0) { // Start time offset was not set 1487 startTimeOffsetUs = kInitialDelayTimeUs; 1488 } 1489 startTimeUs += startTimeOffsetUs; 1490 LOGI("Start time offset: %lld us", startTimeOffsetUs); 1491 } 1492 1493 meta->setInt64(kKeyTime, startTimeUs); 1494 1495 status_t err = mSource->start(meta.get()); 1496 if (err != OK) { 1497 mDone = mReachedEOS = true; 1498 return err; 1499 } 1500 1501 pthread_attr_t attr; 1502 pthread_attr_init(&attr); 1503 pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE); 1504 1505 mDone = false; 1506 mStarted = true; 1507 mTrackDurationUs = 0; 1508 mReachedEOS = false; 1509 mEstimatedTrackSizeBytes = 0; 1510 mNumStcoTableEntries = 0; 1511 mNumStssTableEntries = 0; 1512 mNumStscTableEntries = 0; 1513 mNumSttsTableEntries = 0; 1514 mNumCttsTableEntries = 0; 1515 mMdatSizeBytes = 0; 1516 mIsMediaTimeAdjustmentOn = false; 1517 mPrevMediaTimeAdjustTimestampUs = 0; 1518 mMediaTimeAdjustNumFrames = 0; 1519 mPrevMediaTimeAdjustSample = 0; 1520 mTotalDriftTimeToAdjustUs = 0; 1521 mPrevTotalAccumDriftTimeUs = 0; 1522 mMaxChunkDurationUs = 0; 1523 mHasNegativeCttsDeltaDuration = false; 1524 1525 pthread_create(&mThread, &attr, ThreadWrapper, this); 1526 pthread_attr_destroy(&attr); 1527 1528 return OK; 1529} 1530 1531status_t MPEG4Writer::Track::pause() { 1532 mPaused = true; 1533 return OK; 1534} 1535 1536status_t MPEG4Writer::Track::stop() { 1537 LOGD("Stopping %s track", mIsAudio? "Audio": "Video"); 1538 if (!mStarted) { 1539 LOGE("Stop() called but track is not started"); 1540 return ERROR_END_OF_STREAM; 1541 } 1542 1543 if (mDone) { 1544 return OK; 1545 } 1546 mDone = true; 1547 1548 void *dummy; 1549 pthread_join(mThread, &dummy); 1550 1551 status_t err = (status_t) dummy; 1552 1553 LOGD("Stopping %s track source", mIsAudio? "Audio": "Video"); 1554 { 1555 status_t status = mSource->stop(); 1556 if (err == OK && status != OK && status != ERROR_END_OF_STREAM) { 1557 err = status; 1558 } 1559 } 1560 1561 LOGD("%s track stopped", mIsAudio? "Audio": "Video"); 1562 return err; 1563} 1564 1565bool MPEG4Writer::Track::reachedEOS() { 1566 return mReachedEOS; 1567} 1568 1569// static 1570void *MPEG4Writer::Track::ThreadWrapper(void *me) { 1571 Track *track = static_cast<Track *>(me); 1572 1573 status_t err = track->threadEntry(); 1574 return (void *) err; 1575} 1576 1577static void getNalUnitType(uint8_t byte, uint8_t* type) { 1578 LOGV("getNalUnitType: %d", byte); 1579 1580 // nal_unit_type: 5-bit unsigned integer 1581 *type = (byte & 0x1F); 1582} 1583 1584static const uint8_t *findNextStartCode( 1585 const uint8_t *data, size_t length) { 1586 1587 LOGV("findNextStartCode: %p %d", data, length); 1588 1589 size_t bytesLeft = length; 1590 while (bytesLeft > 4 && 1591 memcmp("\x00\x00\x00\x01", &data[length - bytesLeft], 4)) { 1592 --bytesLeft; 1593 } 1594 if (bytesLeft <= 4) { 1595 bytesLeft = 0; // Last parameter set 1596 } 1597 return &data[length - bytesLeft]; 1598} 1599 1600const uint8_t *MPEG4Writer::Track::parseParamSet( 1601 const uint8_t *data, size_t length, int type, size_t *paramSetLen) { 1602 1603 LOGV("parseParamSet"); 1604 CHECK(type == kNalUnitTypeSeqParamSet || 1605 type == kNalUnitTypePicParamSet); 1606 1607 const uint8_t *nextStartCode = findNextStartCode(data, length); 1608 *paramSetLen = nextStartCode - data; 1609 if (*paramSetLen == 0) { 1610 LOGE("Param set is malformed, since its length is 0"); 1611 return NULL; 1612 } 1613 1614 AVCParamSet paramSet(*paramSetLen, data); 1615 if (type == kNalUnitTypeSeqParamSet) { 1616 if (*paramSetLen < 4) { 1617 LOGE("Seq parameter set malformed"); 1618 return NULL; 1619 } 1620 if (mSeqParamSets.empty()) { 1621 mProfileIdc = data[1]; 1622 mProfileCompatible = data[2]; 1623 mLevelIdc = data[3]; 1624 } else { 1625 if (mProfileIdc != data[1] || 1626 mProfileCompatible != data[2] || 1627 mLevelIdc != data[3]) { 1628 LOGE("Inconsistent profile/level found in seq parameter sets"); 1629 return NULL; 1630 } 1631 } 1632 mSeqParamSets.push_back(paramSet); 1633 } else { 1634 mPicParamSets.push_back(paramSet); 1635 } 1636 return nextStartCode; 1637} 1638 1639status_t MPEG4Writer::Track::copyAVCCodecSpecificData( 1640 const uint8_t *data, size_t size) { 1641 LOGV("copyAVCCodecSpecificData"); 1642 1643 // 2 bytes for each of the parameter set length field 1644 // plus the 7 bytes for the header 1645 if (size < 4 + 7) { 1646 LOGE("Codec specific data length too short: %d", size); 1647 return ERROR_MALFORMED; 1648 } 1649 1650 mCodecSpecificDataSize = size; 1651 mCodecSpecificData = malloc(size); 1652 memcpy(mCodecSpecificData, data, size); 1653 return OK; 1654} 1655 1656status_t MPEG4Writer::Track::parseAVCCodecSpecificData( 1657 const uint8_t *data, size_t size) { 1658 1659 LOGV("parseAVCCodecSpecificData"); 1660 // Data starts with a start code. 1661 // SPS and PPS are separated with start codes. 1662 // Also, SPS must come before PPS 1663 uint8_t type = kNalUnitTypeSeqParamSet; 1664 bool gotSps = false; 1665 bool gotPps = false; 1666 const uint8_t *tmp = data; 1667 const uint8_t *nextStartCode = data; 1668 size_t bytesLeft = size; 1669 size_t paramSetLen = 0; 1670 mCodecSpecificDataSize = 0; 1671 while (bytesLeft > 4 && !memcmp("\x00\x00\x00\x01", tmp, 4)) { 1672 getNalUnitType(*(tmp + 4), &type); 1673 if (type == kNalUnitTypeSeqParamSet) { 1674 if (gotPps) { 1675 LOGE("SPS must come before PPS"); 1676 return ERROR_MALFORMED; 1677 } 1678 if (!gotSps) { 1679 gotSps = true; 1680 } 1681 nextStartCode = parseParamSet(tmp + 4, bytesLeft - 4, type, ¶mSetLen); 1682 } else if (type == kNalUnitTypePicParamSet) { 1683 if (!gotSps) { 1684 LOGE("SPS must come before PPS"); 1685 return ERROR_MALFORMED; 1686 } 1687 if (!gotPps) { 1688 gotPps = true; 1689 } 1690 nextStartCode = parseParamSet(tmp + 4, bytesLeft - 4, type, ¶mSetLen); 1691 } else { 1692 LOGE("Only SPS and PPS Nal units are expected"); 1693 return ERROR_MALFORMED; 1694 } 1695 1696 if (nextStartCode == NULL) { 1697 return ERROR_MALFORMED; 1698 } 1699 1700 // Move on to find the next parameter set 1701 bytesLeft -= nextStartCode - tmp; 1702 tmp = nextStartCode; 1703 mCodecSpecificDataSize += (2 + paramSetLen); 1704 } 1705 1706 { 1707 // Check on the number of seq parameter sets 1708 size_t nSeqParamSets = mSeqParamSets.size(); 1709 if (nSeqParamSets == 0) { 1710 LOGE("Cound not find sequence parameter set"); 1711 return ERROR_MALFORMED; 1712 } 1713 1714 if (nSeqParamSets > 0x1F) { 1715 LOGE("Too