• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2010 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 "ESQueue"
19 #include <media/stagefright/foundation/ADebug.h>
20 
21 #include "ESQueue.h"
22 
23 #include <media/stagefright/foundation/hexdump.h>
24 #include <media/stagefright/foundation/ABitReader.h>
25 #include <media/stagefright/foundation/ABuffer.h>
26 #include <media/stagefright/foundation/AMessage.h>
27 #include <media/stagefright/foundation/ByteUtils.h>
28 #include <media/stagefright/foundation/avc_utils.h>
29 #include <media/stagefright/MediaErrors.h>
30 #include <media/stagefright/MediaDefs.h>
31 #include <media/stagefright/MetaData.h>
32 #include <media/stagefright/MetaDataUtils.h>
33 #include <media/cas/DescramblerAPI.h>
34 #include <media/hardware/CryptoAPI.h>
35 
36 #include <inttypes.h>
37 #include <netinet/in.h>
38 
39 #ifdef ENABLE_CRYPTO
40 #include "HlsSampleDecryptor.h"
41 #endif
42 
43 namespace android {
44 
ElementaryStreamQueue(Mode mode,uint32_t flags)45 ElementaryStreamQueue::ElementaryStreamQueue(Mode mode, uint32_t flags)
46     : mMode(mode),
47       mFlags(flags),
48       mEOSReached(false),
49       mCASystemId(0),
50       mAUIndex(0) {
51 
52     ALOGV("ElementaryStreamQueue(%p) mode %x  flags %x  isScrambled %d  isSampleEncrypted %d",
53             this, mode, flags, isScrambled(), isSampleEncrypted());
54 
55     // Create the decryptor anyway since we don't know the use-case unless key is provided
56     // Won't decrypt if key info not available (e.g., scanner/extractor just parsing ts files)
57     mSampleDecryptor = isSampleEncrypted() ?
58 #ifdef ENABLE_CRYPTO
59         new HlsSampleDecryptor
60 #else
61         new SampleDecryptor
62 #endif
63         : NULL;
64 }
65 
getFormat()66 sp<MetaData> ElementaryStreamQueue::getFormat() {
67     return mFormat;
68 }
69 
clear(bool clearFormat)70 void ElementaryStreamQueue::clear(bool clearFormat) {
71     if (mBuffer != NULL) {
72         mBuffer->setRange(0, 0);
73     }
74 
75     mRangeInfos.clear();
76 
77     if (mScrambledBuffer != NULL) {
78         mScrambledBuffer->setRange(0, 0);
79     }
80     mScrambledRangeInfos.clear();
81 
82     if (clearFormat) {
83         mFormat.clear();
84     }
85 
86     mEOSReached = false;
87 }
88 
isScrambled() const89 bool ElementaryStreamQueue::isScrambled() const {
90     return (mFlags & kFlag_ScrambledData) != 0;
91 }
92 
setCasInfo(int32_t systemId,const std::vector<uint8_t> & sessionId)93 void ElementaryStreamQueue::setCasInfo(
94         int32_t systemId, const std::vector<uint8_t> &sessionId) {
95     mCASystemId = systemId;
96     mCasSessionId = sessionId;
97 }
98 
readVariableBits(ABitReader & bits,int32_t nbits)99 static int32_t readVariableBits(ABitReader &bits, int32_t nbits) {
100     int32_t value = 0;
101     int32_t more_bits = 1;
102 
103     while (more_bits) {
104         value += bits.getBits(nbits);
105         more_bits = bits.getBits(1);
106         if (!more_bits)
107             break;
108         value++;
109         value <<= nbits;
110     }
111     return value;
112 }
113 
114 // Parse AC3 header assuming the current ptr is start position of syncframe,
115 // update metadata only applicable, and return the payload size
parseAC3SyncFrame(const uint8_t * ptr,size_t size,sp<MetaData> * metaData)116 static unsigned parseAC3SyncFrame(
117         const uint8_t *ptr, size_t size, sp<MetaData> *metaData) {
118     static const unsigned channelCountTable[] = {2, 1, 2, 3, 3, 4, 4, 5};
119     static const unsigned samplingRateTable[] = {48000, 44100, 32000};
120 
121     static const unsigned frameSizeTable[19][3] = {
122         { 64, 69, 96 },
123         { 80, 87, 120 },
124         { 96, 104, 144 },
125         { 112, 121, 168 },
126         { 128, 139, 192 },
127         { 160, 174, 240 },
128         { 192, 208, 288 },
129         { 224, 243, 336 },
130         { 256, 278, 384 },
131         { 320, 348, 480 },
132         { 384, 417, 576 },
133         { 448, 487, 672 },
134         { 512, 557, 768 },
135         { 640, 696, 960 },
136         { 768, 835, 1152 },
137         { 896, 975, 1344 },
138         { 1024, 1114, 1536 },
139         { 1152, 1253, 1728 },
140         { 1280, 1393, 1920 },
141     };
142 
143     ABitReader bits(ptr, size);
144     if (bits.numBitsLeft() < 16) {
145         return 0;
146     }
147     if (bits.getBits(16) != 0x0B77) {
148         return 0;
149     }
150 
151     if (bits.numBitsLeft() < 16 + 2 + 6 + 5 + 3 + 3) {
152         ALOGV("Not enough bits left for further parsing");
153         return 0;
154     }
155     bits.skipBits(16);  // crc1
156 
157     unsigned fscod = bits.getBits(2);
158     if (fscod == 3) {
159         ALOGW("Incorrect fscod in AC3 header");
160         return 0;
161     }
162 
163     unsigned frmsizecod = bits.getBits(6);
164     if (frmsizecod > 37) {
165         ALOGW("Incorrect frmsizecod in AC3 header");
166         return 0;
167     }
168 
169     unsigned bsid = bits.getBits(5);
170     if (bsid > 8) {
171         ALOGW("Incorrect bsid in AC3 header. Possibly E-AC-3?");
172         return 0;
173     }
174 
175     bits.skipBits(3); // bsmod
176     unsigned acmod = bits.getBits(3);
177 
178     if ((acmod & 1) > 0 && acmod != 1) {
179         if (bits.numBitsLeft() < 2) {
180             return 0;
181         }
182         bits.skipBits(2); //cmixlev
183     }
184     if ((acmod & 4) > 0) {
185         if (bits.numBitsLeft() < 2) {
186             return 0;
187         }
188         bits.skipBits(2); //surmixlev
189     }
190     if (acmod == 2) {
191         if (bits.numBitsLeft() < 2) {
192             return 0;
193         }
194         bits.skipBits(2); //dsurmod
195     }
196 
197     if (bits.numBitsLeft() < 1) {
198         return 0;
199     }
200     unsigned lfeon = bits.getBits(1);
201 
202     unsigned samplingRate = samplingRateTable[fscod];
203     unsigned payloadSize = frameSizeTable[frmsizecod >> 1][fscod];
204     if (fscod == 1) {
205         payloadSize += frmsizecod & 1;
206     }
207     payloadSize <<= 1;  // convert from 16-bit words to bytes
208 
209     unsigned channelCount = channelCountTable[acmod] + lfeon;
210 
211     if (metaData != NULL) {
212         (*metaData)->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_AC3);
213         (*metaData)->setInt32(kKeyChannelCount, channelCount);
214         (*metaData)->setInt32(kKeySampleRate, samplingRate);
215     }
216 
217     return payloadSize;
218 }
219 
220 // Parse EAC3 header assuming the current ptr is start position of syncframe,
221 // update metadata only applicable, and return the payload size
222 // ATSC A/52:2012 E2.3.1
parseEAC3SyncFrame(const uint8_t * ptr,size_t size,sp<MetaData> * metaData)223 static unsigned parseEAC3SyncFrame(
224     const uint8_t *ptr, size_t size, sp<MetaData> *metaData) {
225     static const unsigned channelCountTable[] = {2, 1, 2, 3, 3, 4, 4, 5};
226     static const unsigned samplingRateTable[] = {48000, 44100, 32000};
227     static const unsigned samplingRateTable2[] = {24000, 22050, 16000};
228 
229     ABitReader bits(ptr, size);
230     if (bits.numBitsLeft() < 16) {
231         ALOGE("Not enough bits left for further parsing");
232         return 0;
233     }
234     if (bits.getBits(16) != 0x0B77) {
235         ALOGE("No valid sync word in EAC3 header");
236         return 0;
237     }
238 
239     // we parse up to bsid so there needs to be at least that many bits
240     if (bits.numBitsLeft() < 2 + 3 + 11 + 2 + 2 + 3 + 1 + 5) {
241         ALOGE("Not enough bits left for further parsing");
242         return 0;
243     }
244 
245     unsigned strmtyp = bits.getBits(2);
246     if (strmtyp == 3) {
247         ALOGE("Incorrect strmtyp in EAC3 header");
248         return 0;
249     }
250 
251     unsigned substreamid = bits.getBits(3);
252     // only the first independent stream is supported
253     if ((strmtyp == 0 || strmtyp == 2) && substreamid != 0)
254         return 0;
255 
256     unsigned frmsiz = bits.getBits(11);
257     unsigned fscod = bits.getBits(2);
258 
259     unsigned samplingRate = 0;
260     if (fscod == 0x3) {
261         unsigned fscod2 = bits.getBits(2);
262         if (fscod2 == 3) {
263             ALOGW("Incorrect fscod2 in EAC3 header");
264             return 0;
265         }
266         samplingRate = samplingRateTable2[fscod2];
267     } else {
268         samplingRate = samplingRateTable[fscod];
269         bits.skipBits(2); // numblkscod
270     }
271 
272     unsigned acmod = bits.getBits(3);
273     unsigned lfeon = bits.getBits(1);
274     unsigned bsid = bits.getBits(5);
275     if (bsid < 11 || bsid > 16) {
276         ALOGW("Incorrect bsid in EAC3 header. Could be AC-3 or some unknown EAC3 format");
277         return 0;
278     }
279 
280     // we currently only support the first independant stream
281     if (metaData != NULL && (strmtyp == 0 || strmtyp == 2)) {
282         unsigned channelCount = channelCountTable[acmod] + lfeon;
283         ALOGV("EAC3 channelCount = %d", channelCount);
284         ALOGV("EAC3 samplingRate = %d", samplingRate);
285         (*metaData)->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_EAC3);
286         (*metaData)->setInt32(kKeyChannelCount, channelCount);
287         (*metaData)->setInt32(kKeySampleRate, samplingRate);
288         (*metaData)->setInt32(kKeyIsSyncFrame, 1);
289     }
290 
291     unsigned payloadSize = frmsiz + 1;
292     payloadSize <<= 1;  // convert from 16-bit words to bytes
293 
294     return payloadSize;
295 }
296 
297 // Parse AC4 header assuming the current ptr is start position of syncframe
298 // and update frameSize and metadata.
parseAC4SyncFrame(const uint8_t * ptr,size_t size,unsigned & frameSize,sp<MetaData> * metaData)299 static status_t parseAC4SyncFrame(
300         const uint8_t *ptr, size_t size, unsigned &frameSize, sp<MetaData> *metaData) {
301     // ETSI TS 103 190-2 V1.1.1 (2015-09), Annex C
302     // The sync_word can be either 0xAC40 or 0xAC41.
