MPEG4Writer.cpp revision b8e9897b8ad5d722e424d7a7e256153307ecbd93
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 26#include <media/stagefright/MPEG4Writer.h> 27#include <media/stagefright/MediaBuffer.h> 28#include <media/stagefright/MetaData.h> 29#include <media/stagefright/MediaDebug.h> 30#include <media/stagefright/MediaDefs.h> 31#include <media/stagefright/MediaErrors.h> 32#include <media/stagefright/MediaSource.h> 33#include <media/stagefright/Utils.h> 34#include <media/mediarecorder.h> 35#include <cutils/properties.h> 36#include <sys/types.h> 37#include <sys/stat.h> 38#include <fcntl.h> 39#include <unistd.h> 40 41#include "include/ESDS.h" 42 43namespace android { 44 45static const int64_t kMax32BitFileSize = 0x007fffffffLL; 46static const uint8_t kNalUnitTypeSeqParamSet = 0x07; 47static const uint8_t kNalUnitTypePicParamSet = 0x08; 48static const int64_t kInitialDelayTimeUs = 700000LL; 49 50// Using longer adjustment period to suppress fluctuations in 51// the audio encoding paths 52static const int64_t kVideoMediaTimeAdjustPeriodTimeUs = 600000000LL; // 10 minutes 53 54class MPEG4Writer::Track { 55public: 56 Track(MPEG4Writer *owner, const sp<MediaSource> &source, size_t trackId); 57 58 ~Track(); 59 60 status_t start(MetaData *params); 61 status_t stop(); 62 status_t pause(); 63 bool reachedEOS(); 64 65 int64_t getDurationUs() const; 66 int64_t getEstimatedTrackSizeBytes() const; 67 void writeTrackHeader(bool use32BitOffset = true); 68 void bufferChunk(int64_t timestampUs); 69 bool isAvc() const { return mIsAvc; } 70 bool isAudio() const { return mIsAudio; } 71 bool isMPEG4() const { return mIsMPEG4; } 72 void addChunkOffset(off64_t offset); 73 int32_t getTrackId() const { return mTrackId; } 74 status_t dump(int fd, const Vector<String16>& args) const; 75 76private: 77 MPEG4Writer *mOwner; 78 sp<MetaData> mMeta; 79 sp<MediaSource> mSource; 80 volatile bool mDone; 81 volatile bool mPaused; 82 volatile bool mResumed; 83 volatile bool mStarted; 84 bool mIsAvc; 85 bool mIsAudio; 86 bool mIsMPEG4; 87 int32_t mTrackId; 88 int64_t mTrackDurationUs; 89 int64_t mMaxChunkDurationUs; 90 91 // For realtime applications, we need to adjust the media clock 92 // for video track based on the audio media clock 93 bool mIsRealTimeRecording; 94 int64_t mMaxTimeStampUs; 95 int64_t mEstimatedTrackSizeBytes; 96 int64_t mMdatSizeBytes; 97 int32_t mTimeScale; 98 99 pthread_t mThread; 100 101 // mNumSamples is used to track how many samples in mSampleSizes List. 102 // This is to reduce the cost associated with mSampleSizes.size() call, 103 // since it is O(n). Ideally, the fix should be in List class. 104 size_t mNumSamples; 105 List<size_t> mSampleSizes; 106 bool mSamplesHaveSameSize; 107 108 List<MediaBuffer *> mChunkSamples; 109 110 size_t mNumStcoTableEntries; 111 List<off64_t> mChunkOffsets; 112 113 size_t mNumStscTableEntries; 114 struct StscTableEntry { 115 116 StscTableEntry(uint32_t chunk, uint32_t samples, uint32_t id) 117 : firstChunk(chunk), 118 samplesPerChunk(samples), 119 sampleDescriptionId(id) {} 120 121 uint32_t firstChunk; 122 uint32_t samplesPerChunk; 123 uint32_t sampleDescriptionId; 124 }; 125 List<StscTableEntry> mStscTableEntries; 126 127 size_t mNumStssTableEntries; 128 List<int32_t> mStssTableEntries; 129 130 struct SttsTableEntry { 131 132 SttsTableEntry(uint32_t count, uint32_t duration) 133 : sampleCount(count), sampleDuration(duration) {} 134 135 uint32_t sampleCount; 136 uint32_t sampleDuration; // time scale based 137 }; 138 size_t mNumSttsTableEntries; 139 List<SttsTableEntry> mSttsTableEntries; 140 141 struct CttsTableEntry { 142 CttsTableEntry(uint32_t count, int32_t timescaledDur) 143 : sampleCount(count), sampleDuration(timescaledDur) {} 144 145 uint32_t sampleCount; 146 int32_t sampleDuration; // time scale based 147 }; 148 bool mHasNegativeCttsDeltaDuration; 149 size_t mNumCttsTableEntries; 150 List<CttsTableEntry> mCttsTableEntries; 151 152 // Sequence parameter set or picture parameter set 153 struct AVCParamSet { 154 AVCParamSet(uint16_t length, const uint8_t *data) 155 : mLength(length), mData(data) {} 156 157 uint16_t mLength; 158 const uint8_t *mData; 159 }; 160 List<AVCParamSet> mSeqParamSets; 161 List<AVCParamSet> mPicParamSets; 162 uint8_t mProfileIdc; 163 uint8_t mProfileCompatible; 164 uint8_t mLevelIdc; 165 166 void *mCodecSpecificData; 167 size_t mCodecSpecificDataSize; 168 bool mGotAllCodecSpecificData; 169 bool mTrackingProgressStatus; 170 171 bool mReachedEOS; 172 int64_t mStartTimestampUs; 173 int64_t mStartTimeRealUs; 174 int64_t mFirstSampleTimeRealUs; 175 int64_t mPreviousTrackTimeUs; 176 int64_t mTrackEveryTimeDurationUs; 177 178 // Has the media time adjustment for video started? 179 bool mIsMediaTimeAdjustmentOn; 180 // The time stamp when previous media time adjustment period starts 181 int64_t mPrevMediaTimeAdjustTimestampUs; 182 // Number of vidoe frames whose time stamp may be adjusted 183 int64_t mMediaTimeAdjustNumFrames; 184 // The sample number when previous meida time adjustmnet period starts 185 int64_t mPrevMediaTimeAdjustSample; 186 // The total accumulated drift time within a period of 187 // kVideoMediaTimeAdjustPeriodTimeUs. 188 int64_t mTotalDriftTimeToAdjustUs; 189 // The total accumalated drift time since the start of the recording 190 // excluding the current time adjustment period 191 int64_t mPrevTotalAccumDriftTimeUs; 192 193 // Update the audio track's drift information. 194 void updateDriftTime(const sp<MetaData>& meta); 195 196 // Adjust the time stamp of the video track according to 197 // the drift time information from the audio track. 198 void adjustMediaTime(int64_t *timestampUs); 199 200 static void *ThreadWrapper(void *me); 201 status_t threadEntry(); 202 203 const uint8_t *parseParamSet( 204 const uint8_t *data, size_t length, int type, size_t *paramSetLen); 205 206 status_t makeAVCCodecSpecificData(const uint8_t *data, size_t size); 207 status_t copyAVCCodecSpecificData(const uint8_t *data, size_t size); 208 status_t parseAVCCodecSpecificData(const uint8_t *data, size_t size); 209 210 // Track authoring progress status 211 void trackProgressStatus(int64_t timeUs, status_t err = OK); 212 void initTrackingProgressStatus(MetaData *params); 213 214 void getCodecSpecificDataFromInputFormatIfPossible(); 215 216 // Determine the track time scale 217 // If it is an audio track, try to use the sampling rate as 218 // the time scale; however, if user chooses the overwrite 219 // value, the user-supplied time scale will be used. 220 void setTimeScale(); 221 222 // Simple validation on the codec specific data 223 status_t checkCodecSpecificData() const; 224 int32_t mRotation; 225 226 void updateTrackSizeEstimate(); 227 void addOneStscTableEntry(size_t chunkId, size_t sampleId); 228 void addOneStssTableEntry(size_t sampleId); 229 230 // Duration is time scale based 231 void addOneSttsTableEntry(size_t sampleCount, int32_t timescaledDur); 232 void addOneCttsTableEntry(size_t sampleCount, int32_t timescaledDur); 233 void sendTrackSummary(bool hasMultipleTracks); 234 235 // Write the boxes 236 void writeStcoBox(bool use32BitOffset); 237 void writeStscBox(); 238 void writeStszBox(); 239 void writeStssBox(); 240 void writeSttsBox(); 241 void writeCttsBox(); 242 void writeD263Box(); 243 void writePaspBox(); 244 void writeAvccBox(); 245 void writeUrlBox(); 246 void writeDrefBox(); 247 void writeDinfBox(); 248 void writeDamrBox(); 249 void writeMdhdBox(time_t now); 250 void writeSmhdBox(); 251 void writeVmhdBox(); 252 void writeHdlrBox(); 253 void writeTkhdBox(time_t now); 254 void writeMp4aEsdsBox(); 255 void writeMp4vEsdsBox(); 256 void writeAudioFourCCBox(); 257 void writeVideoFourCCBox(); 258 void writeStblBox(bool use32BitOffset); 259 260 Track(const Track &); 261 Track &operator=(const Track &); 262}; 263 264MPEG4Writer::MPEG4Writer(const char *filename) 265 : mFd(-1), 266 mInitCheck(NO_INIT), 267 mUse4ByteNalLength(true), 268 mUse32BitOffset(true), 269 mIsFileSizeLimitExplicitlyRequested(false), 270 mPaused(false), 271 mStarted(false), 272 mOffset(0), 273 mMdatOffset(0), 274 mEstimatedMoovBoxSize(0), 275 mInterleaveDurationUs(1000000), 276 mLatitudex10000(0), 277 mLongitudex10000(0), 278 mAreGeoTagsAvailable(false), 279 mStartTimeOffsetMs(-1) { 280 281 mFd = open(filename, O_CREAT | O_LARGEFILE | O_TRUNC | O_RDWR); 282 if (mFd >= 0) { 283 mInitCheck = OK; 284 } 285} 286 287MPEG4Writer::MPEG4Writer(int fd) 288 : mFd(dup(fd)), 289 mInitCheck(mFd < 0? NO_INIT: OK), 290 mUse4ByteNalLength(true), 291 mUse32BitOffset(true), 292 mIsFileSizeLimitExplicitlyRequested(false), 293 mPaused(false), 294 mStarted(false), 295 mOffset(0), 296 mMdatOffset(0), 297 mEstimatedMoovBoxSize(0), 298 mInterleaveDurationUs(1000000), 299 mLatitudex10000(0), 300 mLongitudex10000(0), 301 mAreGeoTagsAvailable(false), 302 mStartTimeOffsetMs(-1) { 303} 304 305MPEG4Writer::~MPEG4Writer() { 306 stop(); 307 308 while (!mTracks.empty()) { 309 List<Track *>::iterator it = mTracks.begin(); 310 delete *it; 311 (*it) = NULL; 312 mTracks.erase(it); 313 } 314 mTracks.clear(); 315} 316 317status_t MPEG4Writer::dump( 318 int fd, const Vector<String16>& args) { 319 const size_t SIZE = 256; 320 char buffer[SIZE]; 321 String8 result; 322 snprintf(buffer, SIZE, " MPEG4Writer %p\n", this); 323 result.append(buffer); 324 snprintf(buffer, SIZE, " mStarted: %s\n", mStarted? "true": "false"); 325 result.append(buffer); 326 ::write(fd, result.string(), result.size()); 327 for (List<Track *>::iterator it = mTracks.begin(); 328 it != mTracks.end(); ++it) { 329 (*it)->dump(fd, args); 330 } 331 return OK; 332} 333 334status_t MPEG4Writer::Track::dump( 335 int fd, const Vector<String16>& args) const { 336 const size_t SIZE = 256; 337 char buffer[SIZE]; 338 String8 result; 339 snprintf(buffer, SIZE, " %s track\n", mIsAudio? "Audio": "Video"); 340 result.append(buffer); 341 snprintf(buffer, SIZE, " reached EOS: %s\n", 342 mReachedEOS? "true": "false"); 343 result.append(buffer); 344 ::write(fd, result.string(), result.size()); 345 return OK; 346} 347 348status_t MPEG4Writer::addSource(const sp<MediaSource> &source) { 349 Mutex::Autolock l(mLock); 350 if (mStarted) { 351 LOGE("Attempt to add source AFTER recording is started"); 352 return UNKNOWN_ERROR; 353 } 354 Track *track = new Track(this, source, mTracks.size()); 355 mTracks.push_back(track); 356 357 return OK; 358} 359 360status_t MPEG4Writer::startTracks(MetaData *params) { 361 for (List<Track *>::iterator it = mTracks.begin(); 362 it != mTracks.end(); ++it) { 363 status_t err = (*it)->start(params); 364 365 if (err != OK) { 366 for (List<Track *>::iterator it2 = mTracks.begin(); 367 it2 != it; ++it2) { 368 (*it2)->stop(); 369 } 370 371 return err; 372 } 373 } 374 return OK; 375} 376 377int64_t MPEG4Writer::estimateMoovBoxSize(int32_t bitRate) { 378 // This implementation is highly experimental/heurisitic. 379 // 380 // Statistical analysis shows that metadata usually accounts 381 // for a small portion of the total file size, usually < 0.6%. 382 383 // The default MIN_MOOV_BOX_SIZE is set to 0.6% x 1MB / 2, 384 // where 1MB is the common file size limit for MMS application. 385 // The default MAX _MOOV_BOX_SIZE value is based on about 3 386 // minute video recording with a bit rate about 3 Mbps, because 387 // statistics also show that most of the video captured are going 388 // to be less than 3 minutes. 389 390 // If the estimation is wrong, we will pay the price of wasting 391 // some reserved space. This should not happen so often statistically. 