many seq parameter sets (%d) found", nSeqParamSets); 1716 return ERROR_MALFORMED; 1717 } 1718 } 1719 1720 { 1721 // Check on the number of pic parameter sets 1722 size_t nPicParamSets = mPicParamSets.size(); 1723 if (nPicParamSets == 0) { 1724 LOGE("Cound not find picture parameter set"); 1725 return ERROR_MALFORMED; 1726 } 1727 if (nPicParamSets > 0xFF) { 1728 LOGE("Too many pic parameter sets (%d) found", nPicParamSets); 1729 return ERROR_MALFORMED; 1730 } 1731 } 1732 1733 { 1734 // Check on the profiles 1735 // These profiles requires additional parameter set extensions 1736 if (mProfileIdc == 100 || mProfileIdc == 110 || 1737 mProfileIdc == 122 || mProfileIdc == 144) { 1738 LOGE("Sorry, no support for profile_idc: %d!", mProfileIdc); 1739 return BAD_VALUE; 1740 } 1741 } 1742 1743 return OK; 1744} 1745 1746status_t MPEG4Writer::Track::makeAVCCodecSpecificData( 1747 const uint8_t *data, size_t size) { 1748 1749 if (mCodecSpecificData != NULL) { 1750 LOGE("Already have codec specific data"); 1751 return ERROR_MALFORMED; 1752 } 1753 1754 if (size < 4) { 1755 LOGE("Codec specific data length too short: %d", size); 1756 return ERROR_MALFORMED; 1757 } 1758 1759 // Data is in the form of AVCCodecSpecificData 1760 if (memcmp("\x00\x00\x00\x01", data, 4)) { 1761 return copyAVCCodecSpecificData(data, size); 1762 } 1763 1764 if (parseAVCCodecSpecificData(data, size) != OK) { 1765 return ERROR_MALFORMED; 1766 } 1767 1768 // ISO 14496-15: AVC file format 1769 mCodecSpecificDataSize += 7; // 7 more bytes in the header 1770 mCodecSpecificData = malloc(mCodecSpecificDataSize); 1771 uint8_t *header = (uint8_t *)mCodecSpecificData; 1772 header[0] = 1; // version 1773 header[1] = mProfileIdc; // profile indication 1774 header[2] = mProfileCompatible; // profile compatibility 1775 header[3] = mLevelIdc; 1776 1777 // 6-bit '111111' followed by 2-bit to lengthSizeMinuusOne 1778 if (mOwner->useNalLengthFour()) { 1779 header[4] = 0xfc | 3; // length size == 4 bytes 1780 } else { 1781 header[4] = 0xfc | 1; // length size == 2 bytes 1782 } 1783 1784 // 3-bit '111' followed by 5-bit numSequenceParameterSets 1785 int nSequenceParamSets = mSeqParamSets.size(); 1786 header[5] = 0xe0 | nSequenceParamSets; 1787 header += 6; 1788 for (List<AVCParamSet>::iterator it = mSeqParamSets.begin(); 1789 it != mSeqParamSets.end(); ++it) { 1790 // 16-bit sequence parameter set length 1791 uint16_t seqParamSetLength = it->mLength; 1792 header[0] = seqParamSetLength >> 8; 1793 header[1] = seqParamSetLength & 0xff; 1794 1795 // SPS NAL unit (sequence parameter length bytes) 1796 memcpy(&header[2], it->mData, seqParamSetLength); 1797 header += (2 + seqParamSetLength); 1798 } 1799 1800 // 8-bit nPictureParameterSets 1801 int nPictureParamSets = mPicParamSets.size(); 1802 header[0] = nPictureParamSets; 1803 header += 1; 1804 for (List<AVCParamSet>::iterator it = mPicParamSets.begin(); 1805 it != mPicParamSets.end(); ++it) { 1806 // 16-bit picture parameter set length 1807 uint16_t picParamSetLength = it->mLength; 1808 header[0] = picParamSetLength >> 8; 1809 header[1] = picParamSetLength & 0xff; 1810 1811 // PPS Nal unit (picture parameter set length bytes) 1812 memcpy(&header[2], it->mData, picParamSetLength); 1813 header += (2 + picParamSetLength); 1814 } 1815 1816 return OK; 1817} 1818 1819/* 1820* The video track's media time adjustment for real-time applications 1821* is described as follows: 1822* 1823* First, the media time adjustment is done for every period of 1824* kVideoMediaTimeAdjustPeriodTimeUs. kVideoMediaTimeAdjustPeriodTimeUs 1825* is currently a fixed value chosen heuristically. The value of 1826* kVideoMediaTimeAdjustPeriodTimeUs should not be very large or very small 1827* for two considerations: on one hand, a relatively large value 1828* helps reduce large fluctuation of drift time in the audio encoding 1829* path; while on the other hand, a relatively small value helps keep 1830* restoring synchronization in audio/video more frequently. Note for the 1831* very first period of kVideoMediaTimeAdjustPeriodTimeUs, there is 1832* no media time adjustment for the video track. 1833* 1834* Second, the total accumulated audio track time drift found 1835* in a period of kVideoMediaTimeAdjustPeriodTimeUs is distributed 1836* over a stream of incoming video frames. The number of video frames 1837* affected is determined based on the number of recorded video frames 1838* within the past kVideoMediaTimeAdjustPeriodTimeUs period. 1839* We choose to distribute the drift time over only a portion 1840* (rather than all) of the total number of recorded video frames 1841* in order to make sure that the video track media time adjustment is 1842* completed for the current period before the next video track media 1843* time adjustment period starts. Currently, the portion chosen is a 1844* half (0.5). 1845* 1846* Last, various additional checks are performed to ensure that 1847* the actual audio encoding path does not have too much drift. 1848* In particular, 1) we want to limit the average incremental time 1849* adjustment for each video frame to be less than a threshold 1850* for a single period of kVideoMediaTimeAdjustPeriodTimeUs. 1851* Currently, the threshold is set to 5 ms. If the average incremental 1852* media time adjustment for a video frame is larger than the 1853* threshold, the audio encoding path has too much time drift. 1854* 2) We also want to limit the total time drift in the audio 1855* encoding path to be less than a threshold for a period of 1856* kVideoMediaTimeAdjustPeriodTimeUs. Currently, the threshold 1857* is 0.5% of kVideoMediaTimeAdjustPeriodTimeUs. If the time drift of 1858* the audio encoding path is larger than the threshold, the audio 1859* encoding path has too much time drift. We treat the large time 1860* drift of the audio encoding path as errors, since there is no 1861* way to keep audio/video in synchronization for real-time 1862* applications if the time drift is too large unless we drop some 1863* video frames, which has its own problems that we don't want 1864* to get into for the time being. 