303     static const int kSyncWordAC40 = 0xAC40;
304     static const int kSyncWordAC41 = 0xAC41;
305 
306     size_t headerSize = 0;
307     ABitReader bits(ptr, size);
308     int32_t syncWord = bits.getBits(16);
309     if ((syncWord != kSyncWordAC40) && (syncWord != kSyncWordAC41)) {
310         ALOGE("Invalid syncword in AC4 header");
311         return ERROR_MALFORMED;
312     }
313     headerSize += 2;
314 
315     frameSize = bits.getBits(16);
316     headerSize += 2;
317     if (frameSize == 0xFFFF) {
318         frameSize = bits.getBits(24);
319         headerSize += 3;
320     }
321 
322     if (frameSize == 0) {
323         ALOGE("Invalid frame size in AC4 header");
324         return ERROR_MALFORMED;
325     }
326     frameSize += headerSize;
327     // If the sync_word is 0xAC41, a crc_word is also transmitted.
328     if (syncWord == kSyncWordAC41) {
329         frameSize += 2; // crc_word
330     }
331     ALOGV("AC4 frameSize = %u", frameSize);
332 
333     // ETSI TS 103 190-2 V1.1.1 6.2.1.1
334     uint32_t bitstreamVersion = bits.getBits(2);
335     if (bitstreamVersion == 3) {
336         bitstreamVersion += readVariableBits(bits, 2);
337     }
338 
339     bits.skipBits(10); // Sequence Counter
340 
341     uint32_t bWaitFrames = bits.getBits(1);
342     if (bWaitFrames) {
343         uint32_t waitFrames = bits.getBits(3);
344         if (waitFrames > 0) {
345             bits.skipBits(2); // br_code;
346         }
347     }
348 
349     // ETSI TS 103 190 V1.1.1 Table 82
350     bool fsIndex = bits.getBits(1);
351     uint32_t samplingRate = fsIndex ? 48000 : 44100;
352 
353     if (metaData != NULL) {
354         ALOGV("dequeueAccessUnitAC4 Setting mFormat");
355         (*metaData)->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_AC4);
356         (*metaData)->setInt32(kKeyIsSyncFrame, 1);
357         // [FIXME] AC4 channel count is defined per presentation. Provide a default channel count
358         // as stereo for the entire stream.
359         (*metaData)->setInt32(kKeyChannelCount, 2);
360         (*metaData)->setInt32(kKeySampleRate, samplingRate);
361     }
362     return OK;
363 }
364 
365 #define RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bitstream, size) \
366     do { \
367         if ((bitstream).numBitsLeft() < (size)) { \
368         ALOGE("Not enough bits left for further parsing"); \
369         return ERROR_MALFORMED; } \
370     } while (0)
371 
372 // Parse DTS Digital Surround and DTS Express(LBR) stream header
parseDTSHDSyncFrame(const uint8_t * ptr,size_t size,unsigned & frameSize,sp<MetaData> * metaData)373 static status_t parseDTSHDSyncFrame(
374     const uint8_t *ptr, size_t size, unsigned &frameSize, sp<MetaData> *metaData) {
375     static const unsigned channelCountTable[] = {1, 2, 2, 2, 2, 3, 3, 4,
376                                                  4, 5, 6, 6, 6, 7, 8, 8};
377     static const unsigned samplingRateTableCoreSS[] = {0, 8000, 16000, 32000, 0, 0, 11025, 22050,
378                                                        44100, 0, 0, 12000, 24000, 48000, 0, 0};
379     static const unsigned samplingRateTableExtSS[] = {8000, 16000, 32000, 64000, 128000,
380                                                       22050, 44100, 88200, 176400, 352800,
381                                                       12000, 24000, 48000, 96000, 192000, 384000};
382 
383     const uint32_t DTSHD_SYNC_CORE_16BIT_BE = 0x7ffe8001;
384     const uint32_t DTSHD_SYNC_EXSS_16BIT_BE = 0x64582025;
385 
386     uint32_t numChannels = 0, samplingRate = 0;
387     bool isLBR = false;
388 
389     ABitReader bits(ptr, size);
390 
391     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 32);
392     uint32_t dtshdSyncWord = bits.getBits(32);
393 
394     // Expecting DTS Digital Surround or DTS Express(LBR) streams only
395     if (dtshdSyncWord == DTSHD_SYNC_CORE_16BIT_BE) { // DTS Digital Surround Header
396         RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (1 + 5 + 1 + 7 + 14 + 6 + 4 + 15 + 2));
397 
398         // FTYPE, SHORT, CRC, NBLKS
399         bits.skipBits(1 + 5 + 1 + 7);
400 
401         frameSize = bits.getBits(14) + 1;
402         uint32_t amode = bits.getBits(6);
403         uint32_t freqIndex = bits.getBits(4);
404 
405         // RATE, FIXEDBIT, DYNF, TIMEF, AUXF, HDCD, EXT_AUDIO_ID, EXT_AUDIO, ASPF
406         bits.skipBits(5 + 1 + 1 + 1 + 1 + 1 + 3 + 1 + 1);
407 
408         uint32_t lfeFlag = bits.getBits(2);
409         numChannels = (amode <= 15) ? channelCountTable[amode] : 0;
410         numChannels += ((lfeFlag == 1) || (lfeFlag == 2)) ? 1 : 0;
411         samplingRate = (freqIndex <= 15) ? samplingRateTableCoreSS[freqIndex] : 0;
412 
413         isLBR = false;
414     } else if (dtshdSyncWord == DTSHD_SYNC_EXSS_16BIT_BE) { // DTS Express(LBR) Header
415         RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (8 + 2 + 1));
416 
417         uint32_t extHeadersize, extSSFsize;
418         uint32_t numAudioPresent = 1, numAssets = 1;
419         uint32_t nuActiveExSSMask[8];
420 
421         // userDefinedBits
422         bits.skipBits(8);
423 
424         uint32_t extSSIndex = bits.getBits(2);
425         uint32_t headerSizeType = bits.getBits(1);
426 
427         if (headerSizeType == 0) {
428             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (8 + 16));
429 
430             extHeadersize = bits.getBits(8) + 1;
431             extSSFsize = bits.getBits(16) + 1;
432         } else {
433             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (12 + 20));
434 
435             extHeadersize = bits.getBits(12) + 1;
436             extSSFsize = bits.getBits(20) + 1;
437         }
438 
439         RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (1));
440 
441         uint32_t staticFieldsPresent = bits.getBits(1);
442 
443         if (staticFieldsPresent) {
444             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (2 + 3 + 1));
445 
446             // nuRefClockCode, nuExSSFrameDurationCode
447             bits.skipBits(2 + 3);
448 
449             if (bits.getBits(1)) {
450                 RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (32 + 4));
451 
452                 bits.skipBits(32 + 4);
453             }
454 
455             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (3 + 3));
456 
457             // numAudioPresent, numAssets
458             bits.skipBits(3 + 3);
459 
460             for (uint32_t nAuPr = 0; nAuPr < numAudioPresent; nAuPr++) {
461                 RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (extSSIndex + 1));
462 
463                 nuActiveExSSMask[nAuPr] = bits.getBits(extSSIndex + 1);
464             }
465 
466             for (uint32_t nAuPr = 0; nAuPr < numAudioPresent; nAuPr++) {
467                 for (uint32_t nSS = 0; nSS < extSSIndex + 1; nSS++) {
468                     if (((nuActiveExSSMask[nAuPr] >> nSS) & 0x1) == 1) {
469                         RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 8);
470 
471                         // nuActiveAssetMask
472                         bits.skipBits(8);
473                     }
474                 }
475             }
476 
477             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 1);
478 
479             // bMixMetadataEnbl
480             if (bits.getBits(1)) {
481                 RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (2 + 2 + 2));
482 
483                 // nuMixMetadataAdjLevel
484                 bits.skipBits(2);
485 
486                 uint32_t bits4MixOutMask = (bits.getBits(2) + 1) << 2;
487                 uint32_t numMixOutConfigs = bits.getBits(2) + 1;
488 
489                 for (int ns = 0; ns < numMixOutConfigs; ns++) {
490                     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, bits4MixOutMask);
491 
492                     // nuMixOutChMask
493                     bits.skipBits(bits4MixOutMask);
494                 }
495             }
496         }
497 
498         for (int nAst = 0; nAst < numAssets; nAst++) {
499             int bits4ExSSFsize = (headerSizeType == 0) ? 16 : 20;
500 
501             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, bits4ExSSFsize);
502 
503             bits.skipBits(bits4ExSSFsize);
504         }
505 
506         /* Asset descriptor */
507         for (int nAst = 0; nAst < numAssets; nAst++) {
508             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (9 + 3));
509 
510             // nuAssetDescriptFsize, nuAssetIndex
511             bits.skipBits(9 + 3);
512 
513             if (staticFieldsPresent) {
514                 RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 1);
515 
516                 // bAssetTypeDescrPresent
517                 if (bits.getBits(1)) {
518                     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 4);
519 
520                     // nuAssetTypeDescriptor
521                     bits.skipBits(4);
522                 }
523 
524                 RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 1);
525 
526                 // bLanguageDescrPresent
527                 if (bits.getBits(1)) {
528                     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 24);
529 
530                     // LanguageDescriptor
531                     bits.skipBits(24);
532                 }
533 
534                 RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 1);
535 
536                 // bInfoTextPresent
537                 if (bits.getBits(1)) {
538                     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 10);
539 
540                     uint32_t nuInfoTextByteSize = bits.getBits(10) + 1;
541 
542                     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (nuInfoTextByteSize * 8));
543 
544                     // InfoTextString
545                     bits.skipBits(nuInfoTextByteSize * 8);
546                 }