392 static const int32_t factor = mUse32BitOffset? 1: 2; 393 static const int64_t MIN_MOOV_BOX_SIZE = 3 * 1024; // 3 KB 394 static const int64_t MAX_MOOV_BOX_SIZE = (180 * 3000000 * 6LL / 8000); 395 int64_t size = MIN_MOOV_BOX_SIZE; 396 397 // Max file size limit is set 398 if (mMaxFileSizeLimitBytes != 0 && mIsFileSizeLimitExplicitlyRequested) { 399 size = mMaxFileSizeLimitBytes * 6 / 1000; 400 } 401 402 // Max file duration limit is set 403 if (mMaxFileDurationLimitUs != 0) { 404 if (bitRate > 0) { 405 int64_t size2 = 406 ((mMaxFileDurationLimitUs * bitRate * 6) / 1000 / 8000000); 407 if (mMaxFileSizeLimitBytes != 0 && mIsFileSizeLimitExplicitlyRequested) { 408 // When both file size and duration limits are set, 409 // we use the smaller limit of the two. 410 if (size > size2) { 411 size = size2; 412 } 413 } else { 414 // Only max file duration limit is set 415 size = size2; 416 } 417 } 418 } 419 420 if (size < MIN_MOOV_BOX_SIZE) { 421 size = MIN_MOOV_BOX_SIZE; 422 } 423 424 // Any long duration recording will be probably end up with 425 // non-streamable mp4 file. 426 if (size > MAX_MOOV_BOX_SIZE) { 427 size = MAX_MOOV_BOX_SIZE; 428 } 429 430 LOGI("limits: %lld/%lld bytes/us, bit rate: %d bps and the estimated" 431 " moov size %lld bytes", 432 mMaxFileSizeLimitBytes, mMaxFileDurationLimitUs, bitRate, size); 433 return factor * size; 434} 435 436status_t MPEG4Writer::start(MetaData *param) { 437 if (mInitCheck != OK) { 438 return UNKNOWN_ERROR; 439 } 440 441 /* 442 * Check mMaxFileSizeLimitBytes at the beginning 443 * since mMaxFileSizeLimitBytes may be implicitly 444 * changed later for 32-bit file offset even if 445 * user does not ask to set it explicitly. 446 */ 447 if (mMaxFileSizeLimitBytes != 0) { 448 mIsFileSizeLimitExplicitlyRequested = true; 449 } 450 451 int32_t use64BitOffset; 452 if (param && 453 param->findInt32(kKey64BitFileOffset, &use64BitOffset) && 454 use64BitOffset) { 455 mUse32BitOffset = false; 456 } 457 458 if (mUse32BitOffset) { 459 // Implicit 32 bit file size limit 460 if (mMaxFileSizeLimitBytes == 0) { 461 mMaxFileSizeLimitBytes = kMax32BitFileSize; 462 } 463 464 // If file size is set to be larger than the 32 bit file 465 // size limit, treat it as an error. 466 if (mMaxFileSizeLimitBytes > kMax32BitFileSize) { 467 LOGW("32-bit file size limit (%lld bytes) too big. " 468 "It is changed to %lld bytes", 469 mMaxFileSizeLimitBytes, kMax32BitFileSize); 470 mMaxFileSizeLimitBytes = kMax32BitFileSize; 471 } 472 } 473 474 int32_t use2ByteNalLength; 475 if (param && 476 param->findInt32(kKey2ByteNalLength, &use2ByteNalLength) && 477 use2ByteNalLength) { 478 mUse4ByteNalLength = false; 479 } 480 481 mStartTimestampUs = -1; 482 483 if (mStarted) { 484 if (mPaused) { 485 mPaused = false; 486 return startTracks(param); 487 } 488 return OK; 489 } 490 491 if (!param || 492 !param->findInt32(kKeyTimeScale, &mTimeScale)) { 493 mTimeScale = 1000; 494 } 495 CHECK(mTimeScale > 0); 496 LOGV("movie time scale: %d", mTimeScale); 497 498 mStreamableFile = true; 499 mWriteMoovBoxToMemory = false; 500 mMoovBoxBuffer = NULL; 501 mMoovBoxBufferOffset = 0; 502 503 writeFtypBox(param); 504 505 mFreeBoxOffset = mOffset; 506 507 if (mEstimatedMoovBoxSize == 0) { 508 int32_t bitRate = -1; 509 if (param) { 510 param->findInt32(kKeyBitRate, &bitRate); 511 } 512 mEstimatedMoovBoxSize = estimateMoovBoxSize(bitRate); 513 } 514 CHECK(mEstimatedMoovBoxSize >= 8); 515 lseek64(mFd, mFreeBoxOffset, SEEK_SET); 516 writeInt32(mEstimatedMoovBoxSize); 517 write("free", 4); 518 519 mMdatOffset = mFreeBoxOffset + mEstimatedMoovBoxSize; 520 mOffset = mMdatOffset; 521 lseek64(mFd, mMdatOffset, SEEK_SET); 522 if (mUse32BitOffset) { 523 write("????mdat", 8); 524 } else { 525 write("\x00\x00\x00\x01mdat????????", 16); 526 } 527 528 status_t err = startWriterThread(); 529 if (err != OK) { 530 return err; 531 } 532 533 err = startTracks(param); 534 if (err != OK) { 535 return err; 536 } 537 538 mStarted = true; 539 return OK; 540} 541 542bool MPEG4Writer::use32BitFileOffset() const { 543 return mUse32BitOffset; 544} 545 546status_t MPEG4Writer::pause() { 547 if (mInitCheck != OK) { 548 return OK; 549 } 550 mPaused = true; 551 status_t err = OK; 552 for (List<Track *>::iterator it = mTracks.begin(); 553 it != mTracks.end(); ++it) { 554 status_t status = (*it)->pause(); 555 if (status != OK) { 556 err = status; 557 } 558 } 559 return err; 560} 561 562void MPEG4Writer::stopWriterThread() { 563 LOGD("Stopping writer thread"); 564 565 { 566 Mutex::Autolock autolock(mLock); 567 568 mDone = true; 569 mChunkReadyCondition.signal(); 570 } 571 572 void *dummy; 573 pthread_join(mThread, &dummy); 574 LOGD("Writer thread stopped"); 575} 576 577/* 578 * MP4 file standard defines a composition matrix: 579 * | a b u | 580 * | c d v | 581 * | x y w | 582 * 583 * the element in the matrix is stored in the following 584 * order: {a, b, u, c, d, v, x, y, w}, 585 * where a, b, c, d, x, and y is in 16.16 format, while 586 * u, v and w is in 2.30 format. 587 */ 588void MPEG4Writer::writeCompositionMatrix(int degrees) { 589 LOGV("writeCompositionMatrix"); 590 uint32_t a = 0x00010000; 591 uint32_t b = 0; 592 uint32_t c = 0; 593 uint32_t d = 0x00010000; 594 switch (degrees) { 595 case 0: 596 break; 597 case 90: 598 a = 0; 599 b = 0x00010000; 600 c = 0xFFFF0000; 601 d = 0; 602 break; 603 case 180: 604 a = 0xFFFF0000; 605 d = 0xFFFF0000; 606 break; 607 case 270: 608 a = 0; 609 b = 0xFFFF0000; 610 c = 0x00010000; 611 d = 0; 612 break; 613 default: 614 CHECK(!"Should never reach this unknown rotation"); 615 break; 616 } 617 618 writeInt32(a); // a 619 writeInt32(b); // b 620 writeInt32(0); // u 621 writeInt32(c); // c 622 writeInt32(d); // d 623 writeInt32(0); // v 624 writeInt32(0); // x 625 writeInt32(0); // y 626 writeInt32(0x40000000); // w 627} 628 629 630status_t MPEG4Writer::stop() { 631 if (mInitCheck != OK) { 632 return OK; 633 } 634 635 status_t err = OK; 636 int64_t maxDurationUs = 0; 637 int64_t minDurationUs = 0x7fffffffffffffffLL; 638 for (List<Track *>::iterator it = mTracks.begin(); 639 it != mTracks.end(); ++it) { 640 status_t status = (*it)->stop(); 641 if (err == OK && status != OK) { 642 err = status; 643 } 644 645 int64_t durationUs = (*it)->getDurationUs(); 646 if (durationUs > maxDurationUs) { 647 maxDurationUs = durationUs; 648 } 649 if (durationUs < minDurationUs) { 650 minDurationUs = durationUs; 651 } 652 } 653 654 if (mTracks.size() > 1) { 655 LOGD("Duration from tracks range is [%lld, %lld] us", 656 minDurationUs, maxDurationUs); 657 } 658 659 stopWriterThread(); 660 661 // Do not write out movie header on error. 662 if (err != OK) { 663 close(mFd); 664 mFd = -1; 665 mInitCheck = NO_INIT; 666 mStarted = false; 667 return err; 668 } 669 670 // Fix up the size of the 'mdat' chunk. 671 if (mUse32BitOffset) { 672 lseek64(mFd, mMdatOffset, SEEK_SET); 673 int32_t size = htonl(static_cast<int32_t>(mOffset - mMdatOffset)); 674 ::write(mFd, &size, 4); 675 } else { 676 lseek64(mFd, mMdatOffset + 8, SEEK_SET); 677 int64_t size = mOffset - mMdatOffset; 678 size = hton64(size); 679 ::write(mFd, &size, 8); 680 } 681 lseek64(mFd, mOffset, SEEK_SET); 682 683 const off64_t moovOffset = mOffset; 684 mWriteMoovBoxToMemory = true; 685 mMoovBoxBuffer = (uint8_t *) malloc(mEstimatedMoovBoxSize); 686 mMoovBoxBufferOffset = 0; 687 CHECK(mMoovBoxBuffer != NULL); 688 writeMoovBox(maxDurationUs); 689 690 mWriteMoovBoxToMemory = false; 691 if (mStreamableFile) { 692 CHECK(mMoovBoxBufferOffset + 8 <= mEstimatedMoovBoxSize); 693 694 // Moov box 695 lseek64(mFd, mFreeBoxOffset, SEEK_SET); 696 mOffset = mFreeBoxOffset; 697 write(mMoovBoxBuffer, 1, mMoovBoxBufferOffset); 698 699 // Free box 700 lseek64(mFd, mOffset, SEEK_SET); 701 writeInt32(mEstimatedMoovBoxSize - mMoovBoxBufferOffset); 702 write("free", 4); 703 704 // Free temp memory 705 free(mMoovBoxBuffer); 706 mMoovBoxBuffer = NULL; 707 mMoovBoxBufferOffset = 0; 708 } else { 709 LOGI("The mp4 file will not be streamable."); 710 } 711 712 CHECK(mBoxes.empty()); 713 714 close(mFd); 715 mFd = -1; 716 mInitCheck = NO_INIT; 717 mStarted = false; 718 719 return err; 720} 721 722void MPEG4Writer::writeMvhdBox(int64_t durationUs) { 723 time_t now = time(NULL); 724 beginBox("mvhd"); 725 writeInt32(0); // version=0, flags=0 726 writeInt32(now); // creation time 727 writeInt32(now); // modification time 728 writeInt32(mTimeScale); // mvhd timescale 729 int32_t duration = (durationUs * mTimeScale + 5E5) / 1E6; 730 writeInt32(duration); 731 writeInt32(0x10000); // rate: 1.0 732 writeInt16(0x100); // volume 733 writeInt16(0); // reserved 734 writeInt32(0); // reserved 735 writeInt32(0); // reserved 736 writeCompositionMatrix(0); // matrix 737 writeInt32(0); // predefined 738 writeInt32(0); // predefined 739 writeInt32(0); // predefined 740 writeInt32(0); // predefined 741 writeInt32(0); // predefined 742 writeInt32(0); // predefined 743 writeInt32(mTracks.size() + 1); // nextTrackID 744 endBox(); // mvhd 745} 746 747void MPEG4Writer::writeMoovBox(int64_t durationUs) { 748 beginBox("moov"); 749 writeMvhdBox(durationUs); 750 if (mAreGeoTagsAvailable) { 751 writeUdtaBox(); 752 } 753 int32_t id = 1; 754 for (List<Track *>::iterator it = mTracks.begin(); 755 it != mTracks.end(); ++it, ++id) { 756 (*it)->writeTrackHeader(mUse32BitOffset); 757 } 758 endBox(); // moov 759} 760 761void MPEG4Writer::writeFtypBox(MetaData *param) { 762 beginBox("ftyp"); 763 764 int32_t fileType; 765 if (param && param->findInt32(kKeyFileType, &fileType) && 766 fileType != OUTPUT_FORMAT_MPEG_4) { 767 writeFourcc("3gp4"); 768 } else { 769 writeFourcc("isom"); 770 } 771 772 writeInt32(0); 773 writeFourcc("isom"); 774 writeFourcc("3gp4"); 775 endBox(); 776} 777 778static bool isTestModeEnabled() { 779#if (PROPERTY_VALUE_MAX < 5) 780#error "PROPERTY_VALUE_MAX must be at least 5" 781#endif 782 783 // Test mode is enabled only if rw.media.record.test system 784 // property is enabled. 785 char value[PROPERTY_VALUE_MAX]; 786 if (property_get("rw.media.record.test", value, NULL) && 787 (!strcasecmp(value, "true") || !strcasecmp(value, "1"))) { 788 return true; 789 } 790 return false; 791} 792 793void MPEG4Writer::sendSessionSummary() { 794 // Send session summary only if test mode is enabled 795 if (!isTestModeEnabled()) { 796 return; 797 } 798 799 for (List<ChunkInfo>::iterator it = mChunkInfos.begin(); 800 it != mChunkInfos.end(); ++it) { 801 int trackNum = it->mTrack->getTrackId() << 28; 802 notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 803 trackNum | MEDIA_RECORDER_TRACK_INTER_CHUNK_TIME_MS, 804 it->mMaxInterChunkDurUs); 805 } 806} 807 808status_t MPEG4Writer::setInterleaveDuration(uint32_t durationUs) { 809 mInterleaveDurationUs = durationUs; 810 return OK; 811} 812 813void MPEG4Writer::lock() { 814 mLock.lock(); 815} 816 817void MPEG4Writer::unlock() { 818 mLock.unlock(); 819} 820 821off64_t MPEG4Writer::addSample_l(MediaBuffer *buffer) { 822 off64_t old_offset = mOffset; 823 824 ::write(mFd, 825 (const uint8_t *)buffer->data() + buffer->range_offset(), 826 buffer->range_length()); 827 828 mOffset += buffer->range_length(); 829 830 return old_offset; 831} 832 833static void StripStartcode(MediaBuffer *buffer) { 834 if (buffer->range_length() < 4) { 835 return; 836 } 837 838 const uint8_t *ptr = 839 (const uint8_t *)buffer->data() + buffer->range_offset(); 840 841 if (!memcmp(ptr, "\x00\x00\x00\x01", 4)) { 842 buffer->set_range( 843 buffer->range_offset() + 4, buffer->range_length() - 4); 844 } 845} 846 847off64_t MPEG4Writer::addLengthPrefixedSample_l(MediaBuffer *buffer) { 848 off64_t old_offset = mOffset; 849 850 size_t length = buffer->range_length(); 851 852 if (mUse4ByteNalLength) { 853 uint8_t x = length >> 24; 854 ::write(mFd, &x, 1); 855 x = (length >> 16) & 0xff; 856 ::write(mFd, &x, 1); 857 x = (length >> 8) & 0xff; 858 ::write(mFd, &x, 1); 859 x = length & 0xff; 860 ::write(mFd, &x, 1); 861 862 ::write(mFd, 863 (const uint8_t *)buffer->data() + buffer->range_offset(), 864 length); 865 866 mOffset += length + 4; 867 } else { 868 CHECK(length < 65536); 869 870 uint8_t x = length >> 8; 871 ::write(mFd, &x, 1); 872 x = length & 0xff; 873 ::write(mFd, &x, 1); 874 ::write(mFd, (const uint8_t *)buffer->data() + buffer->range_offset(), length); 875 mOffset += length + 2; 876 } 877 878 return old_offset; 879} 880 881size_t MPEG4Writer::write( 882 const void *ptr, size_t size, size_t nmemb) { 883 884 const size_t bytes = size * nmemb; 885 if (mWriteMoovBoxToMemory) { 886 // This happens only when we write the moov box at the end of 887 // recording, not for each output video/audio frame we receive. 