1865*/ 1866void MPEG4Writer::Track::adjustMediaTime(int64_t *timestampUs) { 1867 if (*timestampUs - mPrevMediaTimeAdjustTimestampUs >= 1868 kVideoMediaTimeAdjustPeriodTimeUs) { 1869 1870 LOGV("New media time adjustment period at %lld us", *timestampUs); 1871 mIsMediaTimeAdjustmentOn = true; 1872 mMediaTimeAdjustNumFrames = 1873 (mNumSamples - mPrevMediaTimeAdjustSample) >> 1; 1874 1875 mPrevMediaTimeAdjustTimestampUs = *timestampUs; 1876 mPrevMediaTimeAdjustSample = mNumSamples; 1877 int64_t totalAccumDriftTimeUs = mOwner->getDriftTimeUs(); 1878 mTotalDriftTimeToAdjustUs = 1879 totalAccumDriftTimeUs - mPrevTotalAccumDriftTimeUs; 1880 1881 mPrevTotalAccumDriftTimeUs = totalAccumDriftTimeUs; 1882 1883 // Check on incremental adjusted time per frame 1884 int64_t adjustTimePerFrameUs = 1885 mTotalDriftTimeToAdjustUs / mMediaTimeAdjustNumFrames; 1886 1887 if (adjustTimePerFrameUs < 0) { 1888 adjustTimePerFrameUs = -adjustTimePerFrameUs; 1889 } 1890 if (adjustTimePerFrameUs >= 5000) { 1891 LOGE("Adjusted time per video frame is %lld us", 1892 adjustTimePerFrameUs); 1893 CHECK(!"Video frame time adjustment is too large!"); 1894 } 1895 1896 // Check on total accumulated time drift within a period of 1897 // kVideoMediaTimeAdjustPeriodTimeUs. 1898 int64_t driftPercentage = (mTotalDriftTimeToAdjustUs * 1000) 1899 / kVideoMediaTimeAdjustPeriodTimeUs; 1900 1901 if (driftPercentage < 0) { 1902 driftPercentage = -driftPercentage; 1903 } 1904 if (driftPercentage > 5) { 1905 LOGE("Audio track has time drift %lld us over %lld us", 1906 mTotalDriftTimeToAdjustUs, 1907 kVideoMediaTimeAdjustPeriodTimeUs); 1908 1909 CHECK(!"The audio track media time drifts too much!"); 1910 } 1911 1912 } 1913 1914 if (mIsMediaTimeAdjustmentOn) { 1915 if (mNumSamples - mPrevMediaTimeAdjustSample <= 1916 mMediaTimeAdjustNumFrames) { 1917 1918 // Do media time incremental adjustment 1919 int64_t incrementalAdjustTimeUs = 1920 (mTotalDriftTimeToAdjustUs * 1921 (mNumSamples - mPrevMediaTimeAdjustSample)) 1922 / mMediaTimeAdjustNumFrames; 1923 1924 *timestampUs += 1925 (incrementalAdjustTimeUs + mPrevTotalAccumDriftTimeUs); 1926 1927 LOGV("Incremental video frame media time adjustment: %lld us", 1928 (incrementalAdjustTimeUs + mPrevTotalAccumDriftTimeUs)); 1929 } else { 1930 // Within the remaining adjustment period, 1931 // no incremental adjustment is needed. 1932 *timestampUs += 1933 (mTotalDriftTimeToAdjustUs + mPrevTotalAccumDriftTimeUs); 1934 1935 LOGV("Fixed video frame media time adjustment: %lld us", 1936 (mTotalDriftTimeToAdjustUs + mPrevTotalAccumDriftTimeUs)); 1937 } 1938 } 1939} 1940 1941/* 1942 * Updates the drift time from the audio track so that 1943 * the video track can get the updated drift time information 1944 * from the file writer. The fluctuation of the drift time of the audio 1945 * encoding path is smoothed out with a simple filter by giving a larger 1946 * weight to more recently drift time. The filter coefficients, 0.5 and 0.5, 1947 * are heuristically determined. 1948 */ 1949void MPEG4Writer::Track::updateDriftTime(const sp<MetaData>& meta) { 1950 int64_t driftTimeUs = 0; 1951 if (meta->findInt64(kKeyDriftTime, &driftTimeUs)) { 1952 int64_t prevDriftTimeUs = mOwner->getDriftTimeUs(); 1953 int64_t timeUs = (driftTimeUs + prevDriftTimeUs) >> 1; 1954 mOwner->setDriftTimeUs(timeUs); 1955 } 1956} 1957 1958status_t MPEG4Writer::Track::threadEntry() { 1959 int32_t count = 0; 1960 const int64_t interleaveDurationUs = mOwner->interleaveDuration(); 1961 const bool hasMultipleTracks = (mOwner->numTracks() > 1); 1962 int64_t chunkTimestampUs = 0; 1963 int32_t nChunks = 0; 1964 int32_t nZeroLengthFrames = 0; 1965 int64_t lastTimestampUs = 0; // Previous sample time stamp 1966 int64_t lastCttsTimeUs = 0; // Previous sample time stamp 1967 int64_t lastDurationUs = 0; // Between the previous two samples 1968 int64_t currDurationTicks = 0; // Timescale based ticks 1969 int64_t lastDurationTicks = 0; // Timescale based ticks 1970 int32_t sampleCount = 1; // Sample count in the current stts table entry 1971 int64_t currCttsDurTicks = 0; // Timescale based ticks 1972 int64_t lastCttsDurTicks = 0; // Timescale based ticks 1973 int32_t cttsSampleCount = 1; // Sample count in the current ctts table entry 1974 uint32_t previousSampleSize = 0; // Size of the previous sample 1975 int64_t previousPausedDurationUs = 0; 1976 int64_t timestampUs = 0; 1977 int64_t cttsDeltaTimeUs = 0; 1978 bool hasBFrames = false; 1979 1980#if 1 1981 // XXX: Samsung's video encoder's output buffer timestamp 1982 // is not correct. see bug 4724339 1983 char value[PROPERTY_VALUE_MAX]; 1984 if (property_get("rw.media.record.hasb", value, NULL) && 1985 (!strcasecmp(value, "true") || !strcasecmp(value, "1"))) { 1986 hasBFrames = true; 1987 } 1988#endif 1989 if (mIsAudio) { 1990 prctl(PR_SET_NAME, (unsigned long)"AudioTrackEncoding", 0, 0, 0); 1991 } else { 1992 prctl(PR_SET_NAME, (unsigned long)"VideoTrackEncoding", 0, 0, 0); 1993 } 1994 androidSetThreadPriority(0, ANDROID_PRIORITY_AUDIO); 1995 1996 sp<MetaData> meta_data; 1997 1998 mNumSamples = 0; 1999 status_t err = OK; 2000 MediaBuffer *buffer; 2001 while (!mDone && (err = mSource->read(&buffer)) == OK) { 2002 if (buffer->range_length() == 0) { 2003 buffer->release(); 2004 buffer = NULL; 2005 ++nZeroLengthFrames; 2006 continue; 2007 } 2008 2009 // If the codec specific data has not been received yet, delay pause. 2010 // After the codec specific data is received, discard what we received 2011 // when the track is to be paused. 2012 if (mPaused && !mResumed) { 2013 buffer->release(); 2014 buffer = NULL; 2015 continue; 2016 } 2017 2018 ++count; 2019 2020 int32_t isCodecConfig; 2021 if (buffer->meta_data()->findInt32(kKeyIsCodecConfig, &isCodecConfig) 2022 && isCodecConfig) { 2023 CHECK(!mGotAllCodecSpecificData); 2024 2025 if (mIsAvc) { 2026 status_t err = makeAVCCodecSpecificData( 2027 (const uint8_t *)buffer->data() 2028 + buffer->range_offset(), 2029 buffer->range_length()); 2030 CHECK_EQ(OK, err); 2031 } else if (mIsMPEG4) { 2032 mCodecSpecificDataSize = buffer->range_length(); 2033 mCodecSpecificData = malloc(mCodecSpecificDataSize); 2034 memcpy(mCodecSpecificData, 2035 (const uint8_t *)buffer->data() 2036 + buffer->range_offset(), 2037 buffer->range_length()); 2038 } 2039 2040 buffer->release(); 2041 buffer = NULL; 2042 2043 mGotAllCodecSpecificData = true; 2044 continue; 2045 } 2046 2047 // Make a deep copy of the MediaBuffer and Metadata and release 2048 // the original as soon as we can 2049 MediaBuffer *copy = new MediaBuffer(buffer->range_length()); 2050 memcpy(copy->data(), (uint8_t *)buffer->data() + buffer->range_offset(), 2051 buffer->range_length()); 2052 copy->set_range(0, buffer->range_length()); 2053 