547 
548                 RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (5 + 4 + 8));
549 
550                 // nuBitResolution
551                 bits.skipBits(5);
552 
553                 samplingRate = samplingRateTableExtSS[bits.getBits(4)];
554                 numChannels = bits.getBits(8) + 1;
555             }
556         }
557 
558         frameSize = extHeadersize + extSSFsize;
559         isLBR = true;
560     } else {
561         ALOGE("No valid sync word in DTS/DTSHD header");
562         return ERROR_MALFORMED;
563     }
564 
565     if (metaData != NULL) {
566         if (isLBR) {
567             (*metaData)->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_DTS_HD);
568             (*metaData)->setInt32(kKeyAudioProfile, 0x2); // CodecProfileLevel.DTS_HDProfileLBR
569         } else {
570             (*metaData)->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_DTS);
571         }
572         (*metaData)->setInt32(kKeyChannelCount, numChannels);
573         (*metaData)->setInt32(kKeySampleRate, samplingRate);
574     }
575     return OK;
576 }
577 
extractVarLenBitFields(ABitReader * bits,size_t * bitsUsed,uint32_t * value,unsigned ucTable[],bool extractAndAddFlag)578 static status_t extractVarLenBitFields(
579     ABitReader *bits, size_t *bitsUsed, uint32_t *value,
580     unsigned ucTable[], bool extractAndAddFlag) {
581 
582     static const unsigned bitsUsedTbl[8] = {1, 1, 1, 1, 2, 2, 3, 3}; // prefix code lengths
583     static const unsigned indexTbl[8] = {0, 0, 0, 0, 1, 1, 2, 3}; // code to prefix code index map
584 
585     /* Clone the bitstream */
586     ABitReader bitStream(bits->data(), bits->numBitsLeft() / 8);
587     ABitReader bitstreamClone(bits->data(), bits->numBitsLeft() / 8);
588 
589     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bitstreamClone, 3);
590 
591     unsigned code = bitstreamClone.getBits(3);
592     unsigned totalBitsUsed = bitsUsedTbl[code];
593     unsigned unIndex = indexTbl[code];
594 
595     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bitStream, totalBitsUsed);
596 
597     bitStream.skipBits(totalBitsUsed);
598 
599     uint32_t unValue = 0;
600     if (ucTable[unIndex] > 0) {
601         if (extractAndAddFlag) {
602             for (unsigned un = 0; un < unIndex; un++) {
603                 unValue += (1 << ucTable[un]);
604             }
605 
606             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bitStream, ucTable[unIndex]);
607 
608             unValue += bitStream.getBits(ucTable[unIndex]);
609             totalBitsUsed += ucTable[unIndex];
610         } else {
611             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bitStream, ucTable[unIndex]);
612 
613             unValue += bitStream.getBits(ucTable[unIndex]);
614             totalBitsUsed += ucTable[unIndex];
615         }
616     }
617 
618     *bitsUsed = (size_t)totalBitsUsed;
619     *value = unValue;
620     return OK;
621 }
622 
623 // Parse DTS UHD Profile-2 stream header
parseDTSUHDSyncFrame(const uint8_t * ptr,size_t size,unsigned & frameSize,sp<MetaData> * metaData)624 static status_t parseDTSUHDSyncFrame(
625     const uint8_t *ptr, size_t size, unsigned &frameSize, sp<MetaData> *metaData) {
626 
627     static const uint32_t DTSUHD_SYNC_CORE_16BIT_BE = 0x40411BF2;
628     static const uint32_t DTSUHD_NONSYNC_CORE_16BIT_BE = 0x71C442E8;
629 
630     unsigned audioSamplRate = 0;
631 
632     ABitReader bits(ptr, size);
633 
634     RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 32);
635 
636     uint32_t syncWord = bits.getBits(32);
637 
638     bool isSyncFrameFlag = false;
639     switch (syncWord) {
640         case DTSUHD_SYNC_CORE_16BIT_BE:
641             isSyncFrameFlag = true;
642             break;
643         case DTSUHD_NONSYNC_CORE_16BIT_BE:
644             isSyncFrameFlag = false;
645             break;
646         default:
647             ALOGE("No valid sync word in DTSUHD header");
648             return ERROR_MALFORMED; // invalid sync word
649     }
650 
651     unsigned uctable1[4] = { 5, 8, 10, 12 };
652     uint32_t sizeOfFTOCPayload = 0;
653     size_t nuBitsUsed = 0;
654     status_t status = OK;
655 
656     status = extractVarLenBitFields(&bits, &nuBitsUsed, &sizeOfFTOCPayload, uctable1, true);
657 
658     if (status != OK) {
659         ALOGE("Failed to extractVarLenBitFields from DTSUHD header");
660         return ERROR_MALFORMED;
661     }
662 
663     bits.skipBits(nuBitsUsed);
664 
665     if (isSyncFrameFlag) {
666         RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (1 + 2 + 3 + 2 + 1));
667 
668         // FullChannelBasedMixFlag, ETSI TS 103 491 V1.2.1, Section 6.4.6.1
669         if (!(bits.getBits(1))) {
670             // This implementation only supports full channel mask-based
671             // audio presentation (i.e. 2.0, 5.1, 11.1 mix without objects)
672             ALOGE("Objects not supported, only DTSUHD full channel mask-based mix");
673             return ERROR_MALFORMED;
674         }
675 
676         // BaseDuration, FrameDuration
677         bits.skipBits(2 + 3);
678 
679         unsigned clockRateIndex = bits.getBits(2);
680         unsigned clockRateHertz = 0;
681 
682         switch (clockRateIndex) {
683             case 0:
684                 clockRateHertz = 32000;
685                 break;
686             case 1:
687                 clockRateHertz = 44100;
688                 break;
689             case 2:
690                 clockRateHertz = 48000;
691                 break;
692             default:
693                 ALOGE("Invalid clockRateIndex in DTSUHD header");
694                 return ERROR_MALFORMED;
695         }
696 
697         if (bits.getBits(1)) {
698             RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, (32 + 4));
699 
700             bits.skipBits(32 + 4);
701         }
702 
703         RETURN_ERROR_IF_NOT_ENOUGH_BYTES_LEFT(bits, 2);
704 
705         unsigned samplRateMultiplier = (1 << bits.getBits(2));
706         audioSamplRate = clockRateHertz * samplRateMultiplier;
707     }
708 
709     uint32_t chunkPayloadBytes = 0;
710     int numOfMDChunks = isSyncFrameFlag ? 1 : 0; // Metadata chunks
711     for (int nmdc = 0; nmdc < numOfMDChunks; nmdc++) {
712         unsigned uctable2[4] = {6, 9, 12, 15};
713         uint32_t nuMDChunkSize = 0;
714         nuBitsUsed = 0;
715 
716         status = extractVarLenBitFields(&bits, &nuBitsUsed, &nuMDChunkSize, uctable2, true);
717         if (status != OK) {
718             ALOGE("Failed to extractVarLenBitFields from DTSUHD header");
719             return ERROR_MALFORMED;
720         }
721 
722         bits.skipBits(nuBitsUsed);
723 
724         if (nuMDChunkSize > 32767) {
725             ALOGE("Unsupported number of metadata chunks in DTSUHD header");
726             return ERROR_MALFORMED;
727         }
728         chunkPayloadBytes += nuMDChunkSize;
729     }
730 
731     // Ony one audio chunk is supported
732     int numAudioChunks = 1;
733     for (int nac = 0; nac < numAudioChunks; nac++) {
734         uint32_t acID = 256, nuAudioChunkSize = 0;
735 
736         // isSyncFrameFlag means that ACID is present
737         if (isSyncFrameFlag) {
738             unsigned uctable3[4] = {2, 4, 6, 8};
739             nuBitsUsed = 0;
740 
741             status = extractVarLenBitFields(&bits, &nuBitsUsed, &acID, uctable3, true);
742 
743             if (status != OK) {
744                 ALOGE("Failed to extractVarLenBitFields from DTSUHD header");
745                 return ERROR_MALFORMED;
746             }
747 
748             bits.skipBits(nuBitsUsed);
749         }
750 
751         nuBitsUsed = 0;
752         if (acID == 0) {
753             nuAudioChunkSize = 0;
754         } else {
755             unsigned uctable4[4] = {9, 11, 13, 16};
756 
757             status = extractVarLenBitFields(&bits, &nuBitsUsed, &nuAudioChunkSize, uctable4, true);
758 
759             if (status != OK) {
760                 ALOGE("Failed to extractVarLenBitFields from DTSUHD header");
761                 return ERROR_MALFORMED;
762             }
763         }
764 
765         if (nuAudioChunkSize > 65535){
766             ALOGE("Unsupported number of audio chunks in DTSUHD header");
767             return ERROR_MALFORMED;
768         }
769 
770         chunkPayloadBytes += nuAudioChunkSize;
771     }
772 
773     frameSize = (sizeOfFTOCPayload + 1) + chunkPayloadBytes;
774 
775     if (metaData != NULL) {
776         (*metaData)->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_DTS_UHD);
777         (*metaData)->setInt32(kKeyAudioProfile, 0x2); // CodecProfileLevel.DTS_UHDProfileP2
778         (*metaData)->setInt32(kKeyChannelCount, 2); // Setting default channel count as stereo
779         (*metaData)->setInt32(kKeySampleRate, audioSamplRate);
780     }
781 
782     return OK;
783 }
784 
isSeeminglyValidDTSHDHeader(const uint8_t * ptr,size_t size,unsigned & frameSize)785 static status_t isSeeminglyValidDTSHDHeader(const uint8_t *ptr, size_t size,unsigned &frameSize)
786 {
787     return parseDTSHDSyncFrame(ptr, size, frameSize, NULL);
788 }
789 
isSeeminglyValidDTSUHDHeader(const uint8_t * ptr,size_t size,unsigned & frameSize)790 static status_t isSeeminglyValidDTSUHDHeader(const uint8_t *ptr, size_t size,unsigned &frameSize)
791 {
792     return parseDTSUHDSyncFrame(ptr, size, frameSize, NULL);
793 }
794 
IsSeeminglyValidAC4Header(const uint8_t * ptr,size_t size,unsigned & frameSize)795 static status_t IsSeeminglyValidAC4Header(const uint8_t *ptr, size_t size, unsigned &frameSize) {
796     return parseAC4SyncFrame(ptr, size, frameSize, NULL);
797 }
798 
IsSeeminglyValidADTSHeader(const uint8_t * ptr,size_t size,size_t * frameLength)799 static bool IsSeeminglyValidADTSHeader(
800         const uint8_t *ptr, size_t size, size_t *frameLength) {
801     if (size < 7) {
802         // Not enough data to verify header.