888 off64_t moovBoxSize = 8 + mMoovBoxBufferOffset + bytes; 889 if (moovBoxSize > mEstimatedMoovBoxSize) { 890 for (List<off64_t>::iterator it = mBoxes.begin(); 891 it != mBoxes.end(); ++it) { 892 (*it) += mOffset; 893 } 894 lseek64(mFd, mOffset, SEEK_SET); 895 ::write(mFd, mMoovBoxBuffer, mMoovBoxBufferOffset); 896 ::write(mFd, ptr, size * nmemb); 897 mOffset += (bytes + mMoovBoxBufferOffset); 898 free(mMoovBoxBuffer); 899 mMoovBoxBuffer = NULL; 900 mMoovBoxBufferOffset = 0; 901 mWriteMoovBoxToMemory = false; 902 mStreamableFile = false; 903 } else { 904 memcpy(mMoovBoxBuffer + mMoovBoxBufferOffset, ptr, bytes); 905 mMoovBoxBufferOffset += bytes; 906 } 907 } else { 908 ::write(mFd, ptr, size * nmemb); 909 mOffset += bytes; 910 } 911 return bytes; 912} 913 914void MPEG4Writer::beginBox(const char *fourcc) { 915 CHECK_EQ(strlen(fourcc), 4); 916 917 mBoxes.push_back(mWriteMoovBoxToMemory? 918 mMoovBoxBufferOffset: mOffset); 919 920 writeInt32(0); 921 writeFourcc(fourcc); 922} 923 924void MPEG4Writer::endBox() { 925 CHECK(!mBoxes.empty()); 926 927 off64_t offset = *--mBoxes.end(); 928 mBoxes.erase(--mBoxes.end()); 929 930 if (mWriteMoovBoxToMemory) { 931 int32_t x = htonl(mMoovBoxBufferOffset - offset); 932 memcpy(mMoovBoxBuffer + offset, &x, 4); 933 } else { 934 lseek64(mFd, offset, SEEK_SET); 935 writeInt32(mOffset - offset); 936 mOffset -= 4; 937 lseek64(mFd, mOffset, SEEK_SET); 938 } 939} 940 941void MPEG4Writer::writeInt8(int8_t x) { 942 write(&x, 1, 1); 943} 944 945void MPEG4Writer::writeInt16(int16_t x) { 946 x = htons(x); 947 write(&x, 1, 2); 948} 949 950void MPEG4Writer::writeInt32(int32_t x) { 951 x = htonl(x); 952 write(&x, 1, 4); 953} 954 955void MPEG4Writer::writeInt64(int64_t x) { 956 x = hton64(x); 957 write(&x, 1, 8); 958} 959 960void MPEG4Writer::writeCString(const char *s) { 961 size_t n = strlen(s); 962 write(s, 1, n + 1); 963} 964 965void MPEG4Writer::writeFourcc(const char *s) { 966 CHECK_EQ(strlen(s), 4); 967 write(s, 1, 4); 968} 969 970 971// Written in +/-DD.DDDD format 972void MPEG4Writer::writeLatitude(int degreex10000) { 973 bool isNegative = (degreex10000 < 0); 974 char sign = isNegative? '-': '+'; 975 976 // Handle the whole part 977 char str[9]; 978 int wholePart = degreex10000 / 10000; 979 if (wholePart == 0) { 980 snprintf(str, 5, "%c%.2d.", sign, wholePart); 981 } else { 982 snprintf(str, 5, "%+.2d.", wholePart); 983 } 984 985 // Handle the fractional part 986 int fractionalPart = degreex10000 - (wholePart * 10000); 987 if (fractionalPart < 0) { 988 fractionalPart = -fractionalPart; 989 } 990 snprintf(&str[4], 5, "%.4d", fractionalPart); 991 992 // Do not write the null terminator 993 write(str, 1, 8); 994} 995 996// Written in +/- DDD.DDDD format 997void MPEG4Writer::writeLongitude(int degreex10000) { 998 bool isNegative = (degreex10000 < 0); 999 char sign = isNegative? '-': '+'; 1000 1001 // Handle the whole part 1002 char str[10]; 1003 int wholePart = degreex10000 / 10000; 1004 if (wholePart == 0) { 1005 snprintf(str, 6, "%c%.3d.", sign, wholePart); 1006 } else { 1007 snprintf(str, 6, "%+.3d.", wholePart); 1008 } 1009 1010 // Handle the fractional part 1011 int fractionalPart = degreex10000 - (wholePart * 10000); 1012 if (fractionalPart < 0) { 1013 fractionalPart = -fractionalPart; 1014 } 1015 snprintf(&str[5], 5, "%.4d", fractionalPart); 1016 1017 // Do not write the null terminator 1018 write(str, 1, 9); 1019} 1020 1021/* 1022 * Geodata is stored according to ISO-6709 standard. 1023 * latitudex10000 is latitude in degrees times 10000, and 1024 * longitudex10000 is longitude in degrees times 10000. 1025 * The range for the latitude is in [-90, +90], and 1026 * The range for the longitude is in [-180, +180] 1027 */ 1028status_t MPEG4Writer::setGeoData(int latitudex10000, int longitudex10000) { 1029 // Is latitude or longitude out of range? 1030 if (latitudex10000 < -900000 || latitudex10000 > 900000 || 1031 longitudex10000 < -1800000 || longitudex10000 > 1800000) { 1032 return BAD_VALUE; 1033 } 1034 1035 mLatitudex10000 = latitudex10000; 1036 mLongitudex10000 = longitudex10000; 1037 mAreGeoTagsAvailable = true; 1038 return OK; 1039} 1040 1041void MPEG4Writer::write(const void *data, size_t size) { 1042 write(data, 1, size); 1043} 1044 1045bool MPEG4Writer::isFileStreamable() const { 1046 return mStreamableFile; 1047} 1048 1049bool MPEG4Writer::exceedsFileSizeLimit() { 1050 // No limit 1051 if (mMaxFileSizeLimitBytes == 0) { 1052 return false; 1053 } 1054 1055 int64_t nTotalBytesEstimate = static_cast<int64_t>(mEstimatedMoovBoxSize); 1056 for (List<Track *>::iterator it = mTracks.begin(); 1057 it != mTracks.end(); ++it) { 1058 nTotalBytesEstimate += (*it)->getEstimatedTrackSizeBytes(); 1059 } 1060 1061 // Be conservative in the estimate: do not exceed 95% of 1062 // the target file limit. For small target file size limit, though, 1063 // this will not help. 1064 return (nTotalBytesEstimate >= (95 * mMaxFileSizeLimitBytes) / 100); 1065} 1066 1067bool MPEG4Writer::exceedsFileDurationLimit() { 1068 // No limit 1069 if (mMaxFileDurationLimitUs == 0) { 1070 return false; 1071 } 1072 1073 for (List<Track *>::iterator it = mTracks.begin(); 1074 it != mTracks.end(); ++it) { 1075 if ((*it)->getDurationUs() >= mMaxFileDurationLimitUs) { 1076 return true; 1077 } 1078 } 1079 return false; 1080} 1081 1082bool MPEG4Writer::reachedEOS() { 1083 bool allDone = true; 1084 for (List<Track *>::iterator it = mTracks.begin(); 1085 it != mTracks.end(); ++it) { 1086 if (!(*it)->reachedEOS()) { 1087 allDone = false; 1088 break; 1089 } 1090 } 1091 1092 return allDone; 1093} 1094 1095void MPEG4Writer::setStartTimestampUs(int64_t timeUs) { 1096 LOGI("setStartTimestampUs: %lld", timeUs); 1097 CHECK(timeUs >= 0); 1098 Mutex::Autolock autoLock(mLock); 1099 if (mStartTimestampUs < 0 || mStartTimestampUs > timeUs) { 1100 mStartTimestampUs = timeUs; 1101 LOGI("Earliest track starting time: %lld", mStartTimestampUs); 1102 } 1103} 1104 1105int64_t MPEG4Writer::getStartTimestampUs() { 1106 Mutex::Autolock autoLock(mLock); 1107 return mStartTimestampUs; 1108} 1109 1110size_t MPEG4Writer::numTracks() { 1111 Mutex::Autolock autolock(mLock); 1112 return mTracks.size(); 1113} 1114 1115//////////////////////////////////////////////////////////////////////////////// 1116 1117MPEG4Writer::Track::Track( 1118 MPEG4Writer *owner, const sp<MediaSource> &source, size_t trackId) 1119 : mOwner(owner), 1120 mMeta(source->getFormat()), 1121 mSource(source), 1122 mDone(false), 1123 mPaused(false), 1124 mResumed(false), 1125 mStarted(false), 1126 mTrackId(trackId), 1127 mTrackDurationUs(0), 1128 mEstimatedTrackSizeBytes(0), 1129 mSamplesHaveSameSize(true), 1130 mCodecSpecificData(NULL), 1131 mCodecSpecificDataSize(0), 1132 mGotAllCodecSpecificData(false), 1133 mReachedEOS(false), 1134 mRotation(0) { 1135 getCodecSpecificDataFromInputFormatIfPossible(); 1136 1137 const char *mime; 1138 mMeta->findCString(kKeyMIMEType, &mime); 1139 mIsAvc = !strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_AVC); 1140 mIsAudio = !strncasecmp(mime, "audio/", 6); 1141 mIsMPEG4 = !strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_MPEG4) || 1142 !strcasecmp(mime, MEDIA_MIMETYPE_AUDIO_AAC); 1143 1144 setTimeScale(); 1145} 1146 1147void MPEG4Writer::Track::updateTrackSizeEstimate() { 1148 1149 int64_t stcoBoxSizeBytes = mOwner->use32BitFileOffset() 1150 ? mNumStcoTableEntries * 4 1151 : mNumStcoTableEntries * 8; 1152 1153 int64_t stszBoxSizeBytes = mSamplesHaveSameSize? 4: (mNumSamples * 4); 1154 1155 mEstimatedTrackSizeBytes = mMdatSizeBytes; // media data size 1156 if (!mOwner->isFileStreamable()) { 1157 // Reserved free space is not large enough to hold 1158 // all meta data and thus wasted. 1159 mEstimatedTrackSizeBytes += mNumStscTableEntries * 12 + // stsc box size 1160 mNumStssTableEntries * 4 + // stss box size 1161 mNumSttsTableEntries * 8 + // stts box size 1162 mNumCttsTableEntries * 8 + // ctts box size 1163 stcoBoxSizeBytes + // stco box size 1164 stszBoxSizeBytes; // stsz box size 1165 } 1166} 1167 1168void MPEG4Writer::Track::addOneStscTableEntry( 1169 size_t chunkId, size_t sampleId) { 1170 1171 StscTableEntry stscEntry(chunkId, sampleId, 1); 1172 mStscTableEntries.push_back(stscEntry); 1173 ++mNumStscTableEntries; 1174} 1175 1176void MPEG4Writer::Track::addOneStssTableEntry(size_t sampleId) { 1177 mStssTableEntries.push_back(sampleId); 1178 ++mNumStssTableEntries; 1179} 1180 1181void MPEG4Writer::Track::addOneSttsTableEntry( 1182 size_t sampleCount, int32_t duration) { 1183 1184 SttsTableEntry sttsEntry(sampleCount, duration); 1185 mSttsTableEntries.push_back(sttsEntry); 1186 ++mNumSttsTableEntries; 1187} 1188 1189void MPEG4Writer::Track::addOneCttsTableEntry( 1190 size_t sampleCount, int32_t duration) { 1191 1192 if (mIsAudio) { 1193 return; 1194 } 1195 if (duration < 0 && !mHasNegativeCttsDeltaDuration) { 1196 mHasNegativeCttsDeltaDuration = true; 1197 } 1198 CttsTableEntry cttsEntry(sampleCount, duration); 1199 mCttsTableEntries.push_back(cttsEntry); 1200 ++mNumCttsTableEntries; 1201} 1202 1203void MPEG4Writer::Track::addChunkOffset(off64_t offset) { 1204 ++mNumStcoTableEntries; 1205 mChunkOffsets.push_back(offset); 1206} 1207 1208void MPEG4Writer::Track::setTimeScale() { 1209 LOGV("setTimeScale"); 1210 // Default time scale 1211 mTimeScale = 90000; 1212 1213 if (mIsAudio) { 1214 // Use the sampling rate as the default time scale for audio track. 1215 int32_t sampleRate; 1216 bool success = mMeta->findInt32(kKeySampleRate, &sampleRate); 1217 CHECK(success); 1218 mTimeScale = sampleRate; 1219 } 1220 1221 // If someone would like to overwrite the timescale, use user-supplied value. 