meta_data = new MetaData(*buffer->meta_data().get()); 2054 buffer->release(); 2055 buffer = NULL; 2056 2057 if (mIsAvc) StripStartcode(copy); 2058 2059 size_t sampleSize = copy->range_length(); 2060 if (mIsAvc) { 2061 if (mOwner->useNalLengthFour()) { 2062 sampleSize += 4; 2063 } else { 2064 sampleSize += 2; 2065 } 2066 } 2067 2068 // Max file size or duration handling 2069 mMdatSizeBytes += sampleSize; 2070 updateTrackSizeEstimate(); 2071 2072 if (mOwner->exceedsFileSizeLimit()) { 2073 mOwner->notify(MEDIA_RECORDER_EVENT_INFO, MEDIA_RECORDER_INFO_MAX_FILESIZE_REACHED, 0); 2074 break; 2075 } 2076 if (mOwner->exceedsFileDurationLimit()) { 2077 mOwner->notify(MEDIA_RECORDER_EVENT_INFO, MEDIA_RECORDER_INFO_MAX_DURATION_REACHED, 0); 2078 break; 2079 } 2080 2081 2082 int32_t isSync = false; 2083 meta_data->findInt32(kKeyIsSyncFrame, &isSync); 2084 2085 /* 2086 * The original timestamp found in the data buffer will be modified as below: 2087 * 2088 * There is a playback offset into this track if the track's start time 2089 * is not the same as the movie start time, which will be recorded in edst 2090 * box of the output file. The playback offset is to make sure that the 2091 * starting time of the audio/video tracks are synchronized. Although the 2092 * track's media timestamp may be subject to various modifications 2093 * as outlined below, the track's playback offset time remains unchanged 2094 * once the first data buffer of the track is received. 2095 * 2096 * The media time stamp will be calculated by subtracting the playback offset 2097 * (and potential pause durations) from the original timestamp in the buffer. 2098 * 2099 * If this track is a video track for a real-time recording application with 2100 * both audio and video tracks, its media timestamp will subject to further 2101 * modification based on the media clock of the audio track. This modification 2102 * is needed for the purpose of maintaining good audio/video synchronization. 2103 * 2104 * If the recording session is paused and resumed multiple times, the track 2105 * media timestamp will be modified as if the recording session had never been 2106 * paused at all during playback of the recorded output file. In other words, 2107 * the output file will have no memory of pause/resume durations. 2108 * 2109 */ 2110 CHECK(meta_data->findInt64(kKeyTime, ×tampUs)); 2111 2112//////////////////////////////////////////////////////////////////////////////// 2113 if (mNumSamples == 0) { 2114 mFirstSampleTimeRealUs = systemTime() / 1000; 2115 mStartTimestampUs = timestampUs; 2116 mOwner->setStartTimestampUs(mStartTimestampUs); 2117 previousPausedDurationUs = mStartTimestampUs; 2118 } 2119 2120 if (mResumed) { 2121 int64_t durExcludingEarlierPausesUs = timestampUs - previousPausedDurationUs; 2122 CHECK(durExcludingEarlierPausesUs >= 0); 2123 int64_t pausedDurationUs = durExcludingEarlierPausesUs - mTrackDurationUs; 2124 CHECK(pausedDurationUs >= lastDurationUs); 2125 previousPausedDurationUs += pausedDurationUs - lastDurationUs; 2126 mResumed = false; 2127 } 2128 2129 timestampUs -= previousPausedDurationUs; 2130 CHECK(timestampUs >= 0); 2131 if (!mIsAudio && hasBFrames) { 2132 /* 2133 * Composition time: timestampUs 2134 * Decoding time: decodingTimeUs 2135 * Composition time delta = composition time - decoding time 2136 * 2137 * We save picture decoding time stamp delta in stts table entries, 2138 * and composition time delta duration in ctts table entries. 2139 */ 2140 int64_t decodingTimeUs; 2141 CHECK(meta_data->findInt64(kKeyDecodingTime, &decodingTimeUs)); 2142 decodingTimeUs -= previousPausedDurationUs; 2143 int64_t timeUs = decodingTimeUs; 2144 cttsDeltaTimeUs = timestampUs - decodingTimeUs; 2145 timestampUs = decodingTimeUs; 2146 LOGV("decoding time: %lld and ctts delta time: %lld", 2147 timestampUs, cttsDeltaTimeUs); 2148 } 2149 2150 // Media time adjustment for real-time applications 2151 if (mIsRealTimeRecording) { 2152 if (mIsAudio) { 2153 updateDriftTime(meta_data); 2154 } else { 2155 adjustMediaTime(×tampUs); 2156 } 2157 } 2158 2159 CHECK(timestampUs >= 0); 2160 if (mNumSamples > 1) { 2161 if (timestampUs <= lastTimestampUs) { 2162 LOGW("Frame arrives too late!"); 2163 // Don't drop the late frame, since dropping a frame may cause 2164 // problems later during playback 2165 2166 // The idea here is to avoid having two or more samples with the 2167 // same timestamp in the output file. 2168 if (mTimeScale >= 1000000LL) { 2169 timestampUs = lastTimestampUs + 1; 2170 } else { 2171 timestampUs = lastTimestampUs + (1000000LL + (mTimeScale >> 1)) / mTimeScale; 2172 } 2173 } 2174 } 2175 2176 LOGV("%s media time stamp: %lld and previous paused duration %lld", 2177 mIsAudio? "Audio": "Video", timestampUs, previousPausedDurationUs); 2178 if (timestampUs > mTrackDurationUs) { 2179 mTrackDurationUs = timestampUs; 2180 } 2181 2182 mSampleSizes.push_back(sampleSize); 2183 ++mNumSamples; 2184 if (mNumSamples > 2) { 2185 // We need to use the time scale based ticks, rather than the 2186 // timestamp itself to determine whether we have to use a new 2187 // stts entry, since we may have rounding errors. 2188 // The calculation is intended to reduce the accumulated 2189 // rounding errors. 2190 currDurationTicks = 2191 ((timestampUs * mTimeScale + 500000LL) / 1000000LL - 2192 (lastTimestampUs * mTimeScale + 500000LL) / 1000000LL); 2193 2194 // Force the first sample to have its own stts entry so that 2195 // we can adjust its value later to maintain the A/V sync. 2196 if (mNumSamples == 3 || currDurationTicks != lastDurationTicks) { 2197 LOGV("%s lastDurationUs: %lld us, currDurationTicks: %lld us", 2198 mIsAudio? "Audio": "Video", lastDurationUs, currDurationTicks); 2199 addOneSttsTableEntry(sampleCount, lastDurationTicks); 2200 sampleCount = 1; 2201 } else { 2202 ++sampleCount; 2203 } 2204 2205 if (!mIsAudio) { 2206 currCttsDurTicks = 2207 ((cttsDeltaTimeUs * mTimeScale + 500000LL) / 1000000LL - 2208 (lastCttsTimeUs * mTimeScale + 500000LL) / 1000000LL); 2209 if (currCttsDurTicks != lastCttsDurTicks) { 2210 addOneCttsTableEntry(cttsSampleCount, lastCttsDurTicks); 2211 cttsSampleCount = 1; 2212 } else { 2213 ++cttsSampleCount; 2214 } 2215 } 2216 } 2217 if (mSamplesHaveSameSize) { 2218 if (mNumSamples >= 2 && previousSampleSize != sampleSize) { 2219 mSamplesHaveSameSize = false; 2220 } 2221 previousSampleSize = sampleSize; 2222 } 2223 LOGV("%s