803         return false;
804     }
805 
806     if (ptr[0] != 0xff || (ptr[1] >> 4) != 0x0f) {
807         return false;
808     }
809 
810     unsigned layer = (ptr[1] >> 1) & 3;
811 
812     if (layer != 0) {
813         return false;
814     }
815 
816     unsigned ID = (ptr[1] >> 3) & 1;
817     unsigned profile_ObjectType = ptr[2] >> 6;
818 
819     if (ID == 1 && profile_ObjectType == 3) {
820         // MPEG-2 profile 3 is reserved.
821         return false;
822     }
823 
824     size_t frameLengthInHeader =
825             ((ptr[3] & 3) << 11) + (ptr[4] << 3) + ((ptr[5] >> 5) & 7);
826     if (frameLengthInHeader > size) {
827         return false;
828     }
829 
830     *frameLength = frameLengthInHeader;
831     return true;
832 }
833 
IsSeeminglyValidMPEGAudioHeader(const uint8_t * ptr,size_t size)834 static bool IsSeeminglyValidMPEGAudioHeader(const uint8_t *ptr, size_t size) {
835     if (size < 3) {
836         // Not enough data to verify header.
837         return false;
838     }
839 
840     if (ptr[0] != 0xff || (ptr[1] >> 5) != 0x07) {
841         return false;
842     }
843 
844     unsigned ID = (ptr[1] >> 3) & 3;
845 
846     if (ID == 1) {
847         return false;  // reserved
848     }
849 
850     unsigned layer = (ptr[1] >> 1) & 3;
851 
852     if (layer == 0) {
853         return false;  // reserved
854     }
855 
856     unsigned bitrateIndex = (ptr[2] >> 4);
857 
858     if (bitrateIndex == 0x0f) {
859         return false;  // reserved
860     }
861 
862     unsigned samplingRateIndex = (ptr[2] >> 2) & 3;
863 
864     if (samplingRateIndex == 3) {
865         return false;  // reserved
866     }
867 
868     return true;
869 }
870 
appendData(const void * data,size_t size,int64_t timeUs,int32_t payloadOffset,uint32_t pesScramblingControl)871 status_t ElementaryStreamQueue::appendData(
872         const void *data, size_t size, int64_t timeUs,
873         int32_t payloadOffset, uint32_t pesScramblingControl) {
874 
875     if (mEOSReached) {
876         ALOGE("appending data after EOS");
877         return ERROR_MALFORMED;
878     }
879 
880     if (!isScrambled() && (mBuffer == NULL || mBuffer->size() == 0)) {
881         switch (mMode) {
882             case H264:
883             case MPEG_VIDEO:
884             {
885 #if 0
886                 if (size < 4 || memcmp("\x00\x00\x00\x01", data, 4)) {
887                     return ERROR_MALFORMED;
888                 }
889 #else
890                 uint8_t *ptr = (uint8_t *)data;
891 
892                 ssize_t startOffset = -1;
893                 for (size_t i = 0; i + 2 < size; ++i) {
894                     if (!memcmp("\x00\x00\x01", &ptr[i], 3)) {
895                         startOffset = i;
896                         break;
897                     }
898                 }
899 
900                 if (startOffset < 0) {
901                     return ERROR_MALFORMED;
902                 }
903 
904                 if (mFormat == NULL && startOffset > 0) {
905                     ALOGI("found something resembling an H.264/MPEG syncword "
906                           "at offset %zd",
907                           startOffset);
908                 }
909 
910                 data = &ptr[startOffset];
911                 size -= startOffset;
912 #endif
913                 break;
914             }
915 
916             case MPEG4_VIDEO:
917             {
918 #if 0
919                 if (size < 3 || memcmp("\x00\x00\x01", data, 3)) {
920                     return ERROR_MALFORMED;
921                 }
922 #else
923                 uint8_t *ptr = (uint8_t *)data;
924 
925                 ssize_t startOffset = -1;
926                 for (size_t i = 0; i + 2 < size; ++i) {
927                     if (!memcmp("\x00\x00\x01", &ptr[i], 3)) {
928                         startOffset = i;
929                         break;
930                     }
931                 }
932 
933                 if (startOffset < 0) {
934                     return ERROR_MALFORMED;
935                 }
936 
937                 if (startOffset > 0) {
938                     ALOGI("found something resembling an H.264/MPEG syncword "
939                           "at offset %zd",
940                           startOffset);
941                 }
942 
943                 data = &ptr[startOffset];
944                 size -= startOffset;
945 #endif
946                 break;
947             }
948 
949             case AAC:
950             {
951                 uint8_t *ptr = (uint8_t *)data;
952 
953 #if 0
954                 if (size < 2 || ptr[0] != 0xff || (ptr[1] >> 4) != 0x0f) {
955                     return ERROR_MALFORMED;
956                 }
957 #else
958                 ssize_t startOffset = -1;
959                 size_t frameLength;
960                 for (size_t i = 0; i < size; ++i) {
961                     if (IsSeeminglyValidADTSHeader(
962                             &ptr[i], size - i, &frameLength)) {
963                         startOffset = i;
964                         break;
965                     }
966                 }
967 
968                 if (startOffset < 0) {
969                     return ERROR_MALFORMED;
970                 }
971 
972                 if (startOffset > 0) {
973                     ALOGI("found something resembling an AAC syncword at "
974                           "offset %zd",
975                           startOffset);
976                 }
977 
978                 if (frameLength != size - startOffset) {
979                     ALOGV("First ADTS AAC frame length is %zd bytes, "
980                           "while the buffer size is %zd bytes.",
981                           frameLength, size - startOffset);
982                 }
983 
984                 data = &ptr[startOffset];
985                 size -= startOffset;
986 #endif
987                 break;
988             }
989 
990             case AC3:
991             case EAC3:
992             {
993                 uint8_t *ptr = (uint8_t *)data;
994 
995                 ssize_t startOffset = -1;
996                 for (size_t i = 0; i < size; ++i) {
997                     unsigned payloadSize = 0;
998                     if (mMode == AC3) {
999                         payloadSize = parseAC3SyncFrame(&ptr[i], size - i, NULL);
1000                     } else if (mMode == EAC3) {
1001                         payloadSize = parseEAC3SyncFrame(&ptr[i], size - i, NULL);
1002                     }
1003                     if (payloadSize > 0) {
1004                         startOffset = i;
1005                         break;
1006                     }
1007                 }
1008 
1009                 if (startOffset < 0) {
1010                     return ERROR_MALFORMED;
1011                 }
1012 
1013                 if (startOffset > 0) {
1014                     ALOGI("found something resembling an (E)AC3 syncword at "
1015                           "offset %zd",
1016                           startOffset);
1017                 }
1018 
1019                 data = &ptr[startOffset];
1020                 size -= startOffset;
1021                 break;
1022             }
1023 
1024             case AC4:
1025             {
1026                 uint8_t *ptr = (uint8_t *)data;
1027                 unsigned frameSize = 0;
1028                 ssize_t startOffset = -1;
1029 
1030                 // A valid AC4 stream should have minimum of 7 bytes in its buffer.