1222 int32_t timeScale; 1223 if (mMeta->findInt32(kKeyTimeScale, &timeScale)) { 1224 mTimeScale = timeScale; 1225 } 1226 1227 CHECK(mTimeScale > 0); 1228} 1229 1230void MPEG4Writer::Track::getCodecSpecificDataFromInputFormatIfPossible() { 1231 const char *mime; 1232 CHECK(mMeta->findCString(kKeyMIMEType, &mime)); 1233 1234 if (!strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_AVC)) { 1235 uint32_t type; 1236 const void *data; 1237 size_t size; 1238 if (mMeta->findData(kKeyAVCC, &type, &data, &size)) { 1239 mCodecSpecificData = malloc(size); 1240 mCodecSpecificDataSize = size; 1241 memcpy(mCodecSpecificData, data, size); 1242 mGotAllCodecSpecificData = true; 1243 } 1244 } else if (!strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_MPEG4) 1245 || !strcasecmp(mime, MEDIA_MIMETYPE_AUDIO_AAC)) { 1246 uint32_t type; 1247 const void *data; 1248 size_t size; 1249 if (mMeta->findData(kKeyESDS, &type, &data, &size)) { 1250 ESDS esds(data, size); 1251 if (esds.getCodecSpecificInfo(&data, &size) == OK) { 1252 mCodecSpecificData = malloc(size); 1253 mCodecSpecificDataSize = size; 1254 memcpy(mCodecSpecificData, data, size); 1255 mGotAllCodecSpecificData = true; 1256 } 1257 } 1258 } 1259} 1260 1261MPEG4Writer::Track::~Track() { 1262 stop(); 1263 1264 if (mCodecSpecificData != NULL) { 1265 free(mCodecSpecificData); 1266 mCodecSpecificData = NULL; 1267 } 1268} 1269 1270void MPEG4Writer::Track::initTrackingProgressStatus(MetaData *params) { 1271 LOGV("initTrackingProgressStatus"); 1272 mPreviousTrackTimeUs = -1; 1273 mTrackingProgressStatus = false; 1274 mTrackEveryTimeDurationUs = 0; 1275 { 1276 int64_t timeUs; 1277 if (params && params->findInt64(kKeyTrackTimeStatus, &timeUs)) { 1278 LOGV("Receive request to track progress status for every %lld us", timeUs); 1279 mTrackEveryTimeDurationUs = timeUs; 1280 mTrackingProgressStatus = true; 1281 } 1282 } 1283} 1284 1285// static 1286void *MPEG4Writer::ThreadWrapper(void *me) { 1287 LOGV("ThreadWrapper: %p", me); 1288 MPEG4Writer *writer = static_cast<MPEG4Writer *>(me); 1289 writer->threadFunc(); 1290 return NULL; 1291} 1292 1293void MPEG4Writer::bufferChunk(const Chunk& chunk) { 1294 LOGV("bufferChunk: %p", chunk.mTrack); 1295 Mutex::Autolock autolock(mLock); 1296 CHECK_EQ(mDone, false); 1297 1298 for (List<ChunkInfo>::iterator it = mChunkInfos.begin(); 1299 it != mChunkInfos.end(); ++it) { 1300 1301 if (chunk.mTrack == it->mTrack) { // Found owner 1302 it->mChunks.push_back(chunk); 1303 mChunkReadyCondition.signal(); 1304 return; 1305 } 1306 } 1307 1308 CHECK("Received a chunk for a unknown track" == 0); 1309} 1310 1311void MPEG4Writer::writeChunkToFile(Chunk* chunk) { 1312 LOGV("writeChunkToFile: %lld from %s track", 1313 chunk.mTimestampUs, chunk.mTrack->isAudio()? "audio": "video"); 1314 1315 int32_t isFirstSample = true; 1316 while (!chunk->mSamples.empty()) { 1317 List<MediaBuffer *>::iterator it = chunk->mSamples.begin(); 1318 1319 off64_t offset = chunk->mTrack->isAvc() 1320 ? addLengthPrefixedSample_l(*it) 1321 : addSample_l(*it); 1322 1323 if (isFirstSample) { 1324 chunk->mTrack->addChunkOffset(offset); 1325 isFirstSample = false; 1326 } 1327 1328 (*it)->release(); 1329 (*it) = NULL; 1330 chunk->mSamples.erase(it); 1331 } 1332 chunk->mSamples.clear(); 1333} 1334 1335void MPEG4Writer::writeAllChunks() { 1336 LOGV("writeAllChunks"); 1337 size_t outstandingChunks = 0; 1338 Chunk chunk; 1339 while (findChunkToWrite(&chunk)) { 1340 writeChunkToFile(&chunk); 1341 ++outstandingChunks; 1342 } 1343 1344 sendSessionSummary(); 1345 1346 mChunkInfos.clear(); 1347 LOGD("%d chunks are written in the last batch", outstandingChunks); 1348} 1349 1350bool MPEG4Writer::findChunkToWrite(Chunk *chunk) { 1351 LOGV("findChunkToWrite"); 1352 1353 int64_t minTimestampUs = 0x7FFFFFFFFFFFFFFFLL; 1354 Track *track = NULL; 1355 for (List<ChunkInfo>::iterator it = mChunkInfos.begin(); 1356 it != mChunkInfos.end(); ++it) { 1357 if (!it->mChunks.empty()) { 1358 List<Chunk>::iterator chunkIt = it->mChunks.begin(); 1359 if (chunkIt->mTimeStampUs < minTimestampUs) { 1360 minTimestampUs = chunkIt->mTimeStampUs; 1361 track = it->mTrack; 1362 } 1363 } 1364 } 1365 1366 if (track == NULL) { 1367 LOGV("Nothing to be written after all"); 1368 return false; 1369 } 1370 1371 if (mIsFirstChunk) { 1372 mIsFirstChunk = false; 1373 } 1374 1375 for (List<ChunkInfo>::iterator it = mChunkInfos.begin(); 1376 it != mChunkInfos.end(); ++it) { 1377 if (it->mTrack == track) { 1378 *chunk = *(it->mChunks.begin()); 1379 it->mChunks.erase(it->mChunks.begin()); 1380 CHECK_EQ(chunk->mTrack, track); 1381 1382 int64_t interChunkTimeUs = 1383 chunk->mTimeStampUs - it->mPrevChunkTimestampUs; 1384 if (interChunkTimeUs > it->mPrevChunkTimestampUs) { 1385 it->mMaxInterChunkDurUs = interChunkTimeUs; 1386 } 1387 1388 return true; 1389 } 1390 } 1391 1392 return false; 1393} 1394 1395void MPEG4Writer::threadFunc() { 1396 LOGV("threadFunc"); 1397 1398 prctl(PR_SET_NAME, (unsigned long)"MPEG4Writer", 0, 0, 0); 1399 1400 Mutex::Autolock autoLock(mLock); 1401 while (!mDone) { 1402 Chunk chunk; 1403 bool chunkFound = false; 1404 1405 while (!mDone && !(chunkFound = findChunkToWrite(&chunk))) { 1406 mChunkReadyCondition.wait(mLock); 1407 } 1408 1409 // Actual write without holding the lock in order to 1410 // reduce the blocking time for media track threads. 1411 if (chunkFound) { 1412 mLock.unlock(); 1413 writeChunkToFile(&chunk); 1414 mLock.lock(); 1415 } 1416 } 1417 1418 writeAllChunks(); 1419} 1420 1421status_t MPEG4Writer::startWriterThread() { 1422 LOGV("startWriterThread"); 1423 1424 mDone = false; 1425 mIsFirstChunk = true; 1426 mDriftTimeUs = 0; 1427 for (List<Track *>::iterator it = mTracks.begin(); 1428 it != mTracks.end(); ++it) { 1429 ChunkInfo info; 1430 info.mTrack = *it; 1431 info.mPrevChunkTimestampUs = 0; 1432 info.mMaxInterChunkDurUs = 0; 1433 mChunkInfos.push_back(info); 1434 } 1435 1436 pthread_attr_t attr; 1437 pthread_attr_init(&attr); 1438 pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE); 1439 pthread_create(&mThread, &attr, ThreadWrapper, this); 1440 pthread_attr_destroy(&attr); 1441 return OK; 1442} 1443 1444 1445status_t MPEG4Writer::Track::start(MetaData *params) { 1446 if (!mDone && mPaused) { 1447 mPaused = false; 1448 mResumed = true; 1449 return OK; 1450 } 1451 1452 int64_t startTimeUs; 1453 if (params == NULL || !params->findInt64(kKeyTime, &startTimeUs)) { 1454 startTimeUs = 0; 1455 } 1456 mStartTimeRealUs = startTimeUs; 1457 1458 int32_t rotationDegrees; 1459 if (!mIsAudio && params && params->findInt32(kKeyRotation, &rotationDegrees)) { 1460 mRotation = rotationDegrees; 1461 } 1462 1463 mIsRealTimeRecording = true; 1464 { 1465 int32_t isNotRealTime; 1466 if (params && params->findInt32(kKeyNotRealTime, &isNotRealTime)) { 1467 mIsRealTimeRecording = (isNotRealTime == 0); 1468 } 1469 } 1470 1471 initTrackingProgressStatus(params); 1472 1473 sp<MetaData> meta = new MetaData; 1474 if (mIsRealTimeRecording && mOwner->numTracks() > 1) { 1475 /* 1476 * This extra delay of accepting incoming audio/video signals 1477 * helps to align a/v start time at the beginning of a recording 1478 * session, and it also helps eliminate the "recording" sound for 1479 * camcorder applications. 1480 * 1481 * If client does not set the start time offset, we fall back to 1482 * use the default initial delay value. 1483 */ 1484 int64_t startTimeOffsetUs = mOwner->getStartTimeOffsetMs() * 1000LL; 1485 if (startTimeOffsetUs < 0) { // Start time offset was not set 1486 startTimeOffsetUs = kInitialDelayTimeUs; 1487 } 1488 startTimeUs += startTimeOffsetUs; 1489 LOGI("Start time offset: %lld us", startTimeOffsetUs); 1490 } 1491 1492 meta->setInt64(kKeyTime, startTimeUs); 1493 1494 status_t err = mSource->start(meta.get()); 1495 if (err != OK) { 1496 mDone = mReachedEOS = true; 1497 return err; 1498 } 1499 1500 pthread_attr_t attr; 1501 pthread_attr_init(&attr); 1502 pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE); 1503 1504 mDone = false; 1505 mStarted = true; 1506 mTrackDurationUs = 0; 1507 mReachedEOS = false; 1508 mEstimatedTrackSizeBytes = 0; 1509 mNumStcoTableEntries = 0; 1510 mNumStssTableEntries = 0; 1511 mNumStscTableEntries = 0; 1512 mNumSttsTableEntries = 0; 1513 mNumCttsTableEntries = 0; 1514 mMdatSizeBytes = 0; 1515 mIsMediaTimeAdjustmentOn = false; 1516 mPrevMediaTimeAdjustTimestampUs = 0; 1517 mMediaTimeAdjustNumFrames = 0; 1518 mPrevMediaTimeAdjustSample = 0; 1519 mTotalDriftTimeToAdjustUs = 0; 1520 mPrevTotalAccumDriftTimeUs = 0; 1521 mMaxChunkDurationUs = 0; 1522 mHasNegativeCttsDeltaDuration = false; 1523 1524 pthread_create(&mThread, &attr, ThreadWrapper, this); 1525 pthread_attr_destroy(&attr); 1526 1527 return OK; 1528} 1529 1530status_t MPEG4Writer::Track::pause() { 1531 mPaused = true; 1532 return OK; 1533} 1534 1535status_t MPEG4Writer::Track::stop() { 1536 LOGD("Stopping %s track", mIsAudio? "Audio": "Video"); 1537 if (!mStarted) { 1538 LOGE("Stop() called but track is not started"); 1539 return ERROR_END_OF_STREAM; 1540 } 1541 1542 if (mDone) { 1543 return OK; 1544 } 1545 mDone = true; 1546 1547 void *dummy; 1548 pthread_join(mThread, &dummy); 1549 1550 status_t err = (status_t) dummy; 1551 1552 LOGD("Stopping %s track source", mIsAudio? "Audio": "Video"); 1553 { 1554 status_t status = mSource->stop(); 1555 if (err == OK && status != OK && status != ERROR_END_OF_STREAM) { 1556 err = status; 1557 } 1558 } 1559 1560 LOGD("%s track stopped", mIsAudio? "Audio": "Video"); 1561 return err; 1562} 1563 1564bool MPEG4Writer::Track::reachedEOS() { 1565 return mReachedEOS; 1566} 1567 1568// static 1569void *MPEG4Writer::Track::ThreadWrapper(void *me) { 1570 Track *track = static_cast<Track *>(me); 1571 1572 status_t err = track->threadEntry(); 1573 return (void *) err; 1574} 1575 1576static void getNalUnitType(uint8_t byte, uint8_t* type) { 1577 LOGV("getNalUnitType: %d", byte); 1578 1579 // nal_unit_type: 5-bit unsigned integer 1580 *type = (byte & 0x1F); 1581} 1582 1583static const uint8_t *findNextStartCode( 1584 const uint8_t *data, size_t length) { 1585 1586 LOGV("findNextStartCode: %p %d", data, length); 1587 1588 size_t bytesLeft = length; 1589 while (bytesLeft > 4 && 1590 memcmp("\x00\x00\x00\x01", &data[length - bytesLeft], 4)) { 1591 --bytesLeft; 1592 } 1593 if (bytesLeft <= 4) { 1594 bytesLeft = 0; // Last parameter set 1595 } 1596 return &data[length - bytesLeft]; 1597} 1598 1599const uint8_t *MPEG4Writer::Track::parseParamSet( 1600 const uint8_t *data, size_t length, int type, size_t *paramSetLen) { 1601 1602 LOGV("parseParamSet"); 1603 CHECK(type == kNalUnitTypeSeqParamSet || 1604 type == kNalUnitTypePicParamSet); 1605 1606 const uint8_t *nextStartCode = findNextStartCode(data, length); 1607 *paramSetLen = nextStartCode - data; 1608 if (*paramSetLen == 0) { 1609 LOGE("Param set is malformed, since its length is 0"); 1610 return NULL; 1611 } 1612 1613 AVCParamSet paramSet(*paramSetLen, data); 1614 if (type == kNalUnitTypeSeqParamSet) { 1615 if (*paramSetLen < 4) { 1616 LOGE("Seq parameter set malformed"); 1617 return NULL; 1618 } 1619 if (mSeqParamSets.empty()) { 1620 mProfileIdc = data[1]; 1621 mProfileCompatible = data[2]; 1622 mLevelIdc = data[3]; 1623 } else { 1624 if (mProfileIdc != data[1] || 1625 mProfileCompatible != data[2] || 1626 mLevelIdc != data[3]) { 1627 LOGE("Inconsistent profile/level found in seq parameter sets"); 1628 return NULL; 1629 } 1630 } 1631 mSeqParamSets.push_back(paramSet); 1632 } else { 1633 mPicParamSets.push_back(paramSet); 1634 } 1635 return nextStartCode; 1636} 1637 1638status_t MPEG4Writer::Track::copyAVCCodecSpecificData( 1639 const uint8_t *data, size_t size) { 1640 LOGV("copyAVCCodecSpecificData"); 1641 1642 // 2 bytes for each of the parameter set length field 1643 // plus the 7 bytes for the header 1644 if (size < 4 + 7) { 1645 LOGE("Codec specific data length too short: %d", size); 1646 return ERROR_MALFORMED; 1647 } 1648 1649 mCodecSpecificDataSize = size; 1650 mCodecSpecificData = malloc(size); 1651 memcpy(mCodecSpecificData, data, size); 1652 return OK; 1653} 1654 1655status_t MPEG4Writer::Track::parseAVCCodecSpecificData( 1656 const uint8_t *data, size_t size) { 1657 1658 LOGV("parseAVCCodecSpecificData"); 1659 // Data starts with a start code. 1660 // SPS and PPS are separated with start codes. 