timestampUs/lastTimestampUs: %lld/%lld", 2224 mIsAudio? "Audio": "Video", timestampUs, lastTimestampUs); 2225 lastDurationUs = timestampUs - lastTimestampUs; 2226 lastDurationTicks = currDurationTicks; 2227 lastTimestampUs = timestampUs; 2228 2229 if (!mIsAudio) { 2230 lastCttsDurTicks = currCttsDurTicks; 2231 lastCttsTimeUs = cttsDeltaTimeUs; 2232 } 2233 2234 if (isSync != 0) { 2235 addOneStssTableEntry(mNumSamples); 2236 } 2237 2238 if (mTrackingProgressStatus) { 2239 if (mPreviousTrackTimeUs <= 0) { 2240 mPreviousTrackTimeUs = mStartTimestampUs; 2241 } 2242 trackProgressStatus(timestampUs); 2243 } 2244 if (!hasMultipleTracks) { 2245 off64_t offset = mIsAvc? mOwner->addLengthPrefixedSample_l(copy) 2246 : mOwner->addSample_l(copy); 2247 if (mChunkOffsets.empty()) { 2248 addChunkOffset(offset); 2249 } 2250 copy->release(); 2251 copy = NULL; 2252 continue; 2253 } 2254 2255 mChunkSamples.push_back(copy); 2256 if (interleaveDurationUs == 0) { 2257 addOneStscTableEntry(++nChunks, 1); 2258 bufferChunk(timestampUs); 2259 } else { 2260 if (chunkTimestampUs == 0) { 2261 chunkTimestampUs = timestampUs; 2262 } else { 2263 int64_t chunkDurationUs = timestampUs - chunkTimestampUs; 2264 if (chunkDurationUs > interleaveDurationUs) { 2265 if (chunkDurationUs > mMaxChunkDurationUs) { 2266 mMaxChunkDurationUs = chunkDurationUs; 2267 } 2268 ++nChunks; 2269 if (nChunks == 1 || // First chunk 2270 (--(mStscTableEntries.end()))->samplesPerChunk != 2271 mChunkSamples.size()) { 2272 addOneStscTableEntry(nChunks, mChunkSamples.size()); 2273 } 2274 bufferChunk(timestampUs); 2275 chunkTimestampUs = timestampUs; 2276 } 2277 } 2278 } 2279 2280 } 2281 2282 if (mSampleSizes.empty() || // no samples written 2283 (!mIsAudio && mNumStssTableEntries == 0) || // no sync frames for video 2284 (OK != checkCodecSpecificData())) { // no codec specific data 2285 err = ERROR_MALFORMED; 2286 } 2287 mOwner->trackProgressStatus(mTrackId, -1, err); 2288 2289 // Last chunk 2290 if (!hasMultipleTracks) { 2291 addOneStscTableEntry(1, mNumSamples); 2292 } else if (!mChunkSamples.empty()) { 2293 addOneStscTableEntry(++nChunks, mChunkSamples.size()); 2294 bufferChunk(timestampUs); 2295 } 2296 2297 // We don't really know how long the last frame lasts, since 2298 // there is no frame time after it, just repeat the previous 2299 // frame's duration. 2300 if (mNumSamples == 1) { 2301 lastDurationUs = 0; // A single sample's duration 2302 lastDurationTicks = 0; 2303 lastCttsDurTicks = 0; 2304 } else { 2305 ++sampleCount; // Count for the last sample 2306 ++cttsSampleCount; 2307 } 2308 2309 if (mNumSamples <= 2) { 2310 addOneSttsTableEntry(1, lastDurationTicks); 2311 if (sampleCount - 1 > 0) { 2312 addOneSttsTableEntry(sampleCount - 1, lastDurationTicks); 2313 } 2314 } else { 2315 addOneSttsTableEntry(sampleCount, lastDurationTicks); 2316 } 2317 2318 addOneCttsTableEntry(cttsSampleCount, lastCttsDurTicks); 2319 mTrackDurationUs += lastDurationUs; 2320 mReachedEOS = true; 2321 2322 sendTrackSummary(hasMultipleTracks); 2323 2324 LOGI("Received total/0-length (%d/%d) buffers and encoded %d frames. - %s", 2325 count, nZeroLengthFrames, mNumSamples, mIsAudio? "audio": "video"); 2326 if (mIsAudio) { 2327 LOGI("Audio track drift time: %lld us", mOwner->getDriftTimeUs()); 2328 } 2329 2330 if (err == ERROR_END_OF_STREAM) { 2331 return OK; 2332 } 2333 return err; 2334} 2335 2336void MPEG4Writer::Track::sendTrackSummary(bool hasMultipleTracks) { 2337 2338 // Send track summary only if test mode is enabled. 2339 if (!isTestModeEnabled()) { 2340 return; 2341 } 2342 2343 int trackNum = (mTrackId << 28); 2344 2345 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2346 trackNum | MEDIA_RECORDER_TRACK_INFO_TYPE, 2347 mIsAudio? 0: 1); 2348 2349 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2350 trackNum | MEDIA_RECORDER_TRACK_INFO_DURATION_MS, 2351 mTrackDurationUs / 1000); 2352 2353 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2354 trackNum | MEDIA_RECORDER_TRACK_INFO_ENCODED_FRAMES, 2355 mNumSamples); 2356 2357 { 2358 // The system delay time excluding the requested initial delay that 2359 // is used to eliminate the recording sound. 2360 int64_t startTimeOffsetUs = mOwner->getStartTimeOffsetMs() * 1000LL; 2361 if (startTimeOffsetUs < 0) { // Start time offset was not set 2362 startTimeOffsetUs = kInitialDelayTimeUs; 2363 } 2364 int64_t initialDelayUs = 2365 mFirstSampleTimeRealUs - mStartTimeRealUs - startTimeOffsetUs; 2366 2367 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2368 trackNum | MEDIA_RECORDER_TRACK_INFO_INITIAL_DELAY_MS, 2369 (initialDelayUs) / 1000); 2370 } 2371 2372 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2373 trackNum | MEDIA_RECORDER_TRACK_INFO_DATA_KBYTES, 2374 mMdatSizeBytes / 1024); 2375 2376 if (hasMultipleTracks) { 2377 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2378 trackNum | MEDIA_RECORDER_TRACK_INFO_MAX_CHUNK_DUR_MS, 2379 mMaxChunkDurationUs / 1000); 2380 2381 int64_t moovStartTimeUs = mOwner->getStartTimestampUs(); 2382 if (mStartTimestampUs != moovStartTimeUs) { 2383 int64_t startTimeOffsetUs = mStartTimestampUs - moovStartTimeUs; 2384 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2385 trackNum | MEDIA_RECORDER_TRACK_INFO_START_OFFSET_MS, 2386 startTimeOffsetUs / 1000); 2387 } 2388 } 2389} 2390 2391void MPEG4Writer::Track::trackProgressStatus(int64_t timeUs, status_t err) { 2392 LOGV("trackProgressStatus: %lld us", timeUs); 2393 if (mTrackEveryTimeDurationUs > 0 && 2394 timeUs - mPreviousTrackTimeUs >= mTrackEveryTimeDurationUs) { 2395 LOGV("Fire time tracking progress status at %lld us", timeUs); 2396 mOwner->trackProgressStatus(mTrackId, timeUs - mPreviousTrackTimeUs, err); 2397 mPreviousTrackTimeUs = timeUs; 2398 } 2399} 2400 2401void MPEG4Writer::trackProgressStatus( 2402 size_t trackId, int64_t timeUs, status_t err) { 2403 Mutex::Autolock lock(mLock); 2404 int32_t trackNum = (trackId << 28); 2405 2406 // Error notification 2407 // Do not consider ERROR_END_OF_STREAM an error 2408 if (err != OK && err != ERROR_END_OF_STREAM) { 2409 notify(MEDIA_RECORDER_TRACK_EVENT_ERROR, 2410 trackNum | MEDIA_RECORDER_TRACK_ERROR_GENERAL, 2411 err); 2412 return; 2413 } 2414 2415 if (timeUs == -1) { 2416 // Send completion notification 2417 notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2418 trackNum | MEDIA_RECORDER_TRACK_INFO_COMPLETION_STATUS, 2419 err); 2420 } else { 2421 // Send progress