1031                 // (Sync header 4 bytes + AC4 toc 3 bytes)
1032                 if (size < 7) {
1033                     return ERROR_MALFORMED;
1034                 }
1035                 for (size_t i = 0; i < size; ++i) {
1036                     if (IsSeeminglyValidAC4Header(&ptr[i], size - i, frameSize) == OK) {
1037                         startOffset = i;
1038                         break;
1039                     }
1040                 }
1041 
1042                 if (startOffset < 0) {
1043                     return ERROR_MALFORMED;
1044                 }
1045 
1046                 if (startOffset > 0) {
1047                     ALOGI("found something resembling an AC4 syncword at "
1048                           "offset %zd",
1049                           startOffset);
1050                 }
1051                 if (frameSize != size - startOffset) {
1052                     ALOGV("AC4 frame size is %u bytes, while the buffer size is %zd bytes.",
1053                           frameSize, size - startOffset);
1054                 }
1055 
1056                 data = &ptr[startOffset];
1057                 size -= startOffset;
1058                 break;
1059             }
1060 
1061             case MPEG_AUDIO:
1062             {
1063                 uint8_t *ptr = (uint8_t *)data;
1064 
1065                 ssize_t startOffset = -1;
1066                 for (size_t i = 0; i < size; ++i) {
1067                     if (IsSeeminglyValidMPEGAudioHeader(&ptr[i], size - i)) {
1068                         startOffset = i;
1069                         break;
1070                     }
1071                 }
1072 
1073                 if (startOffset < 0) {
1074                     return ERROR_MALFORMED;
1075                 }
1076 
1077                 if (startOffset > 0) {
1078                     ALOGI("found something resembling an MPEG audio "
1079                           "syncword at offset %zd",
1080                           startOffset);
1081                 }
1082 
1083                 data = &ptr[startOffset];
1084                 size -= startOffset;
1085                 break;
1086             }
1087 
1088             case DTS: //  Checking for DTS or DTS-HD syncword
1089             case DTS_HD:
1090             {
1091                 uint8_t *ptr = (uint8_t *)data;
1092                 unsigned frameSize = 0;
1093                 ssize_t startOffset = -1;
1094 
1095                 for (size_t i = 0; i < size; ++i) {
1096                     if (isSeeminglyValidDTSHDHeader(&ptr[i], size - i, frameSize) == OK) {
1097                         startOffset = i;
1098                         break;
1099                     }
1100                 }
1101 
1102                 if (startOffset < 0) {
1103                     return ERROR_MALFORMED;
1104                 }
1105                 if (startOffset > 0) {
1106                     ALOGI("found something resembling a DTS-HD syncword at "
1107                           "offset %zd",
1108                           startOffset);
1109                 }
1110 
1111                 if (frameSize != size - startOffset) {
1112                     ALOGV("DTS-HD frame size is %u bytes, while the buffer size is %zd bytes.",
1113                           frameSize, size - startOffset);
1114                 }
1115 
1116                 data = &ptr[startOffset];
1117                 size -= startOffset;
1118                 break;
1119             }
1120 
1121             case DTS_UHD:
1122             {
1123                 uint8_t *ptr = (uint8_t *)data;
1124                 ssize_t startOffset = -1;
1125                 unsigned frameSize = 0;
1126 
1127                 for (size_t i = 0; i < size; ++i) {
1128                     if (isSeeminglyValidDTSUHDHeader(&ptr[i], size - i, frameSize) == OK) {
1129                         startOffset = i;
1130                         break;
1131                     }
1132                 }
1133 
1134                 if (startOffset < 0) {
1135                     return ERROR_MALFORMED;
1136                 }
1137                 if (startOffset >= 0) {
1138                     ALOGI("found something resembling a DTS UHD syncword"
1139                           "syncword at offset %zd",
1140                           startOffset);
1141                 }
1142 
1143                 if (frameSize != size - startOffset) {
1144                     ALOGV("DTS-UHD frame size is %u bytes, while the buffer size is %zd bytes.",
1145                           frameSize, size - startOffset);
1146                 }
1147                 data = &ptr[startOffset];
1148                 size -= startOffset;
1149                 break;
1150             }
1151 
1152             case PCM_AUDIO:
1153             case METADATA:
1154             {
1155                 break;
1156             }
1157 
1158             default:
1159                 ALOGE("Unknown mode: %d", mMode);
1160                 return ERROR_MALFORMED;
1161         }
1162     }
1163 
1164     size_t neededSize = (mBuffer == NULL ? 0 : mBuffer->size()) + size;
1165     if (mBuffer == NULL || neededSize > mBuffer->capacity()) {
1166         neededSize = (neededSize + 65535) & ~65535;
1167 
1168         ALOGV("resizing buffer to size %zu", neededSize);
1169 
1170         sp<ABuffer> buffer = new ABuffer(neededSize);
1171         if (mBuffer != NULL) {
1172             memcpy(buffer->data(), mBuffer->data(), mBuffer->size());
1173             buffer->setRange(0, mBuffer->size());
1174         } else {
1175             buffer->setRange(0, 0);
1176         }
1177 
1178         mBuffer = buffer;
1179     }
1180 
1181     memcpy(mBuffer->data() + mBuffer->size(), data, size);
1182     mBuffer->setRange(0, mBuffer->size() + size);
1183 
1184     RangeInfo info;
1185     info.mLength = size;
1186     info.mTimestampUs = timeUs;
1187     info.mPesOffset = payloadOffset;
1188     info.mPesScramblingControl = pesScramblingControl;
1189     mRangeInfos.push_back(info);
1190 
1191 #if 0
1192     if (mMode == AAC) {
1193         ALOGI("size = %zu, timeUs = %.2f secs", size, timeUs / 1E6);
1194         hexdump(data, size);
1195     }
1196 #endif
1197 
1198     return OK;
1199 }
1200 
appendScrambledData(const void * data,size_t size,size_t leadingClearBytes,int32_t keyId,bool isSync,sp<ABuffer> clearSizes,sp<ABuffer> encSizes)1201 void ElementaryStreamQueue::appendScrambledData(
1202         const void *data, size_t size,
1203         size_t leadingClearBytes,
1204         int32_t keyId, bool isSync,
1205         sp<ABuffer> clearSizes, sp<ABuffer> encSizes) {
1206     if (!isScrambled()) {
1207         return;
1208     }
1209 
1210     size_t neededSize = (mScrambledBuffer == NULL ? 0 : mScrambledBuffer->size()) + size;
1211     if (mScrambledBuffer == NULL || neededSize > mScrambledBuffer->capacity()) {
1212         neededSize = (neededSize + 65535) & ~65535;
1213 
1214         ALOGI("resizing scrambled buffer to size %zu", neededSize);
1215 
1216         sp<ABuffer> buffer = new ABuffer(neededSize);
1217         if (mScrambledBuffer != NULL) {
1218             memcpy(buffer->data(), mScrambledBuffer->data(), mScrambledBuffer->size());
1219             buffer->setRange(0, mScrambledBuffer->size());
1220         } else {
1221             buffer->setRange(0, 0);
1222         }
1223 
1224         mScrambledBuffer = buffer;
1225     }
1226     memcpy(mScrambledBuffer->data() + mScrambledBuffer->size(), data, size);
1227     mScrambledBuffer->setRange(0, mScrambledBuffer->size() + size);
1228 
1229     ScrambledRangeInfo scrambledInfo;
1230     scrambledInfo.mLength = size;
1231     scrambledInfo.mLeadingClearBytes = leadingClearBytes;
1232     scrambledInfo.mKeyId = keyId;
1233     scrambledInfo.mIsSync = isSync;
1234     scrambledInfo.mClearSizes = clearSizes;
1235     scrambledInfo.mEncSizes = encSizes;
1236 
1237     ALOGV("[stream %d] appending scrambled range: size=%zu", mMode, size);
1238 
1239     mScrambledRangeInfos.push_back(scrambledInfo);
1240 }
1241 
dequeueScrambledAccessUnit()1242 sp<ABuffer> ElementaryStreamQueue::dequeueScrambledAccessUnit() {
1243     size_t nextScan = mBuffer->size();
1244     int32_t pesOffset = 0, pesScramblingControl = 0;
1245     int64_t timeUs = fetchTimestamp(nextScan, &pesOffset, &pesScramblingControl);
1246     if (timeUs < 0ll) {
1247         ALOGE("Negative timeUs");
1248         return NULL;
1249     }
1250 
1251     // return scrambled unit
1252     int32_t keyId = pesScramblingControl, isSync = 0, scrambledLength = 0;
1253     sp<ABuffer> clearSizes, encSizes;
1254     size_t leadingClearBytes;
1255     while (mScrambledRangeInfos.size() > mRangeInfos.size()) {
1256         auto it = mScrambledRangeInfos.begin();
1257         ALOGV("[stream %d] fetching scrambled range: size=%zu", mMode, it->mLength);
1258 
1259         if (scrambledLength > 0) {
1260             // This shouldn't happen since we always dequeue the entire PES.
1261             ALOGW("Discarding srambled length %d", scrambledLength);
1262         }
1263         scrambledLength = it->mLength;
1264 
1265         // TODO: handle key id change, use first non-zero keyId for now
1266         if (keyId == 0) {
1267             keyId = it->mKeyId;
1268         }
1269         clearSizes = it->mClearSizes;
1270         encSizes = it->mEncSizes;
1271         isSync = it->mIsSync;
1272         leadingClearBytes = it->mLeadingClearBytes;
1273         mScrambledRangeInfos.erase(it);
1274     }
1275     if (scrambledLength == 0) {
1276         ALOGE("[stream %d] empty scrambled unit!", mMode);
1277         return NULL;
1278     }
1279 
1280     // Retrieve the leading clear bytes info, and use it to set the clear
1281     // range on mBuffer. Note that the leading clear bytes includes the
1282     // PES header portion, while mBuffer doesn't.
1283     if ((int32_t)leadingClearBytes > pesOffset) {
1284         mBuffer->setRange(0, leadingClearBytes - pesOffset);
1285     } else {
1286         mBuffer->setRange(0, 0);
1287     }
1288 
1289     // Try to parse formats, and if unavailable set up a dummy format.
1290     // Only support the following modes for scrambled content for now.
1291     // (will be expanded later).