1661 // Also, SPS must come before PPS 1662 uint8_t type = kNalUnitTypeSeqParamSet; 1663 bool gotSps = false; 1664 bool gotPps = false; 1665 const uint8_t *tmp = data; 1666 const uint8_t *nextStartCode = data; 1667 size_t bytesLeft = size; 1668 size_t paramSetLen = 0; 1669 mCodecSpecificDataSize = 0; 1670 while (bytesLeft > 4 && !memcmp("\x00\x00\x00\x01", tmp, 4)) { 1671 getNalUnitType(*(tmp + 4), &type); 1672 if (type == kNalUnitTypeSeqParamSet) { 1673 if (gotPps) { 1674 LOGE("SPS must come before PPS"); 1675 return ERROR_MALFORMED; 1676 } 1677 if (!gotSps) { 1678 gotSps = true; 1679 } 1680 nextStartCode = parseParamSet(tmp + 4, bytesLeft - 4, type, ¶mSetLen); 1681 } else if (type == kNalUnitTypePicParamSet) { 1682 if (!gotSps) { 1683 LOGE("SPS must come before PPS"); 1684 return ERROR_MALFORMED; 1685 } 1686 if (!gotPps) { 1687 gotPps = true; 1688 } 1689 nextStartCode = parseParamSet(tmp + 4, bytesLeft - 4, type, ¶mSetLen); 1690 } else { 1691 LOGE("Only SPS and PPS Nal units are expected"); 1692 return ERROR_MALFORMED; 1693 } 1694 1695 if (nextStartCode == NULL) { 1696 return ERROR_MALFORMED; 1697 } 1698 1699 // Move on to find the next parameter set 1700 bytesLeft -= nextStartCode - tmp; 1701 tmp = nextStartCode; 1702 mCodecSpecificDataSize += (2 + paramSetLen); 1703 } 1704 1705 { 1706 // Check on the number of seq parameter sets 1707 size_t nSeqParamSets = mSeqParamSets.size(); 1708 if (nSeqParamSets == 0) { 1709 LOGE("Cound not find sequence parameter set"); 1710 return ERROR_MALFORMED; 1711 } 1712 1713 if (nSeqParamSets > 0x1F) { 1714 LOGE("Too many seq parameter sets (%d) found", nSeqParamSets); 1715 return ERROR_MALFORMED; 1716 } 1717 } 1718 1719 { 1720 // Check on the number of pic parameter sets 1721 size_t nPicParamSets = mPicParamSets.size(); 1722 if (nPicParamSets == 0) { 1723 LOGE("Cound not find picture parameter set"); 1724 return ERROR_MALFORMED; 1725 } 1726 if (nPicParamSets > 0xFF) { 1727 LOGE("Too many pic parameter sets (%d) found", nPicParamSets); 1728 return ERROR_MALFORMED; 1729 } 1730 } 1731 1732 { 1733 // Check on the profiles 1734 // These profiles requires additional parameter set extensions 1735 if (mProfileIdc == 100 || mProfileIdc == 110 || 1736 mProfileIdc == 122 || mProfileIdc == 144) { 1737 LOGE("Sorry, no support for profile_idc: %d!", mProfileIdc); 1738 return BAD_VALUE; 1739 } 1740 } 1741 1742 return OK; 1743} 1744 1745status_t MPEG4Writer::Track::makeAVCCodecSpecificData( 1746 const uint8_t *data, size_t size) { 1747 1748 if (mCodecSpecificData != NULL) { 1749 LOGE("Already have codec specific data"); 1750 return ERROR_MALFORMED; 1751 } 1752 1753 if (size < 4) { 1754 LOGE("Codec specific data length too short: %d", size); 1755 return ERROR_MALFORMED; 1756 } 1757 1758 // Data is in the form of AVCCodecSpecificData 1759 if (memcmp("\x00\x00\x00\x01", data, 4)) { 1760 return copyAVCCodecSpecificData(data, size); 1761 } 1762 1763 if (parseAVCCodecSpecificData(data, size) != OK) { 1764 return ERROR_MALFORMED; 1765 } 1766 1767 // ISO 14496-15: AVC file format 1768 mCodecSpecificDataSize += 7; // 7 more bytes in the header 1769 mCodecSpecificData = malloc(mCodecSpecificDataSize); 1770 uint8_t *header = (uint8_t *)mCodecSpecificData; 1771 header[0] = 1; // version 1772 header[1] = mProfileIdc; // profile indication 1773 header[2] = mProfileCompatible; // profile compatibility 1774 header[3] = mLevelIdc; 1775 1776 // 6-bit '111111' followed by 2-bit to lengthSizeMinuusOne 1777 if (mOwner->useNalLengthFour()) { 1778 header[4] = 0xfc | 3; // length size == 4 bytes 1779 } else { 1780 header[4] = 0xfc | 1; // length size == 2 bytes 1781 } 1782 1783 // 3-bit '111' followed by 5-bit numSequenceParameterSets 1784 int nSequenceParamSets = mSeqParamSets.size(); 1785 header[5] = 0xe0 | nSequenceParamSets; 1786 header += 6; 1787 for (List<AVCParamSet>::iterator it = mSeqParamSets.begin(); 1788 it != mSeqParamSets.end(); ++it) { 1789 // 16-bit sequence parameter set length 1790 uint16_t seqParamSetLength = it->mLength; 1791 header[0] = seqParamSetLength >> 8; 1792 header[1] = seqParamSetLength & 0xff; 1793 1794 // SPS NAL unit (sequence parameter length bytes) 1795 memcpy(&header[2], it->mData, seqParamSetLength); 1796 header += (2 + seqParamSetLength); 1797 } 1798 1799 // 8-bit nPictureParameterSets 1800 int nPictureParamSets = mPicParamSets.size(); 1801 header[0] = nPictureParamSets; 1802 header += 1; 1803 for (List<AVCParamSet>::iterator it = mPicParamSets.begin(); 1804 it != mPicParamSets.end(); ++it) { 1805 // 16-bit picture parameter set length 1806 uint16_t picParamSetLength = it->mLength; 1807 header[0] = picParamSetLength >> 8; 1808 header[1] = picParamSetLength & 0xff; 1809 1810 // PPS Nal unit (picture parameter set length bytes) 1811 memcpy(&header[2], it->mData, picParamSetLength); 1812 header += (2 + picParamSetLength); 1813 } 1814 1815 return OK; 1816} 1817 1818/* 1819* The video track's media time adjustment for real-time applications 1820* is described as follows: 1821* 1822* First, the media time adjustment is done for every period of 1823* kVideoMediaTimeAdjustPeriodTimeUs. kVideoMediaTimeAdjustPeriodTimeUs 1824* is currently a fixed value chosen heuristically. The value of 1825* kVideoMediaTimeAdjustPeriodTimeUs should not be very large or very small 1826* for two considerations: on one hand, a relatively large value 1827* helps reduce large fluctuation of drift time in the audio encoding 1828* path; while on the other hand, a relatively small value helps keep 1829* restoring synchronization in audio/video more frequently. Note for the 1830* very first period of kVideoMediaTimeAdjustPeriodTimeUs, there is 1831* no media time adjustment for the video track. 1832* 1833* Second, the total accumulated audio track time drift found 1834* in a period of kVideoMediaTimeAdjustPeriodTimeUs is distributed 1835* over a stream of incoming video frames. The number of video frames 1836* affected is determined based on the number of recorded video frames 1837* within the past kVideoMediaTimeAdjustPeriodTimeUs period. 1838* We choose to distribute the drift time over only a portion 1839* (rather than all) of the total number of recorded video frames 1840* in order to make sure that the video track media time adjustment is 1841* completed for the current period before the next video track media 1842* time adjustment period starts. Currently, the portion chosen is a 1843* half (0.5). 1844* 1845* Last, various additional checks are performed to ensure that 1846* the actual audio encoding path does not have too much drift. 1847* In particular, 1) we want to limit the average incremental time 1848* adjustment for each video frame to be less than a threshold 1849* for a single period of kVideoMediaTimeAdjustPeriodTimeUs. 1850* Currently, the threshold is set to 5 ms. If the average incremental 1851* media time adjustment for a video frame is larger than the 1852* threshold, the audio encoding path has too much time drift. 1853* 2) We also want to limit the total time drift in the audio 1854* encoding path to be less than a threshold for a period of 1855* kVideoMediaTimeAdjustPeriodTimeUs. Currently, the threshold 1856* is 0.5% of kVideoMediaTimeAdjustPeriodTimeUs. If the time drift of 1857* the audio encoding path is larger than the threshold, the audio 1858* encoding path has too much time drift. We treat the large time 1859* drift of the audio encoding path as errors, since there is no 1860* way to keep audio/video in synchronization for real-time 1861* applications if the time drift is too large unless we drop some 1862* video frames, which has its own problems that we don't want 1863* to get into for the time being. 1864*/ 1865void MPEG4Writer::Track::adjustMediaTime(int64_t *timestampUs) { 1866 if (*timestampUs - mPrevMediaTimeAdjustTimestampUs >= 1867 kVideoMediaTimeAdjustPeriodTimeUs) { 1868 1869 LOGV("New media time adjustment period at %lld us", *timestampUs); 1870 mIsMediaTimeAdjustmentOn = true; 1871 mMediaTimeAdjustNumFrames = 1872 (mNumSamples - mPrevMediaTimeAdjustSample) >> 1; 1873 1874 mPrevMediaTimeAdjustTimestampUs = *timestampUs; 1875 mPrevMediaTimeAdjustSample = mNumSamples; 1876 int64_t totalAccumDriftTimeUs = mOwner->getDriftTimeUs(); 1877 mTotalDriftTimeToAdjustUs = 1878 totalAccumDriftTimeUs - mPrevTotalAccumDriftTimeUs; 1879 1880 mPrevTotalAccumDriftTimeUs = totalAccumDriftTimeUs; 1881 1882 // Check on incremental adjusted time per frame 1883 int64_t adjustTimePerFrameUs = 1884 mTotalDriftTimeToAdjustUs / mMediaTimeAdjustNumFrames; 1885 1886 if (adjustTimePerFrameUs < 0) { 1887 adjustTimePerFrameUs = -adjustTimePerFrameUs; 1888 } 1889 if (adjustTimePerFrameUs >= 5000) { 1890 LOGE("Adjusted time per video frame is %lld us", 1891 adjustTimePerFrameUs); 1892 CHECK(!"Video frame time adjustment is too large!"); 1893 } 1894 1895 // Check on total accumulated time drift within a period of 1896 // kVideoMediaTimeAdjustPeriodTimeUs. 1897 int64_t driftPercentage = (mTotalDriftTimeToAdjustUs * 1000) 1898 / kVideoMediaTimeAdjustPeriodTimeUs; 1899 1900 if (driftPercentage < 0) { 1901 driftPercentage = -driftPercentage; 1902 } 1903 if (driftPercentage > 5) { 1904 LOGE("Audio track has time drift %lld us over %lld us", 1905 mTotalDriftTimeToAdjustUs, 1906 kVideoMediaTimeAdjustPeriodTimeUs); 1907 1908 CHECK(!"The audio track media time drifts too much!"); 1909 } 1910 1911 } 1912 1913 if (mIsMediaTimeAdjustmentOn) { 1914 if (mNumSamples - mPrevMediaTimeAdjustSample <= 1915 mMediaTimeAdjustNumFrames) { 1916 1917 // Do media time incremental adjustment 1918 int64_t incrementalAdjustTimeUs = 1919 (mTotalDriftTimeToAdjustUs * 1920 (mNumSamples - mPrevMediaTimeAdjustSample)) 1921 / mMediaTimeAdjustNumFrames; 1922 1923 *timestampUs += 1924 (incrementalAdjustTimeUs + mPrevTotalAccumDriftTimeUs); 1925 1926 LOGV("Incremental video frame media time adjustment: %lld us", 1927 (incrementalAdjustTimeUs + mPrevTotalAccumDriftTimeUs)); 1928 } else { 1929 // Within the remaining adjustment period, 1930 // no incremental adjustment is needed. 1931 *timestampUs += 1932 (mTotalDriftTimeToAdjustUs + mPrevTotalAccumDriftTimeUs); 1933 1934 LOGV("Fixed video frame media time adjustment: %lld us", 1935 (mTotalDriftTimeToAdjustUs + mPrevTotalAccumDriftTimeUs)); 1936 } 1937 } 1938} 1939 1940/* 1941 * Updates the drift time from the audio track so that 1942 * the video track can get the updated drift time information 1943 * from the file writer. The fluctuation of the drift time of the audio 1944 * encoding path is smoothed out with a simple filter by giving a larger 1945 * weight to more recently drift time. The filter coefficients, 0.5 and 0.5, 1946 * are heuristically determined. 