status 2422 notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2423 trackNum | MEDIA_RECORDER_TRACK_INFO_PROGRESS_IN_TIME, 2424 timeUs / 1000); 2425 } 2426} 2427 2428void MPEG4Writer::setDriftTimeUs(int64_t driftTimeUs) { 2429 LOGV("setDriftTimeUs: %lld us", driftTimeUs); 2430 Mutex::Autolock autolock(mLock); 2431 mDriftTimeUs = driftTimeUs; 2432} 2433 2434int64_t MPEG4Writer::getDriftTimeUs() { 2435 LOGV("getDriftTimeUs: %lld us", mDriftTimeUs); 2436 Mutex::Autolock autolock(mLock); 2437 return mDriftTimeUs; 2438} 2439 2440bool MPEG4Writer::useNalLengthFour() { 2441 return mUse4ByteNalLength; 2442} 2443 2444void MPEG4Writer::Track::bufferChunk(int64_t timestampUs) { 2445 LOGV("bufferChunk"); 2446 2447 Chunk chunk(this, timestampUs, mChunkSamples); 2448 mOwner->bufferChunk(chunk); 2449 mChunkSamples.clear(); 2450} 2451 2452int64_t MPEG4Writer::Track::getDurationUs() const { 2453 return mTrackDurationUs; 2454} 2455 2456int64_t MPEG4Writer::Track::getEstimatedTrackSizeBytes() const { 2457 return mEstimatedTrackSizeBytes; 2458} 2459 2460status_t MPEG4Writer::Track::checkCodecSpecificData() const { 2461 const char *mime; 2462 CHECK(mMeta->findCString(kKeyMIMEType, &mime)); 2463 if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AAC, mime) || 2464 !strcasecmp(MEDIA_MIMETYPE_VIDEO_MPEG4, mime) || 2465 !strcasecmp(MEDIA_MIMETYPE_VIDEO_AVC, mime)) { 2466 if (!mCodecSpecificData || 2467 mCodecSpecificDataSize <= 0) { 2468 LOGE("Missing codec specific data"); 2469 return ERROR_MALFORMED; 2470 } 2471 } else { 2472 if (mCodecSpecificData || 2473 mCodecSpecificDataSize > 0) { 2474 LOGE("Unexepected codec specific data found"); 2475 return ERROR_MALFORMED; 2476 } 2477 } 2478 return OK; 2479} 2480 2481void MPEG4Writer::Track::writeTrackHeader(bool use32BitOffset) { 2482 2483 LOGV("%s track time scale: %d", 2484 mIsAudio? "Audio": "Video", mTimeScale); 2485 2486 time_t now = time(NULL); 2487 mOwner->beginBox("trak"); 2488 writeTkhdBox(now); 2489 mOwner->beginBox("mdia"); 2490 writeMdhdBox(now); 2491 writeHdlrBox(); 2492 mOwner->beginBox("minf"); 2493 if (mIsAudio) { 2494 writeSmhdBox(); 2495 } else { 2496 writeVmhdBox(); 2497 } 2498 writeDinfBox(); 2499 writeStblBox(use32BitOffset); 2500 mOwner->endBox(); // minf 2501 mOwner->endBox(); // mdia 2502 mOwner->endBox(); // trak 2503} 2504 2505void MPEG4Writer::Track::writeStblBox(bool use32BitOffset) { 2506 mOwner->beginBox("stbl"); 2507 mOwner->beginBox("stsd"); 2508 mOwner->writeInt32(0); // version=0, flags=0 2509 mOwner->writeInt32(1); // entry count 2510 if (mIsAudio) { 2511 writeAudioFourCCBox(); 2512 } else { 2513 writeVideoFourCCBox(); 2514 } 2515 mOwner->endBox(); // stsd 2516 writeSttsBox(); 2517 writeCttsBox(); 2518 if (!mIsAudio) { 2519 writeStssBox(); 2520 } 2521 writeStszBox(); 2522 writeStscBox(); 2523 writeStcoBox(use32BitOffset); 2524 mOwner->endBox(); // stbl 2525} 2526 2527void MPEG4Writer::Track::writeVideoFourCCBox() { 2528 const char *mime; 2529 bool success = mMeta->findCString(kKeyMIMEType, &mime); 2530 CHECK(success); 2531 if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_MPEG4, mime)) { 2532 mOwner->beginBox("mp4v"); 2533 } else if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_H263, mime)) { 2534 mOwner->beginBox("s263"); 2535 } else if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_AVC, mime)) { 2536 mOwner->beginBox("avc1"); 2537 } else { 2538 LOGE("Unknown mime type '%s'.", mime); 2539 CHECK(!"should not be here, unknown mime type."); 2540 } 2541 2542 mOwner->writeInt32(0); // reserved 2543 mOwner->writeInt16(0); // reserved 2544 mOwner->writeInt16(1); // data ref index 2545 mOwner->writeInt16(0); // predefined 2546 mOwner->writeInt16(0); // reserved 2547 mOwner->writeInt32(0); // predefined 2548 mOwner->writeInt32(0); // predefined 2549 mOwner->writeInt32(0); // predefined 2550 2551 int32_t width, height; 2552 success = mMeta->findInt32(kKeyWidth, &width); 2553 success = success && mMeta->findInt32(kKeyHeight, &height); 2554 CHECK(success); 2555 2556 mOwner->writeInt16(width); 2557 mOwner->writeInt16(height); 2558 mOwner->writeInt32(0x480000); // horiz resolution 2559 mOwner->writeInt32(0x480000); // vert resolution 2560 mOwner->writeInt32(0); // reserved 2561 mOwner->writeInt16(1); // frame count 2562 mOwner->write(" ", 32); 2563 mOwner->writeInt16(0x18); // depth 2564 mOwner->writeInt16(-1); // predefined 2565 2566 CHECK(23 + mCodecSpecificDataSize < 128); 2567 2568 if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_MPEG4, mime)) { 2569 writeMp4vEsdsBox(); 2570 } else if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_H263, mime)) { 2571 writeD263Box(); 2572 } else if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_AVC, mime)) { 2573 writeAvccBox(); 2574 } 2575 2576 writePaspBox(); 2577 mOwner->endBox(); // mp4v, s263 or avc1 2578} 2579 2580void MPEG4Writer::Track::writeAudioFourCCBox() { 2581 const char *mime; 2582 bool success = mMeta->findCString(kKeyMIMEType, &mime); 2583 CHECK(success); 2584 const char *fourcc = NULL; 2585 if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_NB, mime)) { 2586 fourcc = "samr"; 2587 } else if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_WB, mime)) { 2588 fourcc = "sawb"; 2589 } else if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AAC, mime)) { 2590 fourcc = "mp4a"; 2591 } else { 2592 LOGE("Unknown mime type '%s'.", mime); 2593 CHECK(!"should not be here, unknown mime type."); 2594 } 2595 2596 mOwner->beginBox(fourcc); // audio format 2597 mOwner->writeInt32(0); // reserved 2598 mOwner->writeInt16(0); // reserved 2599 mOwner->writeInt16(0x1); // data ref index 2600 mOwner->writeInt32(0); // reserved 2601 mOwner->writeInt32(0); // reserved 2602 int32_t nChannels; 2603 CHECK_EQ(true, mMeta->findInt32(kKeyChannelCount, &nChannels)); 2604 mOwner->writeInt16(nChannels); // channel count 2605 mOwner->writeInt16(16); // sample size 2606 mOwner->writeInt16(0); // predefined 2607 mOwner->writeInt16(0); // reserved 2608 2609 int32_t samplerate; 2610 success = mMeta->findInt32(kKeySampleRate, &samplerate); 2611 CHECK(success); 2612 mOwner->writeInt32(samplerate << 16); 2613 if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AAC, mime)) { 2614 writeMp4aEsdsBox(); 2615 } else if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_NB, mime) || 2616 !strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_WB, mime)) { 2617 writeDamrBox(); 2618 } 2619 