1292     if (mFormat == NULL) {
1293         mFormat = new MetaData;
1294         switch (mMode) {
1295             case H264:
1296             {
1297                 if (!MakeAVCCodecSpecificData(
1298                         *mFormat, mBuffer->data(), mBuffer->size())) {
1299                     ALOGI("Creating dummy AVC format for scrambled content");
1300 
1301                     mFormat->setCString(kKeyMIMEType, MEDIA_MIMETYPE_VIDEO_AVC);
1302                     mFormat->setInt32(kKeyWidth, 1280);
1303                     mFormat->setInt32(kKeyHeight, 720);
1304                 }
1305                 break;
1306             }
1307             case AAC:
1308             {
1309                 if (!MakeAACCodecSpecificData(
1310                         *mFormat, mBuffer->data(), mBuffer->size())) {
1311                     ALOGI("Creating dummy AAC format for scrambled content");
1312 
1313                     MakeAACCodecSpecificData(*mFormat,
1314                             1 /*profile*/, 7 /*sampling_freq_index*/, 1 /*channel_config*/);
1315                     mFormat->setInt32(kKeyIsADTS, true);
1316                 }
1317 
1318                 break;
1319             }
1320             case MPEG_VIDEO:
1321             {
1322                 ALOGI("Creating dummy MPEG format for scrambled content");
1323 
1324                 mFormat->setCString(kKeyMIMEType, MEDIA_MIMETYPE_VIDEO_MPEG2);
1325                 mFormat->setInt32(kKeyWidth, 1280);
1326                 mFormat->setInt32(kKeyHeight, 720);
1327                 break;
1328             }
1329             default:
1330             {
1331                 ALOGE("Unknown mode for scrambled content");
1332                 return NULL;
1333             }
1334         }
1335 
1336         // for MediaExtractor.CasInfo
1337         mFormat->setInt32(kKeyCASystemID, mCASystemId);
1338         mFormat->setData(kKeyCASessionID,
1339                 0, mCasSessionId.data(), mCasSessionId.size());
1340     }
1341 
1342     mBuffer->setRange(0, 0);
1343 
1344     // copy into scrambled access unit
1345     sp<ABuffer> scrambledAccessUnit = ABuffer::CreateAsCopy(
1346             mScrambledBuffer->data(), scrambledLength);
1347 
1348     scrambledAccessUnit->meta()->setInt64("timeUs", timeUs);
1349     if (isSync) {
1350         scrambledAccessUnit->meta()->setInt32("isSync", 1);
1351     }
1352 
1353     // fill in CryptoInfo fields for AnotherPacketSource::read()
1354     // MediaCas doesn't use cryptoMode, but set to non-zero value here.
1355     scrambledAccessUnit->meta()->setInt32(
1356             "cryptoMode", CryptoPlugin::kMode_AES_CTR);
1357     scrambledAccessUnit->meta()->setInt32("cryptoKey", keyId);
1358     scrambledAccessUnit->meta()->setBuffer("clearBytes", clearSizes);
1359     scrambledAccessUnit->meta()->setBuffer("encBytes", encSizes);
1360     scrambledAccessUnit->meta()->setInt32("pesOffset", pesOffset);
1361 
1362     memmove(mScrambledBuffer->data(),
1363             mScrambledBuffer->data() + scrambledLength,
1364             mScrambledBuffer->size() - scrambledLength);
1365 
1366     mScrambledBuffer->setRange(0, mScrambledBuffer->size() - scrambledLength);
1367 
1368     ALOGV("[stream %d] dequeued scrambled AU: timeUs=%lld, size=%zu",
1369             mMode, (long long)timeUs, scrambledAccessUnit->size());
1370 
1371     return scrambledAccessUnit;
1372 }
1373 
dequeueAccessUnit()1374 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnit() {
1375     if (isScrambled()) {
1376         return dequeueScrambledAccessUnit();
1377     }
1378 
1379     if ((mFlags & kFlag_AlignedData) && mMode == H264) {
1380         if (mRangeInfos.empty()) {
1381             return NULL;
1382         }
1383 
1384         RangeInfo info = *mRangeInfos.begin();
1385         mRangeInfos.erase(mRangeInfos.begin());
1386 
1387         sp<ABuffer> accessUnit = new ABuffer(info.mLength);
1388         memcpy(accessUnit->data(), mBuffer->data(), info.mLength);
1389         accessUnit->meta()->setInt64("timeUs", info.mTimestampUs);
1390 
1391         memmove(mBuffer->data(),
1392                 mBuffer->data() + info.mLength,
1393                 mBuffer->size() - info.mLength);
1394 
1395         mBuffer->setRange(0, mBuffer->size() - info.mLength);
1396 
1397         if (mFormat == NULL) {
1398             mFormat = new MetaData;
1399             if (!MakeAVCCodecSpecificData(*mFormat, accessUnit->data(), accessUnit->size())) {
1400                 mFormat.clear();
1401             }
1402         }
1403 
1404         return accessUnit;
1405     }
1406 
1407     switch (mMode) {
1408         case H264:
1409             return dequeueAccessUnitH264();
1410         case AAC:
1411             return dequeueAccessUnitAAC();
1412         case AC3:
1413         case EAC3:
1414             return dequeueAccessUnitEAC3();
1415         case AC4:
1416             return dequeueAccessUnitAC4();
1417         case MPEG_VIDEO:
1418             return dequeueAccessUnitMPEGVideo();
1419         case MPEG4_VIDEO:
1420             return dequeueAccessUnitMPEG4Video();
1421         case PCM_AUDIO:
1422             return dequeueAccessUnitPCMAudio();
1423         case METADATA:
1424             return dequeueAccessUnitMetadata();
1425         case DTS: // Using same dequeue function for both DTS and DTS-HD types.
1426         case DTS_HD:
1427             return dequeueAccessUnitDTSOrDTSHD();
1428         case DTS_UHD:
1429             return dequeueAccessUnitDTSUHD();
1430         default:
1431             if (mMode != MPEG_AUDIO) {
1432                 ALOGE("Unknown mode");
1433                 return NULL;
1434             }
1435             return dequeueAccessUnitMPEGAudio();
1436     }
1437 }
1438 
dequeueAccessUnitDTSOrDTSHD()1439 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitDTSOrDTSHD() {
1440     unsigned syncStartPos = 0; // in bytes
1441     unsigned payloadSize = 0;
1442     sp<MetaData> format = new MetaData;
1443 
1444     ALOGV("dequeueAccessUnitDTSOrDTSHD[%d]: mBuffer %p(%zu)", mAUIndex,
1445           mBuffer->data(), mBuffer->size());
1446 
1447     while (true) {
1448         if (syncStartPos + 4 >= mBuffer->size()) {
1449             return NULL;
1450         }
1451         uint8_t *ptr = mBuffer->data() + syncStartPos;
1452         size_t size = mBuffer->size() - syncStartPos;
1453         status_t status = parseDTSHDSyncFrame(ptr, size, payloadSize, &format);
1454         if (status == 0) {
1455             break;
1456         }
1457         ++syncStartPos;
1458     }
1459 
1460     if (mBuffer->size() < syncStartPos + payloadSize) {
1461         ALOGV("Not enough buffer size for DTS/DTS-HD");
1462         return NULL;
1463     }
1464 
1465     if (mFormat == NULL) {
1466         mFormat = format;
1467     }
1468 
1469     int64_t timeUs = fetchTimestamp(syncStartPos + payloadSize);
1470     if (timeUs < 0LL) {
1471         ALOGE("negative timeUs");
1472         return NULL;
1473     }
1474     mAUIndex++;
1475 
1476     sp<ABuffer> accessUnit = new ABuffer(syncStartPos + payloadSize);
1477     memcpy(accessUnit->data(), mBuffer->data(), syncStartPos + payloadSize);
1478 
1479     accessUnit->meta()->setInt64("timeUs", timeUs);
1480     accessUnit->meta()->setInt32("isSync", 1);
1481 
1482     memmove(
1483         mBuffer->data(),
1484         mBuffer->data() + syncStartPos + payloadSize,
1485         mBuffer->size() - syncStartPos - payloadSize);
1486 
1487     mBuffer->setRange(0, mBuffer->size() - syncStartPos - payloadSize);
1488 
1489     return accessUnit;
1490 }
1491 
dequeueAccessUnitDTSUHD()1492 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitDTSUHD()
1493 {
1494     unsigned syncStartPos = 0; // in bytes
1495     unsigned payloadSize = 0;
1496     sp<MetaData> format = new MetaData;
1497 
1498     ALOGV("dequeueAccessUnitDTSUHD[%d]: mBuffer %p(%zu)", mAUIndex,
1499           mBuffer->data(), mBuffer->size());
1500 
1501     while (true) {
1502         if (syncStartPos + 4 >= mBuffer->size()) {
1503             return NULL;
1504         }
1505         uint8_t *ptr = mBuffer->data() + syncStartPos;
1506         size_t size = mBuffer->size() - syncStartPos;
1507         status_t status = parseDTSUHDSyncFrame(ptr, size, payloadSize, &format);
1508         if (status == 0) {
1509             break;
1510         }
1511         ++syncStartPos;
1512     }
1513 
1514     if (mBuffer->size() < syncStartPos + payloadSize) {
1515         ALOGV("Not enough buffer size for DTS-UHD");
1516         return NULL;
1517     }
1518 
1519     if (mFormat == NULL) {
1520         mFormat = format;
1521     }
1522 
1523     int64_t timeUs = fetchTimestamp(syncStartPos + payloadSize);
1524     if (timeUs < 0LL) {
1525         ALOGE("negative timeUs");
1526         return NULL;
1527     }
1528     mAUIndex++;
1529 
1530     sp<ABuffer> accessUnit = new ABuffer(syncStartPos + payloadSize);
1531     memcpy(accessUnit->data(), mBuffer->data(), syncStartPos + payloadSize);
1532 
1533     accessUnit->meta()->setInt64("timeUs", timeUs);
1534     accessUnit->meta()->setInt32("isSync", 1);
1535 
1536     memmove(
1537         mBuffer->data(),
1538         mBuffer->data() + syncStartPos + payloadSize,
1539         mBuffer->size() - syncStartPos - payloadSize);
1540 
1541     mBuffer->setRange(0, mBuffer->size() - syncStartPos - payloadSize);
1542 
1543     return accessUnit;
1544 }
1545 
dequeueAccessUnitEAC3()1546 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitEAC3() {
1547     unsigned syncStartPos = 0;  // in bytes
1548     unsigned payloadSize = 0;
1549     sp<MetaData> format = new MetaData;
1550 
1551     ALOGV("dequeueAccessUnitEAC3[%d]: mBuffer %p(%zu)", mAUIndex,
1552             mBuffer->data(), mBuffer->size());
1553 
1554     while (true) {
1555         if (syncStartPos + 2 >= mBuffer->size()) {
1556             return NULL;
1557         }
1558 
1559         uint8_t *ptr = mBuffer->data() + syncStartPos;
1560         size_t size = mBuffer->size() - syncStartPos;
1561         if (mMode == AC3) {
1562             payloadSize = parseAC3SyncFrame(ptr, size, &format);
1563         } else if (mMode == EAC3) {
1564             payloadSize = parseEAC3SyncFrame(ptr, size, &format);
1565         }
1566         if (payloadSize > 0) {
1567             break;
1568         }
1569 
1570         ALOGV("dequeueAccessUnitEAC3[%d]: syncStartPos %u payloadSize %u",
1571                 mAUIndex, syncStartPos, payloadSize);
1572 
1573         ++syncStartPos;
1574     }
1575 
1576     if (mBuffer->size() < syncStartPos + payloadSize) {
1577         ALOGV("Not enough buffer size for E/AC3");
1578         return NULL;
1579     }
1580 
1581     if (mFormat == NULL) {
1582         mFormat = format;
1583     }
1584 
1585     int64_t timeUs = fetchTimestamp(syncStartPos + payloadSize);
1586     if (timeUs < 0ll) {
1587         ALOGE("negative timeUs");
1588         return NULL;
1589     }
1590 
1591     // Not decrypting if key info not available (e.g., scanner/extractor parsing ts files)
1592     if (mSampleDecryptor != NULL) {
1593         if (mMode == AC3) {
1594             mSampleDecryptor->processAC3(mBuffer->data() + syncStartPos, payloadSize);
1595         } else if (mMode == EAC3) {
1596             ALOGE("EAC3 AU is encrypted and decryption is not supported");
1597             return NULL;
1598         }
1599     }
1600     mAUIndex++;
1601 
1602     sp<ABuffer> accessUnit = new ABuffer(syncStartPos + payloadSize);
1603     memcpy(accessUnit->data(), mBuffer->data(), syncStartPos + payloadSize);
1604 
1605     accessUnit->meta()->setInt64("timeUs", timeUs);
1606     accessUnit->meta()->setInt32("isSync", 1);
1607 
1608     memmove(
1609             mBuffer->data(),
1610             mBuffer->data() + syncStartPos + payloadSize,
1611             mBuffer->size() - syncStartPos - payloadSize);
1612 
1613     mBuffer->setRange(0, mBuffer->size() - syncStartPos - payloadSize);
1614 
1615     return accessUnit;
1616 }
1617 
dequeueAccessUnitAC4()1618 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitAC4() {
1619     unsigned syncStartPos = 0;
1620     unsigned payloadSize = 0;
1621     sp<MetaData> format = new MetaData;
1622     ALOGV("dequeueAccessUnit_AC4[%d]: mBuffer %p(%zu)", mAUIndex, mBuffer->data(), mBuffer->size());
1623 
1624     // A valid AC4 stream should have minimum of 7 bytes in its buffer.