1947 */ 1948void MPEG4Writer::Track::updateDriftTime(const sp<MetaData>& meta) { 1949 int64_t driftTimeUs = 0; 1950 if (meta->findInt64(kKeyDriftTime, &driftTimeUs)) { 1951 int64_t prevDriftTimeUs = mOwner->getDriftTimeUs(); 1952 int64_t timeUs = (driftTimeUs + prevDriftTimeUs) >> 1; 1953 mOwner->setDriftTimeUs(timeUs); 1954 } 1955} 1956 1957status_t MPEG4Writer::Track::threadEntry() { 1958 int32_t count = 0; 1959 const int64_t interleaveDurationUs = mOwner->interleaveDuration(); 1960 const bool hasMultipleTracks = (mOwner->numTracks() > 1); 1961 int64_t chunkTimestampUs = 0; 1962 int32_t nChunks = 0; 1963 int32_t nZeroLengthFrames = 0; 1964 int64_t lastTimestampUs = 0; // Previous sample time stamp 1965 int64_t lastCttsTimeUs = 0; // Previous sample time stamp 1966 int64_t lastDurationUs = 0; // Between the previous two samples 1967 int64_t currDurationTicks = 0; // Timescale based ticks 1968 int64_t lastDurationTicks = 0; // Timescale based ticks 1969 int32_t sampleCount = 1; // Sample count in the current stts table entry 1970 int64_t currCttsDurTicks = 0; // Timescale based ticks 1971 int64_t lastCttsDurTicks = 0; // Timescale based ticks 1972 int32_t cttsSampleCount = 1; // Sample count in the current ctts table entry 1973 uint32_t previousSampleSize = 0; // Size of the previous sample 1974 int64_t previousPausedDurationUs = 0; 1975 int64_t timestampUs = 0; 1976 int64_t cttsDeltaTimeUs = 0; 1977 1978 if (mIsAudio) { 1979 prctl(PR_SET_NAME, (unsigned long)"AudioTrackEncoding", 0, 0, 0); 1980 } else { 1981 prctl(PR_SET_NAME, (unsigned long)"VideoTrackEncoding", 0, 0, 0); 1982 } 1983 androidSetThreadPriority(0, ANDROID_PRIORITY_AUDIO); 1984 1985 sp<MetaData> meta_data; 1986 1987 mNumSamples = 0; 1988 status_t err = OK; 1989 MediaBuffer *buffer; 1990 while (!mDone && (err = mSource->read(&buffer)) == OK) { 1991 if (buffer->range_length() == 0) { 1992 buffer->release(); 1993 buffer = NULL; 1994 ++nZeroLengthFrames; 1995 continue; 1996 } 1997 1998 // If the codec specific data has not been received yet, delay pause. 1999 // After the codec specific data is received, discard what we received 2000 // when the track is to be paused. 2001 if (mPaused && !mResumed) { 2002 buffer->release(); 2003 buffer = NULL; 2004 continue; 2005 } 2006 2007 ++count; 2008 2009 int32_t isCodecConfig; 2010 if (buffer->meta_data()->findInt32(kKeyIsCodecConfig, &isCodecConfig) 2011 && isCodecConfig) { 2012 CHECK(!mGotAllCodecSpecificData); 2013 2014 if (mIsAvc) { 2015 status_t err = makeAVCCodecSpecificData( 2016 (const uint8_t *)buffer->data() 2017 + buffer->range_offset(), 2018 buffer->range_length()); 2019 CHECK_EQ(OK, err); 2020 } else if (mIsMPEG4) { 2021 mCodecSpecificDataSize = buffer->range_length(); 2022 mCodecSpecificData = malloc(mCodecSpecificDataSize); 2023 memcpy(mCodecSpecificData, 2024 (const uint8_t *)buffer->data() 2025 + buffer->range_offset(), 2026 buffer->range_length()); 2027 } 2028 2029 buffer->release(); 2030 buffer = NULL; 2031 2032 mGotAllCodecSpecificData = true; 2033 continue; 2034 } 2035 2036 // Make a deep copy of the MediaBuffer and Metadata and release 2037 // the original as soon as we can 2038 MediaBuffer *copy = new MediaBuffer(buffer->range_length()); 2039 memcpy(copy->data(), (uint8_t *)buffer->data() + buffer->range_offset(), 2040 buffer->range_length()); 2041 copy->set_range(0, buffer->range_length()); 2042 meta_data = new MetaData(*buffer->meta_data().get()); 2043 buffer->release(); 2044 buffer = NULL; 2045 2046 if (mIsAvc) StripStartcode(copy); 2047 2048 size_t sampleSize = copy->range_length(); 2049 if (mIsAvc) { 2050 if (mOwner->useNalLengthFour()) { 2051 sampleSize += 4; 2052 } else { 2053 sampleSize += 2; 2054 } 2055 } 2056 2057 // Max file size or duration handling 2058 mMdatSizeBytes += sampleSize; 2059 updateTrackSizeEstimate(); 2060 2061 if (mOwner->exceedsFileSizeLimit()) { 2062 mOwner->notify(MEDIA_RECORDER_EVENT_INFO, MEDIA_RECORDER_INFO_MAX_FILESIZE_REACHED, 0); 2063 break; 2064 } 2065 if (mOwner->exceedsFileDurationLimit()) { 2066 mOwner->notify(MEDIA_RECORDER_EVENT_INFO, MEDIA_RECORDER_INFO_MAX_DURATION_REACHED, 0); 2067 break; 2068 } 2069 2070 2071 int32_t isSync = false; 2072 meta_data->findInt32(kKeyIsSyncFrame, &isSync); 2073 2074 /* 2075 * The original timestamp found in the data buffer will be modified as below: 2076 * 2077 * There is a playback offset into this track if the track's start time 2078 * is not the same as the movie start time, which will be recorded in edst 2079 * box of the output file. The playback offset is to make sure that the 2080 * starting time of the audio/video tracks are synchronized. Although the 2081 * track's media timestamp may be subject to various modifications 2082 * as outlined below, the track's playback offset time remains unchanged 2083 * once the first data buffer of the track is received. 2084 * 2085 * The media time stamp will be calculated by subtracting the playback offset 2086 * (and potential pause durations) from the original timestamp in the buffer. 2087 * 2088 * If this track is a video track for a real-time recording application with 2089 * both audio and video tracks, its media timestamp will subject to further 2090 * modification based on the media clock of the audio track. This modification 2091 * is needed for the purpose of maintaining good audio/video synchronization. 2092 * 2093 * If the recording session is paused and resumed multiple times, the track 2094 * media timestamp will be modified as if the recording session had never been 2095 * paused at all during playback of the recorded output file. In other words, 2096 * the output file will have no memory of pause/resume durations. 2097 * 2098 */ 2099 CHECK(meta_data->findInt64(kKeyTime, ×tampUs)); 2100 2101//////////////////////////////////////////////////////////////////////////////// 2102 if (mNumSamples == 0) { 2103 mFirstSampleTimeRealUs = systemTime() / 1000; 2104 mStartTimestampUs = timestampUs; 2105 mOwner->setStartTimestampUs(mStartTimestampUs); 2106 previousPausedDurationUs = mStartTimestampUs; 2107 } 2108 2109 if (mResumed) { 2110 int64_t durExcludingEarlierPausesUs = timestampUs - previousPausedDurationUs; 2111 CHECK(durExcludingEarlierPausesUs >= 0); 2112 int64_t pausedDurationUs = durExcludingEarlierPausesUs - mTrackDurationUs; 2113 CHECK(pausedDurationUs >= lastDurationUs); 2114 previousPausedDurationUs += pausedDurationUs - lastDurationUs; 2115 mResumed = false; 2116 } 2117 2118 timestampUs -= previousPausedDurationUs; 2119 CHECK(timestampUs >= 0); 2120 if (!mIsAudio) { 2121 /* 2122 * Composition time: timestampUs 2123 * Decoding time: decodingTimeUs 2124 * Composition time delta = composition time - decoding time 2125 * 2126 * We save picture decoding time stamp delta in stts table entries, 2127 * and composition time delta duration in ctts table entries. 2128 */ 2129 int64_t decodingTimeUs; 2130 CHECK(meta_data->findInt64(kKeyDecodingTime, &decodingTimeUs)); 2131 decodingTimeUs -= previousPausedDurationUs; 2132 int64_t timeUs = decodingTimeUs; 2133 cttsDeltaTimeUs = timestampUs - decodingTimeUs; 2134 timestampUs = decodingTimeUs; 2135 LOGV("decoding time: %lld and ctts delta time: %lld", 2136 timestampUs, cttsDeltaTimeUs); 2137 } 2138 2139 // Media time adjustment for real-time applications 2140 if (mIsRealTimeRecording) { 2141 if (mIsAudio) { 2142 updateDriftTime(meta_data); 2143 } else { 2144 adjustMediaTime(×tampUs); 2145 } 2146 } 2147 2148 CHECK(timestampUs >= 0); 2149 if (mNumSamples > 1) { 2150 if (timestampUs <= lastTimestampUs) { 2151 LOGW("Frame arrives too late!"); 2152 // Don't drop the late frame, since dropping a frame may cause 2153 // problems later during playback 2154 2155 // The idea here is to avoid having two or more samples with the 2156 // same timestamp in the output file. 2157 if (mTimeScale >= 1000000LL) { 2158 timestampUs = lastTimestampUs + 1; 2159 } else { 2160 timestampUs = lastTimestampUs + (1000000LL + (mTimeScale >> 1)) / mTimeScale; 2161 } 2162 } 2163 } 2164 2165 LOGV("%s media time stamp: %lld and previous paused duration %lld", 2166 mIsAudio? "Audio": "Video", timestampUs, previousPausedDurationUs); 2167 if (timestampUs > mTrackDurationUs) { 2168 mTrackDurationUs = timestampUs; 2169 } 2170 2171 mSampleSizes.push_back(sampleSize); 2172 ++mNumSamples; 2173 if (mNumSamples > 2) { 2174 // We need to use the time scale based ticks, rather than the 2175 // timestamp itself to determine whether we have to use a new 2176 // stts entry, since we may have rounding errors. 2177 // The calculation is intended to reduce the accumulated 2178 // rounding errors. 2179 currDurationTicks = 2180 ((timestampUs * mTimeScale + 500000LL) / 1000000LL - 2181 (lastTimestampUs * mTimeScale + 500000LL) / 1000000LL); 2182 2183 // Force the first sample to have its own stts entry so that 2184 // we can adjust its value later to maintain the A/V sync. 2185 if (mNumSamples == 3 || currDurationTicks != lastDurationTicks) { 2186 LOGV("%s lastDurationUs: %lld us, currDurationTicks: %lld us", 2187 mIsAudio? "Audio": "Video", lastDurationUs, currDurationTicks); 2188 addOneSttsTableEntry(sampleCount, lastDurationTicks); 2189 sampleCount = 1; 2190 } else { 2191 ++sampleCount; 2192 } 2193 2194 if (!mIsAudio) { 2195 currCttsDurTicks = 2196 ((cttsDeltaTimeUs * mTimeScale + 500000LL) / 1000000LL - 2197 (lastCttsTimeUs * mTimeScale + 500000LL) / 1000000LL); 2198 if (currCttsDurTicks != lastCttsDurTicks) { 2199 addOneCttsTableEntry(cttsSampleCount, lastCttsDurTicks); 2200 cttsSampleCount = 1; 2201 } else { 2202 ++cttsSampleCount; 2203 } 2204 } 2205 } 2206 if (mSamplesHaveSameSize) { 2207 if (mNumSamples >= 2 && previousSampleSize != sampleSize) { 2208 mSamplesHaveSameSize = false; 2209 } 2210 previousSampleSize = sampleSize; 2211 } 2212 LOGV("%s timestampUs/lastTimestampUs: %lld/%lld", 2213 mIsAudio? "Audio": "Video", timestampUs, lastTimestampUs); 2214 lastDurationUs = timestampUs - lastTimestampUs; 2215 lastDurationTicks = currDurationTicks; 2216 lastTimestampUs = timestampUs; 2217 2218 if (!mIsAudio) { 2219 lastCttsDurTicks = currCttsDurTicks; 2220 lastCttsTimeUs = cttsDeltaTimeUs; 2221 } 2222 2223 if (isSync != 0) { 2224 addOneStssTableEntry(mNumSamples); 2225 } 2226 2227 if (mTrackingProgressStatus) { 2228 if (mPreviousTrackTimeUs <= 0) { 2229 mPreviousTrackTimeUs = mStartTimestampUs; 2230 } 2231 trackProgressStatus(timestampUs); 2232 } 2233 if (!hasMultipleTracks) { 2234 off64_t offset = mIsAvc? mOwner->addLengthPrefixedSample_l(copy) 2235 : mOwner->addSample_l(copy); 2236 if (mChunkOffsets.empty()) { 2237 addChunkOffset(offset); 2238 } 2239 copy->release(); 2240 copy = NULL; 2241 continue; 2242 } 2243 2244 mChunkSamples.push_back(copy); 2245 if (interleaveDurationUs == 0) { 2246 addOneStscTableEntry(++nChunks, 1); 2247 bufferChunk(timestampUs); 2248 } else { 2249 if (chunkTimestampUs == 0) { 2250 chunkTimestampUs = timestampUs; 2251 } else { 2252 int64_t chunkDurationUs = timestampUs - chunkTimestampUs; 2253 if (chunkDurationUs > interleaveDurationUs) { 2254 if (chunkDurationUs > mMaxChunkDurationUs) { 2255 mMaxChunkDurationUs = chunkDurationUs; 2256 } 2257 ++nChunks; 2258 if (nChunks == 1 || // First chunk 2259 (--(mStscTableEntries.end()))->samplesPerChunk != 2260 mChunkSamples.size()) { 2261 addOneStscTableEntry(nChunks, mChunkSamples.size()); 2262 } 2263 bufferChunk(timestampUs); 2264 chunkTimestampUs = timestampUs; 2265 } 2266 } 2267 } 2268 2269 } 2270 2271 if (mSampleSizes.empty() || // no samples written 2272 (!mIsAudio && mNumStssTableEntries == 0) || // no sync frames for video 2273 (OK != checkCodecSpecificData())) { // no codec specific data 2274 err = ERROR_MALFORMED; 2275 } 2276 mOwner->trackProgressStatus(mTrackId, -1, err); 2277 2278 // Last chunk 2279 if (!hasMultipleTracks) { 2280 addOneStscTableEntry(1, mNumSamples); 2281 } else if (!mChunkSamples.empty()) { 2282 addOneStscTableEntry(++nChunks, mChunkSamples.size()); 2283 bufferChunk(timestampUs); 2284 } 2285 2286 // We don't really know how long the last frame lasts, since 2287 // there is no frame time after it, just repeat the previous 2288 // frame's duration. 