mOwner->endBox(); 2620} 2621 2622void MPEG4Writer::Track::writeMp4aEsdsBox() { 2623 mOwner->beginBox("esds"); 2624 CHECK(mCodecSpecificData); 2625 CHECK(mCodecSpecificDataSize > 0); 2626 2627 // Make sure all sizes encode to a single byte. 2628 CHECK(mCodecSpecificDataSize + 23 < 128); 2629 2630 mOwner->writeInt32(0); // version=0, flags=0 2631 mOwner->writeInt8(0x03); // ES_DescrTag 2632 mOwner->writeInt8(23 + mCodecSpecificDataSize); 2633 mOwner->writeInt16(0x0000);// ES_ID 2634 mOwner->writeInt8(0x00); 2635 2636 mOwner->writeInt8(0x04); // DecoderConfigDescrTag 2637 mOwner->writeInt8(15 + mCodecSpecificDataSize); 2638 mOwner->writeInt8(0x40); // objectTypeIndication ISO/IEC 14492-2 2639 mOwner->writeInt8(0x15); // streamType AudioStream 2640 2641 mOwner->writeInt16(0x03); // XXX 2642 mOwner->writeInt8(0x00); // buffer size 24-bit 2643 mOwner->writeInt32(96000); // max bit rate 2644 mOwner->writeInt32(96000); // avg bit rate 2645 2646 mOwner->writeInt8(0x05); // DecoderSpecificInfoTag 2647 mOwner->writeInt8(mCodecSpecificDataSize); 2648 mOwner->write(mCodecSpecificData, mCodecSpecificDataSize); 2649 2650 static const uint8_t kData2[] = { 2651 0x06, // SLConfigDescriptorTag 2652 0x01, 2653 0x02 2654 }; 2655 mOwner->write(kData2, sizeof(kData2)); 2656 2657 mOwner->endBox(); // esds 2658} 2659 2660void MPEG4Writer::Track::writeMp4vEsdsBox() { 2661 CHECK(mCodecSpecificData); 2662 CHECK(mCodecSpecificDataSize > 0); 2663 mOwner->beginBox("esds"); 2664 2665 mOwner->writeInt32(0); // version=0, flags=0 2666 2667 mOwner->writeInt8(0x03); // ES_DescrTag 2668 mOwner->writeInt8(23 + mCodecSpecificDataSize); 2669 mOwner->writeInt16(0x0000); // ES_ID 2670 mOwner->writeInt8(0x1f); 2671 2672 mOwner->writeInt8(0x04); // DecoderConfigDescrTag 2673 mOwner->writeInt8(15 + mCodecSpecificDataSize); 2674 mOwner->writeInt8(0x20); // objectTypeIndication ISO/IEC 14492-2 2675 mOwner->writeInt8(0x11); // streamType VisualStream 2676 2677 static const uint8_t kData[] = { 2678 0x01, 0x77, 0x00, 2679 0x00, 0x03, 0xe8, 0x00, 2680 0x00, 0x03, 0xe8, 0x00 2681 }; 2682 mOwner->write(kData, sizeof(kData)); 2683 2684 mOwner->writeInt8(0x05); // DecoderSpecificInfoTag 2685 2686 mOwner->writeInt8(mCodecSpecificDataSize); 2687 mOwner->write(mCodecSpecificData, mCodecSpecificDataSize); 2688 2689 static const uint8_t kData2[] = { 2690 0x06, // SLConfigDescriptorTag 2691 0x01, 2692 0x02 2693 }; 2694 mOwner->write(kData2, sizeof(kData2)); 2695 2696 mOwner->endBox(); // esds 2697} 2698 2699void MPEG4Writer::Track::writeTkhdBox(time_t now) { 2700 mOwner->beginBox("tkhd"); 2701 // Flags = 7 to indicate that the track is enabled, and 2702 // part of the presentation 2703 mOwner->writeInt32(0x07); // version=0, flags=7 2704 mOwner->writeInt32(now); // creation time 2705 mOwner->writeInt32(now); // modification time 2706 mOwner->writeInt32(mTrackId + 1); // track id starts with 1 2707 mOwner->writeInt32(0); // reserved 2708 int64_t trakDurationUs = getDurationUs(); 2709 int32_t mvhdTimeScale = mOwner->getTimeScale(); 2710 int32_t tkhdDuration = 2711 (trakDurationUs * mvhdTimeScale + 5E5) / 1E6; 2712 mOwner->writeInt32(tkhdDuration); // in mvhd timescale 2713 mOwner->writeInt32(0); // reserved 2714 mOwner->writeInt32(0); // reserved 2715 mOwner->writeInt16(0); // layer 2716 mOwner->writeInt16(0); // alternate group 2717 mOwner->writeInt16(mIsAudio ? 0x100 : 0); // volume 2718 mOwner->writeInt16(0); // reserved 2719 2720 mOwner->writeCompositionMatrix(mRotation); // matrix 2721 2722 if (mIsAudio) { 2723 mOwner->writeInt32(0); 2724 mOwner->writeInt32(0); 2725 } else { 2726 int32_t width, height; 2727 bool success = mMeta->findInt32(kKeyWidth, &width); 2728 success = success && mMeta->findInt32(kKeyHeight, &height); 2729 CHECK(success); 2730 2731 mOwner->writeInt32(width << 16); // 32-bit fixed-point value 2732 mOwner->writeInt32(height << 16); // 32-bit fixed-point value 2733 } 2734 mOwner->endBox(); // tkhd 2735} 2736 2737void MPEG4Writer::Track::writeVmhdBox() { 2738 mOwner->beginBox("vmhd"); 2739 mOwner->writeInt32(0x01); // version=0, flags=1 2740 mOwner->writeInt16(0); // graphics mode 2741 mOwner->writeInt16(0); // opcolor 2742 mOwner->writeInt16(0); 2743 mOwner->writeInt16(0); 2744 mOwner->endBox(); 2745} 2746 2747void MPEG4Writer::Track::writeSmhdBox() { 2748 mOwner->beginBox("smhd"); 2749 mOwner->writeInt32(0); // version=0, flags=0 2750 mOwner->writeInt16(0); // balance 2751 mOwner->writeInt16(0); // reserved 2752 mOwner->endBox(); 2753} 2754 2755void MPEG4Writer::Track::writeHdlrBox() { 2756 mOwner->beginBox("hdlr"); 2757 mOwner->writeInt32(0); // version=0, flags=0 2758 mOwner->writeInt32(0); // component type: should be mhlr 2759 mOwner->writeFourcc(mIsAudio ? "soun" : "vide"); // component subtype 2760 mOwner->writeInt32(0); // reserved 2761 mOwner->writeInt32(0); // reserved 2762 mOwner->writeInt32(0); // reserved 2763 // Removing "r" for the name string just makes the string 4 byte aligned 2764 mOwner->writeCString(mIsAudio ? "SoundHandle": "VideoHandle"); // name 2765 mOwner->endBox(); 2766} 2767 2768void MPEG4Writer::Track::writeMdhdBox(time_t now) { 2769 int64_t trakDurationUs = getDurationUs(); 2770 mOwner->beginBox("mdhd"); 2771 mOwner->writeInt32(0); // version=0, flags=0 2772 mOwner->writeInt32(now); // creation time 2773 mOwner->writeInt32(now); // modification time 2774 mOwner->writeInt32(mTimeScale); // media timescale 2775 int32_t mdhdDuration = (trakDurationUs * mTimeScale + 5E5) / 1E6; 2776 mOwner->writeInt32(mdhdDuration); // use media timescale 2777 // Language follows the three letter standard ISO-639-2/T 2778 // 'e', 'n', 'g' for "English", for instance. 2779 // Each character is packed as the difference between its ASCII value and 0x60. 2780 // For "English", these are 00101, 01110, 00111. 2781 // XXX: Where is the padding bit located: 0x15C7? 2782 mOwner->writeInt16(0); // language code 2783 mOwner->writeInt16(0); // predefined 2784 mOwner->endBox(); 2785} 2786 2787void MPEG4Writer::Track::writeDamrBox() { 2788 // 3gpp2 Spec AMRSampleEntry fields 2789 mOwner->beginBox("damr"); 2790 mOwner->writeCString(" "); // vendor: 4 bytes 2791 mOwner->writeInt8(0); // decoder version 2792 mOwner->writeInt16(0x83FF); // mode set: all enabled 2793 mOwner->writeInt8(0); // mode change period 2794 mOwner->writeInt8(1); // frames per sample 2795 mOwner->endBox(); 2796} 2797 2798void MPEG4Writer::Track::writeUrlBox() { 2799 // The table index here refers to the sample description index 2800 // in the sample table entries. 