1625     // (Sync header 4 bytes + AC4 toc 3 bytes)
1626     if (mBuffer->size() < 7) {
1627         return NULL;
1628     }
1629 
1630     while (true) {
1631         if (syncStartPos + 2 >= mBuffer->size()) {
1632             return NULL;
1633         }
1634 
1635         status_t status = parseAC4SyncFrame(
1636                     mBuffer->data() + syncStartPos,
1637                     mBuffer->size() - syncStartPos,
1638                     payloadSize,
1639                     &format);
1640         if (status == OK) {
1641             break;
1642         }
1643 
1644         ALOGV("dequeueAccessUnit_AC4[%d]: syncStartPos %u payloadSize %u",
1645                 mAUIndex, syncStartPos, payloadSize);
1646 
1647         ++syncStartPos;
1648     }
1649 
1650     if (mBuffer->size() < syncStartPos + payloadSize) {
1651         ALOGV("Not enough buffer size for AC4");
1652         return NULL;
1653     }
1654 
1655     if (mFormat == NULL) {
1656         mFormat = format;
1657     }
1658 
1659     int64_t timeUs = fetchTimestamp(syncStartPos + payloadSize);
1660     if (timeUs < 0ll) {
1661         ALOGE("negative timeUs");
1662         return NULL;
1663     }
1664     mAUIndex++;
1665 
1666     sp<ABuffer> accessUnit = new ABuffer(syncStartPos + payloadSize);
1667     memcpy(accessUnit->data(), mBuffer->data(), syncStartPos + payloadSize);
1668 
1669     accessUnit->meta()->setInt64("timeUs", timeUs);
1670     accessUnit->meta()->setInt32("isSync", 1);
1671 
1672     memmove(
1673             mBuffer->data(),
1674             mBuffer->data() + syncStartPos + payloadSize,
1675             mBuffer->size() - syncStartPos - payloadSize);
1676 
1677     mBuffer->setRange(0, mBuffer->size() - syncStartPos - payloadSize);
1678     return accessUnit;
1679 }
1680 
dequeueAccessUnitPCMAudio()1681 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitPCMAudio() {
1682     if (mBuffer->size() < 4) {
1683         return NULL;
1684     }
1685 
1686     ABitReader bits(mBuffer->data(), 4);
1687     if (bits.getBits(8) != 0xa0) {
1688         ALOGE("Unexpected bit values");
1689         return NULL;
1690     }
1691     unsigned numAUs = bits.getBits(8);
1692     bits.skipBits(8);
1693     bits.skipBits(2); // quantization_word_length
1694     unsigned audio_sampling_frequency = bits.getBits(3);
1695     unsigned num_channels = bits.getBits(3);
1696 
1697     if (audio_sampling_frequency != 2) {
1698         ALOGE("Wrong sampling freq");
1699         return NULL;
1700     }
1701     if (num_channels != 1u) {
1702         ALOGE("Wrong channel #");
1703         return NULL;
1704     }
1705 
1706     if (mFormat == NULL) {
1707         mFormat = new MetaData;
1708         mFormat->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_RAW);
1709         mFormat->setInt32(kKeyChannelCount, 2);
1710         mFormat->setInt32(kKeySampleRate, 48000);
1711         mFormat->setInt32(kKeyPcmEncoding, kAudioEncodingPcm16bit);
1712     }
1713 
1714     static const size_t kFramesPerAU = 80;
1715     size_t frameSize = 2 /* numChannels */ * sizeof(int16_t);
1716 
1717     size_t payloadSize = numAUs * frameSize * kFramesPerAU;
1718 
1719     if (mBuffer->size() < 4 + payloadSize) {
1720         return NULL;
1721     }
1722 
1723     sp<ABuffer> accessUnit = new ABuffer(payloadSize);
1724     memcpy(accessUnit->data(), mBuffer->data() + 4, payloadSize);
1725 
1726     int64_t timeUs = fetchTimestamp(payloadSize + 4);
1727     if (timeUs < 0LL) {
1728         ALOGE("Negative timeUs");
1729         return NULL;
1730     }
1731     accessUnit->meta()->setInt64("timeUs", timeUs);
1732     accessUnit->meta()->setInt32("isSync", 1);
1733 
1734     int16_t *ptr = (int16_t *)accessUnit->data();
1735     for (size_t i = 0; i < payloadSize / sizeof(int16_t); ++i) {
1736         ptr[i] = ntohs(ptr[i]);
1737     }
1738 
1739     memmove(
1740             mBuffer->data(),
1741             mBuffer->data() + 4 + payloadSize,
1742             mBuffer->size() - 4 - payloadSize);
1743 
1744     mBuffer->setRange(0, mBuffer->size() - 4 - payloadSize);
1745 
1746     return accessUnit;
1747 }
1748 
dequeueAccessUnitAAC()1749 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitAAC() {
1750     if (mBuffer->size() == 0) {
1751         return NULL;
1752     }
1753 
1754     if (mRangeInfos.empty()) {
1755         return NULL;
1756     }
1757 
1758     const RangeInfo &info = *mRangeInfos.begin();
1759     if (info.mLength == 0 || mBuffer->size() < info.mLength) {
1760         return NULL;
1761     }
1762 
1763     if (info.mTimestampUs < 0LL) {
1764         ALOGE("Negative info.mTimestampUs");
1765         return NULL;
1766     }
1767 
1768     ALOGV("dequeueAccessUnit_AAC[%d]: mBuffer %zu info.mLength %zu",
1769             mAUIndex, mBuffer->size(), info.mLength);
1770 
1771     struct ADTSPosition {
1772         size_t offset;
1773         size_t headerSize;
1774         size_t length;
1775     };
1776 
1777     Vector<ADTSPosition> frames;
1778 
1779     // The idea here is consume all AAC frames starting at offsets before
1780     // info.mLength so we can assign a meaningful timestamp without
1781     // having to interpolate.
1782     // The final AAC frame may well extend into the next RangeInfo but
1783     // that's ok.
1784     size_t offset = 0;
1785     while (offset < info.mLength) {
1786         if (offset + 7 > mBuffer->size()) {
1787             return NULL;
1788         }
1789 
1790         ABitReader bits(mBuffer->data() + offset, mBuffer->size() - offset);
1791 
1792         // adts_fixed_header
1793 
1794         if (bits.getBits(12) != 0xfffu) {
1795             ALOGE("Wrong atds_fixed_header");
1796             return NULL;
1797         }
1798         bits.skipBits(3);  // ID, layer
1799         bool protection_absent = bits.getBits(1) != 0;
1800 
1801         if (mFormat == NULL) {
1802             mFormat = new MetaData;
1803             if (!MakeAACCodecSpecificData(
1804                     *mFormat, mBuffer->data() + offset, mBuffer->size() - offset)) {
1805                 return NULL;
1806             }
1807 
1808             int32_t sampleRate;
1809             int32_t numChannels;
1810             if (!mFormat->findInt32(kKeySampleRate, &sampleRate)) {
1811                 ALOGE("SampleRate not found");
1812                 return NULL;
1813             }
1814             if (!mFormat->findInt32(kKeyChannelCount, &numChannels)) {
1815                 ALOGE("ChannelCount not found");
1816                 return NULL;
1817             }
1818 
1819             ALOGI("found AAC codec config (%d Hz, %d channels)",
1820                  sampleRate, numChannels);
1821         }
1822 
1823         // profile_ObjectType, sampling_frequency_index, private_bits,
1824         // channel_configuration, original_copy, home
1825         bits.skipBits(12);
1826 
1827         // adts_variable_header
1828 
1829         // copyright_identification_bit, copyright_identification_start
1830         bits.skipBits(2);
1831 
1832         unsigned aac_frame_length = bits.getBits(13);
1833         if (aac_frame_length == 0){
1834             ALOGE("b/62673179, Invalid AAC frame length!");
1835             android_errorWriteLog(0x534e4554, "62673179");
1836             return NULL;
1837         }
1838 
1839         bits.skipBits(11);  // adts_buffer_fullness
1840 
1841         unsigned number_of_raw_data_blocks_in_frame = bits.getBits(2);
1842 
1843         if (number_of_raw_data_blocks_in_frame != 0) {
1844             // To be implemented.