2289 if (mNumSamples == 1) { 2290 lastDurationUs = 0; // A single sample's duration 2291 lastDurationTicks = 0; 2292 lastCttsDurTicks = 0; 2293 } else { 2294 ++sampleCount; // Count for the last sample 2295 ++cttsSampleCount; 2296 } 2297 2298 if (mNumSamples <= 2) { 2299 addOneSttsTableEntry(1, lastDurationTicks); 2300 if (sampleCount - 1 > 0) { 2301 addOneSttsTableEntry(sampleCount - 1, lastDurationTicks); 2302 } 2303 } else { 2304 addOneSttsTableEntry(sampleCount, lastDurationTicks); 2305 } 2306 2307 addOneCttsTableEntry(cttsSampleCount, lastCttsDurTicks); 2308 mTrackDurationUs += lastDurationUs; 2309 mReachedEOS = true; 2310 2311 sendTrackSummary(hasMultipleTracks); 2312 2313 LOGI("Received total/0-length (%d/%d) buffers and encoded %d frames. - %s", 2314 count, nZeroLengthFrames, mNumSamples, mIsAudio? "audio": "video"); 2315 if (mIsAudio) { 2316 LOGI("Audio track drift time: %lld us", mOwner->getDriftTimeUs()); 2317 } 2318 2319 if (err == ERROR_END_OF_STREAM) { 2320 return OK; 2321 } 2322 return err; 2323} 2324 2325void MPEG4Writer::Track::sendTrackSummary(bool hasMultipleTracks) { 2326 2327 // Send track summary only if test mode is enabled. 2328 if (!isTestModeEnabled()) { 2329 return; 2330 } 2331 2332 int trackNum = (mTrackId << 28); 2333 2334 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2335 trackNum | MEDIA_RECORDER_TRACK_INFO_TYPE, 2336 mIsAudio? 0: 1); 2337 2338 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2339 trackNum | MEDIA_RECORDER_TRACK_INFO_DURATION_MS, 2340 mTrackDurationUs / 1000); 2341 2342 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2343 trackNum | MEDIA_RECORDER_TRACK_INFO_ENCODED_FRAMES, 2344 mNumSamples); 2345 2346 { 2347 // The system delay time excluding the requested initial delay that 2348 // is used to eliminate the recording sound. 2349 int64_t startTimeOffsetUs = mOwner->getStartTimeOffsetMs() * 1000LL; 2350 if (startTimeOffsetUs < 0) { // Start time offset was not set 2351 startTimeOffsetUs = kInitialDelayTimeUs; 2352 } 2353 int64_t initialDelayUs = 2354 mFirstSampleTimeRealUs - mStartTimeRealUs - startTimeOffsetUs; 2355 2356 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2357 trackNum | MEDIA_RECORDER_TRACK_INFO_INITIAL_DELAY_MS, 2358 (initialDelayUs) / 1000); 2359 } 2360 2361 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2362 trackNum | MEDIA_RECORDER_TRACK_INFO_DATA_KBYTES, 2363 mMdatSizeBytes / 1024); 2364 2365 if (hasMultipleTracks) { 2366 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2367 trackNum | MEDIA_RECORDER_TRACK_INFO_MAX_CHUNK_DUR_MS, 2368 mMaxChunkDurationUs / 1000); 2369 2370 int64_t moovStartTimeUs = mOwner->getStartTimestampUs(); 2371 if (mStartTimestampUs != moovStartTimeUs) { 2372 int64_t startTimeOffsetUs = mStartTimestampUs - moovStartTimeUs; 2373 mOwner->notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2374 trackNum | MEDIA_RECORDER_TRACK_INFO_START_OFFSET_MS, 2375 startTimeOffsetUs / 1000); 2376 } 2377 } 2378} 2379 2380void MPEG4Writer::Track::trackProgressStatus(int64_t timeUs, status_t err) { 2381 LOGV("trackProgressStatus: %lld us", timeUs); 2382 if (mTrackEveryTimeDurationUs > 0 && 2383 timeUs - mPreviousTrackTimeUs >= mTrackEveryTimeDurationUs) { 2384 LOGV("Fire time tracking progress status at %lld us", timeUs); 2385 mOwner->trackProgressStatus(mTrackId, timeUs - mPreviousTrackTimeUs, err); 2386 mPreviousTrackTimeUs = timeUs; 2387 } 2388} 2389 2390void MPEG4Writer::trackProgressStatus( 2391 size_t trackId, int64_t timeUs, status_t err) { 2392 Mutex::Autolock lock(mLock); 2393 int32_t trackNum = (trackId << 28); 2394 2395 // Error notification 2396 // Do not consider ERROR_END_OF_STREAM an error 2397 if (err != OK && err != ERROR_END_OF_STREAM) { 2398 notify(MEDIA_RECORDER_TRACK_EVENT_ERROR, 2399 trackNum | MEDIA_RECORDER_TRACK_ERROR_GENERAL, 2400 err); 2401 return; 2402 } 2403 2404 if (timeUs == -1) { 2405 // Send completion notification 2406 notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2407 trackNum | MEDIA_RECORDER_TRACK_INFO_COMPLETION_STATUS, 2408 err); 2409 } else { 2410 // Send progress status 2411 notify(MEDIA_RECORDER_TRACK_EVENT_INFO, 2412 trackNum | MEDIA_RECORDER_TRACK_INFO_PROGRESS_IN_TIME, 2413 timeUs / 1000); 2414 } 2415} 2416 2417void MPEG4Writer::setDriftTimeUs(int64_t driftTimeUs) { 2418 LOGV("setDriftTimeUs: %lld us", driftTimeUs); 2419 Mutex::Autolock autolock(mLock); 2420 mDriftTimeUs = driftTimeUs; 2421} 2422 2423int64_t MPEG4Writer::getDriftTimeUs() { 2424 LOGV("getDriftTimeUs: %lld us", mDriftTimeUs); 2425 Mutex::Autolock autolock(mLock); 2426 return mDriftTimeUs; 2427} 2428 2429bool MPEG4Writer::useNalLengthFour() { 2430 return mUse4ByteNalLength; 2431} 2432 2433void MPEG4Writer::Track::bufferChunk(int64_t timestampUs) { 2434 LOGV("bufferChunk"); 2435 2436 Chunk chunk(this, timestampUs, mChunkSamples); 2437 mOwner->bufferChunk(chunk); 2438 mChunkSamples.clear(); 2439} 2440 2441int64_t MPEG4Writer::Track::getDurationUs() const { 2442 return mTrackDurationUs; 2443} 2444 2445int64_t MPEG4Writer::Track::getEstimatedTrackSizeBytes() const { 2446 return mEstimatedTrackSizeBytes; 2447} 2448 2449status_t MPEG4Writer::Track::checkCodecSpecificData() const { 2450 const char *mime; 2451 CHECK(mMeta->findCString(kKeyMIMEType, &mime)); 2452 if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AAC, mime) || 2453 !strcasecmp(MEDIA_MIMETYPE_VIDEO_MPEG4, mime) || 2454 !strcasecmp(MEDIA_MIMETYPE_VIDEO_AVC, mime)) { 2455 if (!mCodecSpecificData || 2456 mCodecSpecificDataSize <= 0) { 2457 LOGE("Missing codec specific data"); 2458 return ERROR_MALFORMED; 2459 } 2460 } else { 2461 if (mCodecSpecificData || 2462 mCodecSpecificDataSize > 0) { 2463 LOGE("Unexepected codec specific data found"); 2464 return ERROR_MALFORMED; 2465 } 2466 } 2467 return OK; 2468} 2469 2470void MPEG4Writer::Track::writeTrackHeader(bool use32BitOffset) { 2471 2472 LOGV("%s track time scale: %d", 2473 mIsAudio? "Audio": "Video", mTimeScale); 2474 2475 time_t now = time(NULL); 2476 mOwner->beginBox("trak"); 2477 writeTkhdBox(now); 2478 mOwner->beginBox("mdia"); 2479 writeMdhdBox(now); 2480 writeHdlrBox(); 2481 mOwner->beginBox("minf"); 2482 if (mIsAudio) { 2483 writeSmhdBox(); 2484 } else { 2485 writeVmhdBox(); 2486 } 2487 writeDinfBox(); 2488 writeStblBox(use32BitOffset); 2489 mOwner->endBox(); // minf 2490 mOwner->endBox(); // mdia 2491 mOwner->endBox(); // trak 2492} 2493 2494void MPEG4Writer::Track::writeStblBox(bool use32BitOffset) { 2495 mOwner->beginBox("stbl"); 2496 mOwner->beginBox("stsd"); 2497 mOwner->writeInt32(0); // version=0, flags=0 2498 mOwner->writeInt32(1); // entry count 2499 if (mIsAudio) { 2500 writeAudioFourCCBox(); 2501 } else { 2502 writeVideoFourCCBox(); 2503 } 2504 mOwner->endBox(); // stsd 2505 writeSttsBox(); 2506 writeCttsBox(); 2507 if (!mIsAudio) { 2508 writeStssBox(); 2509 } 2510 writeStszBox(); 2511 writeStscBox(); 2512 writeStcoBox(use32BitOffset); 2513 mOwner->endBox(); // stbl 2514} 2515 2516void MPEG4Writer::Track::writeVideoFourCCBox() { 2517 const char *mime; 2518 bool success = mMeta->findCString(kKeyMIMEType, &mime); 2519 CHECK(success); 2520 if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_MPEG4, mime)) { 2521 mOwner->beginBox("mp4v"); 2522 } else if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_H263, mime)) { 2523 mOwner->beginBox("s263"); 2524 } else if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_AVC, mime)) { 2525 mOwner->beginBox("avc1"); 2526 } else { 2527 LOGE("Unknown mime type '%s'.", mime); 2528 CHECK(!"should not be here, unknown mime type."); 2529 } 2530 2531 mOwner->writeInt32(0); // reserved 2532 mOwner->writeInt16(0); // reserved 2533 mOwner->writeInt16(1); // data ref index 2534 mOwner->writeInt16(0); // predefined 2535 mOwner->writeInt16(0); // reserved 2536 mOwner->writeInt32(0); // predefined 2537 mOwner->writeInt32(0); // predefined 2538 mOwner->writeInt32(0); // predefined 2539 2540 int32_t width, height; 2541 success = mMeta->findInt32(kKeyWidth, &width); 2542 success = success && mMeta->findInt32(kKeyHeight, &height); 2543 CHECK(success); 2544 2545 mOwner->writeInt16(width); 2546 mOwner->writeInt16(height); 2547 mOwner->writeInt32(0x480000); // horiz resolution 2548 mOwner->writeInt32(0x480000); // vert resolution 2549 mOwner->writeInt32(0); // reserved 2550 mOwner->writeInt16(1); // frame count 2551 mOwner->write(" ", 32); 2552 mOwner->writeInt16(0x18); // depth 2553 mOwner->writeInt16(-1); // predefined 2554 2555 CHECK(23 + mCodecSpecificDataSize < 128); 2556 2557 if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_MPEG4, mime)) { 2558 writeMp4vEsdsBox(); 2559 } else if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_H263, mime)) { 2560 writeD263Box(); 2561 } else if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_AVC, mime)) { 2562 writeAvccBox(); 2563 } 2564 2565 writePaspBox(); 2566 mOwner->endBox(); // mp4v, s263 or avc1 2567} 2568 2569void MPEG4Writer::Track::writeAudioFourCCBox() { 2570 const char *mime; 2571 bool success = mMeta->findCString(kKeyMIMEType, &mime); 2572 CHECK(success); 2573 const char *fourcc = NULL; 2574 if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_NB, mime)) { 2575 fourcc = "samr"; 2576 } else if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_WB, mime)) { 2577 fourcc = "sawb"; 2578 } else if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AAC, mime)) { 2579 fourcc = "mp4a"; 2580 } else { 2581 LOGE("Unknown mime type '%s'.", mime); 2582 CHECK(!"should not be here, unknown mime type."); 2583 } 2584 2585 mOwner->beginBox(fourcc); // audio format 2586 mOwner->writeInt32(0); // reserved 2587 mOwner->writeInt16(0); // reserved 2588 mOwner->writeInt16(0x1); // data ref index 2589 mOwner->writeInt32(0); // reserved 2590 mOwner->writeInt32(0); // reserved 2591 int32_t nChannels; 2592 CHECK_EQ(true, mMeta->findInt32(kKeyChannelCount, &nChannels)); 2593 mOwner->writeInt16(nChannels); // channel count 2594 mOwner->writeInt16(16); // sample size 2595 mOwner->writeInt16(0); // predefined 2596 mOwner->writeInt16(0); // reserved 2597 2598 int32_t samplerate; 2599 success = mMeta->findInt32(kKeySampleRate, &samplerate); 2600 CHECK(success); 2601 mOwner->writeInt32(samplerate << 16); 2602 if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AAC, mime)) { 2603 writeMp4aEsdsBox(); 2604 } else if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_NB, mime) || 2605 !strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_WB, mime)) { 2606 writeDamrBox(); 2607 } 2608 mOwner->endBox(); 2609} 2610 2611void MPEG4Writer::Track::writeMp4aEsdsBox() { 2612 mOwner->beginBox("esds"); 2613 CHECK(mCodecSpecificData); 2614 CHECK(mCodecSpecificDataSize > 0); 2615 2616 // Make sure all sizes encode to a single byte. 2617 CHECK(mCodecSpecificDataSize + 23 < 128); 2618 2619 mOwner->writeInt32(0); // version=0, flags=0 2620 mOwner->writeInt8(0x03); // ES_DescrTag 2621 mOwner->writeInt8(23 + mCodecSpecificDataSize); 2622 mOwner->writeInt16(0x0000);// ES_ID 2623 mOwner->writeInt8(0x00); 2624 2625 mOwner->writeInt8(0x04); // DecoderConfigDescrTag 2626 mOwner->writeInt8(15 + mCodecSpecificDataSize); 2627 mOwner->writeInt8(0x40); // objectTypeIndication ISO/IEC 14492-2 2628 mOwner->writeInt8(0x15); // streamType AudioStream 2629 2630 mOwner->writeInt16(0x03); // XXX 2631 mOwner->writeInt8(0x00); // buffer size 24-bit 2632 mOwner->writeInt32(96000); // max bit rate 2633 mOwner->writeInt32(96000); // avg bit rate 2634 2635 mOwner->writeInt8(0x05); // DecoderSpecificInfoTag 2636 mOwner->writeInt8(mCodecSpecificDataSize); 2637 mOwner->write(mCodecSpecificData, mCodecSpecificDataSize); 2638 2639 static const uint8_t kData2[] = { 2640 0x06, // SLConfigDescriptorTag 2641 0x01, 2642 0x02 2643 }; 2644 mOwner->write(kData2, sizeof(kData2)); 2645 2646 mOwner->endBox(); // esds 