2801 mOwner->beginBox("url "); 2802 mOwner->writeInt32(1); // version=0, flags=1 (self-contained) 2803 mOwner->endBox(); // url 2804} 2805 2806void MPEG4Writer::Track::writeDrefBox() { 2807 mOwner->beginBox("dref"); 2808 mOwner->writeInt32(0); // version=0, flags=0 2809 mOwner->writeInt32(1); // entry count (either url or urn) 2810 writeUrlBox(); 2811 mOwner->endBox(); // dref 2812} 2813 2814void MPEG4Writer::Track::writeDinfBox() { 2815 mOwner->beginBox("dinf"); 2816 writeDrefBox(); 2817 mOwner->endBox(); // dinf 2818} 2819 2820void MPEG4Writer::Track::writeAvccBox() { 2821 CHECK(mCodecSpecificData); 2822 CHECK(mCodecSpecificDataSize >= 5); 2823 2824 // Patch avcc's lengthSize field to match the number 2825 // of bytes we use to indicate the size of a nal unit. 2826 uint8_t *ptr = (uint8_t *)mCodecSpecificData; 2827 ptr[4] = (ptr[4] & 0xfc) | (mOwner->useNalLengthFour() ? 3 : 1); 2828 mOwner->beginBox("avcC"); 2829 mOwner->write(mCodecSpecificData, mCodecSpecificDataSize); 2830 mOwner->endBox(); // avcC 2831} 2832 2833void MPEG4Writer::Track::writeD263Box() { 2834 mOwner->beginBox("d263"); 2835 mOwner->writeInt32(0); // vendor 2836 mOwner->writeInt8(0); // decoder version 2837 mOwner->writeInt8(10); // level: 10 2838 mOwner->writeInt8(0); // profile: 0 2839 mOwner->endBox(); // d263 2840} 2841 2842// This is useful if the pixel is not square 2843void MPEG4Writer::Track::writePaspBox() { 2844 mOwner->beginBox("pasp"); 2845 mOwner->writeInt32(1 << 16); // hspacing 2846 mOwner->writeInt32(1 << 16); // vspacing 2847 mOwner->endBox(); // pasp 2848} 2849 2850void MPEG4Writer::Track::writeSttsBox() { 2851 mOwner->beginBox("stts"); 2852 mOwner->writeInt32(0); // version=0, flags=0 2853 mOwner->writeInt32(mNumSttsTableEntries); 2854 2855 // Compensate for small start time difference from different media tracks 2856 int64_t trackStartTimeOffsetUs = 0; 2857 int64_t moovStartTimeUs = mOwner->getStartTimestampUs(); 2858 if (mStartTimestampUs != moovStartTimeUs) { 2859 CHECK(mStartTimestampUs > moovStartTimeUs); 2860 trackStartTimeOffsetUs = mStartTimestampUs - moovStartTimeUs; 2861 } 2862 List<SttsTableEntry>::iterator it = mSttsTableEntries.begin(); 2863 CHECK(it != mSttsTableEntries.end() && it->sampleCount == 1); 2864 mOwner->writeInt32(it->sampleCount); 2865 int32_t dur = (trackStartTimeOffsetUs * mTimeScale + 500000LL) / 1000000LL; 2866 mOwner->writeInt32(dur + it->sampleDuration); 2867 2868 int64_t totalCount = 1; 2869 while (++it != mSttsTableEntries.end()) { 2870 mOwner->writeInt32(it->sampleCount); 2871 mOwner->writeInt32(it->sampleDuration); 2872 totalCount += it->sampleCount; 2873 } 2874 CHECK(totalCount == mNumSamples); 2875 mOwner->endBox(); // stts 2876} 2877 2878void MPEG4Writer::Track::writeCttsBox() { 2879 if (mIsAudio) { // ctts is not for audio 2880 return; 2881 } 2882 2883 // Do not write ctts box when there is no need to have it. 2884 if ((mNumCttsTableEntries == 1 && 2885 mCttsTableEntries.begin()->sampleDuration == 0) || 2886 mNumCttsTableEntries == 0) { 2887 return; 2888 } 2889 2890 LOGV("ctts box has %d entries", mNumCttsTableEntries); 2891 2892 mOwner->beginBox("ctts"); 2893 if (mHasNegativeCttsDeltaDuration) { 2894 mOwner->writeInt32(0x00010000); // version=1, flags=0 2895 } else { 2896 mOwner->writeInt32(0); // version=0, flags=0 2897 } 2898 mOwner->writeInt32(mNumCttsTableEntries); 2899 2900 int64_t totalCount = 0; 2901 for (List<CttsTableEntry>::iterator it = mCttsTableEntries.begin(); 2902 it != mCttsTableEntries.end(); ++it) { 2903 mOwner->writeInt32(it->sampleCount); 2904 mOwner->writeInt32(it->sampleDuration); 2905 totalCount += it->sampleCount; 2906 } 2907 CHECK(totalCount == mNumSamples); 2908 mOwner->endBox(); // ctts 2909} 2910 2911void MPEG4Writer::Track::writeStssBox() { 2912 mOwner->beginBox("stss"); 2913 mOwner->writeInt32(0); // version=0, flags=0 2914 mOwner->writeInt32(mNumStssTableEntries); // number of sync frames 2915 for (List<int32_t>::iterator it = mStssTableEntries.begin(); 2916 it != mStssTableEntries.end(); ++it) { 2917 mOwner->writeInt32(*it); 2918 } 2919 mOwner->endBox(); // stss 2920} 2921 2922void MPEG4Writer::Track::writeStszBox() { 2923 mOwner->beginBox("stsz"); 2924 mOwner->writeInt32(0); // version=0, flags=0 2925 if (mSamplesHaveSameSize) { 2926 List<size_t>::iterator it = mSampleSizes.begin(); 2927 mOwner->writeInt32(*it); // default sample size 2928 } else { 2929 mOwner->writeInt32(0); 2930 } 2931 mOwner->writeInt32(mNumSamples); 2932 if (!mSamplesHaveSameSize) { 2933 for (List<size_t>::iterator it = mSampleSizes.begin(); 2934 it != mSampleSizes.end(); ++it) { 2935 mOwner->writeInt32(*it); 2936 } 2937 } 2938 mOwner->endBox(); // stsz 2939} 2940 2941void MPEG4Writer::Track::writeStscBox() { 2942 mOwner->beginBox("stsc"); 2943 mOwner->writeInt32(0); // version=0, flags=0 2944 mOwner->writeInt32(mNumStscTableEntries); 2945 for (List<StscTableEntry>::iterator it = mStscTableEntries.begin(); 2946 it != mStscTableEntries.end(); ++it) { 2947 mOwner->writeInt32(it->firstChunk); 2948 mOwner->writeInt32(it->samplesPerChunk); 2949 mOwner->writeInt32(it->sampleDescriptionId); 2950 } 2951 mOwner->endBox(); // stsc 2952} 2953 2954void MPEG4Writer::Track::writeStcoBox(bool use32BitOffset) { 2955 mOwner->beginBox(use32BitOffset? "stco": "co64"); 2956 mOwner->writeInt32(0); // version=0, flags=0 2957 mOwner->writeInt32(mNumStcoTableEntries); 2958 for (List<off64_t>::iterator it = mChunkOffsets.begin(); 2959 it != mChunkOffsets.end(); ++it) { 2960 if (use32BitOffset) { 2961 mOwner->writeInt32(static_cast<int32_t>(*it)); 2962 } else { 2963 mOwner->writeInt64((*it)); 2964 } 2965 } 2966 mOwner->endBox(); // stco or co64 2967} 2968 2969void MPEG4Writer::writeUdtaBox() { 2970 beginBox("udta"); 2971 writeGeoDataBox(); 2972 endBox(); 2973} 2974 2975/* 2976 * Geodata is stored according to ISO-6709 standard. 2977 */ 2978void MPEG4Writer::writeGeoDataBox() { 2979 beginBox("\xA9xyz"); 2980 /* 2981 * For historical reasons, any user data start 2982 * with "\0xA9", must be followed by its assoicated 2983 * language code. 2984 * 0x0012: text string length 2985 * 0x15c7: lang (locale) code: en 2986 */ 2987 writeInt32(0x001215c7); 2988 writeLatitude(mLatitudex10000); 2989 writeLongitude(mLongitudex10000); 2990 writeInt8(0x2F); 2991 endBox(); 2992} 2993 2994} // namespace android 2995