2647} 2648 2649void MPEG4Writer::Track::writeMp4vEsdsBox() { 2650 CHECK(mCodecSpecificData); 2651 CHECK(mCodecSpecificDataSize > 0); 2652 mOwner->beginBox("esds"); 2653 2654 mOwner->writeInt32(0); // version=0, flags=0 2655 2656 mOwner->writeInt8(0x03); // ES_DescrTag 2657 mOwner->writeInt8(23 + mCodecSpecificDataSize); 2658 mOwner->writeInt16(0x0000); // ES_ID 2659 mOwner->writeInt8(0x1f); 2660 2661 mOwner->writeInt8(0x04); // DecoderConfigDescrTag 2662 mOwner->writeInt8(15 + mCodecSpecificDataSize); 2663 mOwner->writeInt8(0x20); // objectTypeIndication ISO/IEC 14492-2 2664 mOwner->writeInt8(0x11); // streamType VisualStream 2665 2666 static const uint8_t kData[] = { 2667 0x01, 0x77, 0x00, 2668 0x00, 0x03, 0xe8, 0x00, 2669 0x00, 0x03, 0xe8, 0x00 2670 }; 2671 mOwner->write(kData, sizeof(kData)); 2672 2673 mOwner->writeInt8(0x05); // DecoderSpecificInfoTag 2674 2675 mOwner->writeInt8(mCodecSpecificDataSize); 2676 mOwner->write(mCodecSpecificData, mCodecSpecificDataSize); 2677 2678 static const uint8_t kData2[] = { 2679 0x06, // SLConfigDescriptorTag 2680 0x01, 2681 0x02 2682 }; 2683 mOwner->write(kData2, sizeof(kData2)); 2684 2685 mOwner->endBox(); // esds 2686} 2687 2688void MPEG4Writer::Track::writeTkhdBox(time_t now) { 2689 mOwner->beginBox("tkhd"); 2690 // Flags = 7 to indicate that the track is enabled, and 2691 // part of the presentation 2692 mOwner->writeInt32(0x07); // version=0, flags=7 2693 mOwner->writeInt32(now); // creation time 2694 mOwner->writeInt32(now); // modification time 2695 mOwner->writeInt32(mTrackId + 1); // track id starts with 1 2696 mOwner->writeInt32(0); // reserved 2697 int64_t trakDurationUs = getDurationUs(); 2698 int32_t mvhdTimeScale = mOwner->getTimeScale(); 2699 int32_t tkhdDuration = 2700 (trakDurationUs * mvhdTimeScale + 5E5) / 1E6; 2701 mOwner->writeInt32(tkhdDuration); // in mvhd timescale 2702 mOwner->writeInt32(0); // reserved 2703 mOwner->writeInt32(0); // reserved 2704 mOwner->writeInt16(0); // layer 2705 mOwner->writeInt16(0); // alternate group 2706 mOwner->writeInt16(mIsAudio ? 0x100 : 0); // volume 2707 mOwner->writeInt16(0); // reserved 2708 2709 mOwner->writeCompositionMatrix(mRotation); // matrix 2710 2711 if (mIsAudio) { 2712 mOwner->writeInt32(0); 2713 mOwner->writeInt32(0); 2714 } else { 2715 int32_t width, height; 2716 bool success = mMeta->findInt32(kKeyWidth, &width); 2717 success = success && mMeta->findInt32(kKeyHeight, &height); 2718 CHECK(success); 2719 2720 mOwner->writeInt32(width << 16); // 32-bit fixed-point value 2721 mOwner->writeInt32(height << 16); // 32-bit fixed-point value 2722 } 2723 mOwner->endBox(); // tkhd 2724} 2725 2726void MPEG4Writer::Track::writeVmhdBox() { 2727 mOwner->beginBox("vmhd"); 2728 mOwner->writeInt32(0x01); // version=0, flags=1 2729 mOwner->writeInt16(0); // graphics mode 2730 mOwner->writeInt16(0); // opcolor 2731 mOwner->writeInt16(0); 2732 mOwner->writeInt16(0); 2733 mOwner->endBox(); 2734} 2735 2736void MPEG4Writer::Track::writeSmhdBox() { 2737 mOwner->beginBox("smhd"); 2738 mOwner->writeInt32(0); // version=0, flags=0 2739 mOwner->writeInt16(0); // balance 2740 mOwner->writeInt16(0); // reserved 2741 mOwner->endBox(); 2742} 2743 2744void MPEG4Writer::Track::writeHdlrBox() { 2745 mOwner->beginBox("hdlr"); 2746 mOwner->writeInt32(0); // version=0, flags=0 2747 mOwner->writeInt32(0); // component type: should be mhlr 2748 mOwner->writeFourcc(mIsAudio ? "soun" : "vide"); // component subtype 2749 mOwner->writeInt32(0); // reserved 2750 mOwner->writeInt32(0); // reserved 2751 mOwner->writeInt32(0); // reserved 2752 // Removing "r" for the name string just makes the string 4 byte aligned 2753 mOwner->writeCString(mIsAudio ? "SoundHandle": "VideoHandle"); // name 2754 mOwner->endBox(); 2755} 2756 2757void MPEG4Writer::Track::writeMdhdBox(time_t now) { 2758 int64_t trakDurationUs = getDurationUs(); 2759 mOwner->beginBox("mdhd"); 2760 mOwner->writeInt32(0); // version=0, flags=0 2761 mOwner->writeInt32(now); // creation time 2762 mOwner->writeInt32(now); // modification time 2763 mOwner->writeInt32(mTimeScale); // media timescale 2764 int32_t mdhdDuration = (trakDurationUs * mTimeScale + 5E5) / 1E6; 2765 mOwner->writeInt32(mdhdDuration); // use media timescale 2766 // Language follows the three letter standard ISO-639-2/T 2767 // 'e', 'n', 'g' for "English", for instance. 2768 // Each character is packed as the difference between its ASCII value and 0x60. 2769 // For "English", these are 00101, 01110, 00111. 2770 // XXX: Where is the padding bit located: 0x15C7? 2771 mOwner->writeInt16(0); // language code 2772 mOwner->writeInt16(0); // predefined 2773 mOwner->endBox(); 2774} 2775 2776void MPEG4Writer::Track::writeDamrBox() { 2777 // 3gpp2 Spec AMRSampleEntry fields 2778 mOwner->beginBox("damr"); 2779 mOwner->writeCString(" "); // vendor: 4 bytes 2780 mOwner->writeInt8(0); // decoder version 2781 mOwner->writeInt16(0x83FF); // mode set: all enabled 2782 mOwner->writeInt8(0); // mode change period 2783 mOwner->writeInt8(1); // frames per sample 2784 mOwner->endBox(); 2785} 2786 2787void MPEG4Writer::Track::writeUrlBox() { 2788 // The table index here refers to the sample description index 2789 // in the sample table entries. 2790 mOwner->beginBox("url "); 2791 mOwner->writeInt32(1); // version=0, flags=1 (self-contained) 2792 mOwner->endBox(); // url 2793} 2794 2795void MPEG4Writer::Track::writeDrefBox() { 2796 mOwner->beginBox("dref"); 2797 mOwner->writeInt32(0); // version=0, flags=0 2798 mOwner->writeInt32(1); // entry count (either url or urn) 2799 writeUrlBox(); 2800 mOwner->endBox(); // dref 2801} 2802 2803void MPEG4Writer::Track::writeDinfBox() { 2804 mOwner->beginBox("dinf"); 2805 writeDrefBox(); 2806 mOwner->endBox(); // dinf 2807} 2808 2809void MPEG4Writer::Track::writeAvccBox() { 2810 CHECK(mCodecSpecificData); 2811 CHECK(mCodecSpecificDataSize >= 5); 2812 2813 // Patch avcc's lengthSize field to match the number 2814 // of bytes we use to indicate the size of a nal unit. 2815 uint8_t *ptr = (uint8_t *)mCodecSpecificData; 2816 ptr[4] = (ptr[4] & 0xfc) | (mOwner->useNalLengthFour() ? 3 : 1); 2817 mOwner->beginBox("avcC"); 2818 mOwner->write(mCodecSpecificData, mCodecSpecificDataSize); 2819 mOwner->endBox(); // avcC 2820} 2821 2822void MPEG4Writer::Track::writeD263Box() { 2823 mOwner->beginBox("d263"); 2824 mOwner->writeInt32(0); // vendor 2825 mOwner->writeInt8(0); // decoder version 2826 mOwner->writeInt8(10); // level: 10 2827 mOwner->writeInt8(0); // profile: 0 2828 mOwner->endBox(); // d263 2829} 2830 2831// This is useful if the pixel is not square 2832void MPEG4Writer::Track::writePaspBox() { 2833 mOwner->beginBox("pasp"); 2834 mOwner->writeInt32(1 << 16); // hspacing 2835 mOwner->writeInt32(1 << 16); // vspacing 2836 mOwner->endBox(); // pasp 2837} 2838 2839void MPEG4Writer::Track::writeSttsBox() { 2840 mOwner->beginBox("stts"); 2841 mOwner->writeInt32(0); // version=0, flags=0 2842 mOwner->writeInt32(mNumSttsTableEntries); 2843 2844 // Compensate for small start time difference from different media tracks 2845 int64_t trackStartTimeOffsetUs = 0; 2846 int64_t moovStartTimeUs = mOwner->getStartTimestampUs(); 2847 if (mStartTimestampUs != moovStartTimeUs) { 2848 CHECK(mStartTimestampUs > moovStartTimeUs); 2849 trackStartTimeOffsetUs = mStartTimestampUs - moovStartTimeUs; 2850 } 2851 List<SttsTableEntry>::iterator it = mSttsTableEntries.begin(); 2852 CHECK(it != mSttsTableEntries.end() && it->sampleCount == 1); 2853 mOwner->writeInt32(it->sampleCount); 2854 int32_t dur = (trackStartTimeOffsetUs * mTimeScale + 500000LL) / 1000000LL; 2855 mOwner->writeInt32(dur + it->sampleDuration); 2856 2857 int64_t totalCount = 1; 2858 while (++it != mSttsTableEntries.end()) { 2859 mOwner->writeInt32(it->sampleCount); 2860 mOwner->writeInt32(it->sampleDuration); 2861 totalCount += it->sampleCount; 2862 } 2863 CHECK(totalCount == mNumSamples); 2864 mOwner->endBox(); // stts 2865} 2866 2867void MPEG4Writer::Track::writeCttsBox() { 2868 if (mIsAudio) { // ctts is not for audio 2869 return; 2870 } 2871 2872 // Do not write ctts box when there is no need to have it. 2873 if ((mNumCttsTableEntries == 1 && 2874 mCttsTableEntries.begin()->sampleDuration == 0) || 2875 mNumCttsTableEntries == 0) { 2876 return; 2877 } 2878 2879 LOGV("ctts box has %d entries", mNumCttsTableEntries); 2880 2881 mOwner->beginBox("ctts"); 2882 if (mHasNegativeCttsDeltaDuration) { 2883 mOwner->writeInt32(0x00010000); // version=1, flags=0 2884 } else { 2885 mOwner->writeInt32(0); // version=0, flags=0 2886 } 2887 mOwner->writeInt32(mNumCttsTableEntries); 2888 2889 int64_t totalCount = 0; 2890 for (List<CttsTableEntry>::iterator it = mCttsTableEntries.begin(); 2891 it != mCttsTableEntries.end(); ++it) { 2892 mOwner->writeInt32(it->sampleCount); 2893 mOwner->writeInt32(it->sampleDuration); 2894 totalCount += it->sampleCount; 2895 } 2896 CHECK(totalCount == mNumSamples); 2897 mOwner->endBox(); // ctts 2898} 2899 2900void MPEG4Writer::Track::writeStssBox() { 2901 mOwner->beginBox("stss"); 2902 mOwner->writeInt32(0); // version=0, flags=0 2903 mOwner->writeInt32(mNumStssTableEntries); // number of sync frames 2904 for (List<int32_t>::iterator it = mStssTableEntries.begin(); 2905 it != mStssTableEntries.end(); ++it) { 2906 mOwner->writeInt32(*it); 2907 } 2908 mOwner->endBox(); // stss 2909} 2910 2911void MPEG4Writer::Track::writeStszBox() { 2912 mOwner->beginBox("stsz"); 2913 mOwner->writeInt32(0); // version=0, flags=0 2914 if (mSamplesHaveSameSize) { 2915 List<size_t>::iterator it = mSampleSizes.begin(); 2916 mOwner->writeInt32(*it); // default sample size 2917 } else { 2918 mOwner->writeInt32(0); 2919 } 2920 mOwner->writeInt32(mNumSamples); 2921 if (!mSamplesHaveSameSize) { 2922 for (List<size_t>::iterator it = mSampleSizes.begin(); 2923 it != mSampleSizes.end(); ++it) { 2924 mOwner->writeInt32(*it); 2925 } 2926 } 2927 mOwner->endBox(); // stsz 2928} 2929 2930void MPEG4Writer::Track::writeStscBox() { 2931 mOwner->beginBox("stsc"); 2932 mOwner->writeInt32(0); // version=0, flags=0 2933 mOwner->writeInt32(mNumStscTableEntries); 2934 for (List<StscTableEntry>::iterator it = mStscTableEntries.begin(); 2935 it != mStscTableEntries.end(); ++it) { 2936 mOwner->writeInt32(it->firstChunk); 2937 mOwner->writeInt32(it->samplesPerChunk); 2938 mOwner->writeInt32(it->sampleDescriptionId); 2939 } 2940 mOwner->endBox(); // stsc 2941} 2942 2943void MPEG4Writer::Track::writeStcoBox(bool use32BitOffset) { 2944 mOwner->beginBox(use32BitOffset? "stco": "co64"); 2945 mOwner->writeInt32(0); // version=0, flags=0 2946 mOwner->writeInt32(mNumStcoTableEntries); 2947 for (List<off64_t>::iterator it = mChunkOffsets.begin(); 2948 it != mChunkOffsets.end(); ++it) { 2949 if (use32BitOffset) { 2950 mOwner->writeInt32(static_cast<int32_t>(*it)); 2951 } else { 2952 mOwner->writeInt64((*it)); 2953 } 2954 } 2955 mOwner->endBox(); // stco or co64 2956} 2957 2958void MPEG4Writer::writeUdtaBox() { 2959 beginBox("udta"); 2960 writeGeoDataBox(); 2961 endBox(); 2962} 2963 2964/* 2965 * Geodata is stored according to ISO-6709 standard. 2966 */ 2967void MPEG4Writer::writeGeoDataBox() { 2968 beginBox("\xA9xyz"); 2969 /* 2970 * For historical reasons, any user data start 2971 * with "\0xA9", must be followed by its assoicated 2972 * language code. 2973 * 0x0012: text string length 2974 * 0x15c7: lang (locale) code: en 2975 */ 2976 writeInt32(0x001215c7); 2977 writeLatitude(mLatitudex10000); 2978 writeLongitude(mLongitudex10000); 2979 writeInt8(0x2F); 2980 endBox(); 2981} 2982 2983} // namespace android 2984