1845             ALOGE("Should not reach here.");
1846             return NULL;
1847         }
1848 
1849         if (offset + aac_frame_length > mBuffer->size()) {
1850             return NULL;
1851         }
1852 
1853         size_t headerSize = protection_absent ? 7 : 9;
1854 
1855         // tracking the frame positions first then decrypt only if an accessUnit to be generated
1856         if (mSampleDecryptor != NULL) {
1857             ADTSPosition frame = {
1858                 .offset     = offset,
1859                 .headerSize = headerSize,
1860                 .length     = aac_frame_length
1861             };
1862 
1863             frames.push(frame);
1864         }
1865 
1866         offset += aac_frame_length;
1867     }
1868 
1869     // Decrypting only if the loop didn't exit early and an accessUnit is about to be generated
1870     // Not decrypting if key info not available (e.g., scanner/extractor parsing ts files)
1871     if (mSampleDecryptor != NULL) {
1872         for (size_t frameId = 0; frameId < frames.size(); frameId++) {
1873             const ADTSPosition &frame = frames.itemAt(frameId);
1874 
1875             mSampleDecryptor->processAAC(frame.headerSize,
1876                     mBuffer->data() + frame.offset, frame.length);
1877 //            ALOGV("dequeueAccessUnitAAC[%zu]: while offset %zu headerSize %zu frame_len %zu",
1878 //                    frameId, frame.offset, frame.headerSize, frame.length);
1879         }
1880     }
1881     mAUIndex++;
1882 
1883     int64_t timeUs = fetchTimestamp(offset);
1884 
1885     sp<ABuffer> accessUnit = new ABuffer(offset);
1886     memcpy(accessUnit->data(), mBuffer->data(), offset);
1887 
1888     memmove(mBuffer->data(), mBuffer->data() + offset,
1889             mBuffer->size() - offset);
1890     mBuffer->setRange(0, mBuffer->size() - offset);
1891 
1892     accessUnit->meta()->setInt64("timeUs", timeUs);
1893     accessUnit->meta()->setInt32("isSync", 1);
1894 
1895     return accessUnit;
1896 }
1897 
fetchTimestamp(size_t size,int32_t * pesOffset,int32_t * pesScramblingControl)1898 int64_t ElementaryStreamQueue::fetchTimestamp(
1899         size_t size, int32_t *pesOffset, int32_t *pesScramblingControl) {
1900     int64_t timeUs = -1;
1901     bool first = true;
1902 
1903     while (size > 0) {
1904         if (mRangeInfos.empty()) {
1905             return timeUs;
1906         }
1907 
1908         RangeInfo *info = &*mRangeInfos.begin();
1909 
1910         if (first) {
1911             timeUs = info->mTimestampUs;
1912             if (pesOffset != NULL) {
1913                 *pesOffset = info->mPesOffset;
1914             }
1915             if (pesScramblingControl != NULL) {
1916                 *pesScramblingControl = info->mPesScramblingControl;
1917             }
1918             first = false;
1919         }
1920 
1921         if (info->mLength > size) {
1922             info->mLength -= size;
1923             size = 0;
1924         } else {
1925             size -= info->mLength;
1926 
1927             mRangeInfos.erase(mRangeInfos.begin());
1928             info = NULL;
1929         }
1930 
1931     }
1932 
1933     if (timeUs == 0LL) {
1934         ALOGV("Returning 0 timestamp");
1935     }
1936 
1937     return timeUs;
1938 }
1939 
dequeueAccessUnitH264()1940 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitH264() {
1941     const uint8_t *data = mBuffer->data();
1942 
1943     size_t size = mBuffer->size();
1944     Vector<NALPosition> nals;
1945 
1946     size_t totalSize = 0;
1947     size_t seiCount = 0;
1948 
1949     status_t err;
1950     const uint8_t *nalStart;
1951     size_t nalSize;
1952     bool foundSlice = false;
1953     bool foundIDR = false;
1954 
1955     ALOGV("dequeueAccessUnit_H264[%d] %p/%zu", mAUIndex, data, size);
1956 
1957     while ((err = getNextNALUnit(&data, &size, &nalStart, &nalSize)) == OK) {
1958         if (nalSize == 0) continue;
1959 
1960         unsigned nalType = nalStart[0] & 0x1f;
1961         bool flush = false;
1962 
1963         if (nalType == 1 || nalType == 5) {
1964             if (nalType == 5) {
1965                 foundIDR = true;
1966             }
1967             if (foundSlice) {
1968                 //TODO: Shouldn't this have been called with nalSize-1?
1969                 ABitReader br(nalStart + 1, nalSize);
1970                 unsigned first_mb_in_slice = parseUE(&br);
1971 
1972                 if (first_mb_in_slice == 0) {
1973                     // This slice starts a new frame.
1974 
1975                     flush = true;
1976                 }
1977             }
1978 
1979             foundSlice = true;
1980         } else if ((nalType == 9 || nalType == 7) && foundSlice) {
1981             // Access unit delimiter and SPS will be associated with the
1982             // next frame.
1983 
1984             flush = true;
1985         } else if (nalType == 6 && nalSize > 0) {
1986             // found non-zero sized SEI
1987             ++seiCount;
1988         }
1989 
1990         if (flush) {
1991             // The access unit will contain all nal units up to, but excluding
1992             // the current one, separated by 0x00 0x00 0x00 0x01 startcodes.
1993 
1994             size_t auSize = 4 * nals.size() + totalSize;
1995             sp<ABuffer> accessUnit = new ABuffer(auSize);
1996             sp<ABuffer> sei;
1997 
1998             if (seiCount > 0) {
1999                 sei = new ABuffer(seiCount * sizeof(NALPosition));
2000                 accessUnit->meta()->setBuffer("sei", sei);
2001             }
2002 
2003 #if !LOG_NDEBUG
2004             AString out;
2005 #endif
2006 
2007             size_t dstOffset = 0;
2008             size_t seiIndex = 0;
2009             size_t shrunkBytes = 0;
2010             for (size_t i = 0; i < nals.size(); ++i) {
2011                 const NALPosition &pos = nals.itemAt(i);
2012 
2013                 unsigned nalType = mBuffer->data()[pos.nalOffset] & 0x1f;
2014 
2015                 if (nalType == 6 && pos.nalSize > 0) {
2016                     if (seiIndex >= sei->size() / sizeof(NALPosition)) {
2017                         ALOGE("Wrong seiIndex");
2018                         return NULL;
2019                     }
2020                     NALPosition &seiPos = ((NALPosition *)sei->data())[seiIndex++];
2021                     seiPos.nalOffset = dstOffset + 4;
2022                     seiPos.nalSize = pos.nalSize;
2023                 }
2024 
2025 #if !LOG_NDEBUG
2026                 char tmp[128];
2027                 sprintf(tmp, "0x%02x", nalType);
2028                 if (i > 0) {
2029                     out.append(", ");
2030                 }
2031                 out.append(tmp);
2032 #endif
2033 
2034                 memcpy(accessUnit->data() + dstOffset, "\x00\x00\x00\x01", 4);
2035 
2036                 if (mSampleDecryptor != NULL && (nalType == 1 || nalType == 5)) {
2037                     uint8_t *nalData = mBuffer->data() + pos.nalOffset;
2038                     size_t newSize = mSampleDecryptor->processNal(nalData, pos.nalSize);
2039                     // Note: the data can shrink due to unescaping, but it can never grow
2040                     if (newSize > pos.nalSize) {
2041                         // don't log unless verbose, since this can get called a lot if
2042                         // the caller is trying to resynchronize
2043                         ALOGV("expected sample size < %u, got %zu", pos.nalSize, newSize);
2044                         return NULL;
2045                     }
2046                     memcpy(accessUnit->data() + dstOffset + 4,
2047                             nalData,
2048                             newSize);
2049                     dstOffset += newSize + 4;
2050 
2051                     size_t thisShrunkBytes = pos.nalSize - newSize;
2052                     //ALOGV("dequeueAccessUnitH264[%d]: nalType: %d -> %zu (%zu)",
2053                     //        nalType, (int)pos.nalSize, newSize, thisShrunkBytes);
2054 
2055                     shrunkBytes += thisShrunkBytes;
2056                 }
2057                 else {
2058                     memcpy(accessUnit->data() + dstOffset + 4,
2059                             mBuffer->data() + pos.nalOffset,
2060                             pos.nalSize);
2061 
2062                     dstOffset += pos.nalSize + 4;
2063                     //ALOGV("dequeueAccessUnitH264 [%d] %d @%d",
2064                     //        nalType, (int)pos.nalSize, (int)pos.nalOffset);
2065                 }
2066             }
2067 
2068 #if !LOG_NDEBUG
2069             ALOGV("accessUnit contains nal types %s", out.c_str());
2070 #endif
2071 
2072             const NALPosition &pos = nals.itemAt(nals.size() - 1);
2073             size_t nextScan = pos.nalOffset + pos.nalSize;
2074 
2075             memmove(mBuffer->data(),
2076                     mBuffer->data() + nextScan,
2077                     mBuffer->size() - nextScan);
2078 
2079             mBuffer->setRange(0, mBuffer->size() - nextScan);
2080 
2081             int64_t timeUs = fetchTimestamp(nextScan);
2082             if (timeUs < 0LL) {
2083                 ALOGE("Negative timeUs");
2084                 return NULL;
2085             }
2086 
2087             accessUnit->meta()->setInt64("timeUs", timeUs);
2088             if (foundIDR) {
2089                 accessUnit->meta()->setInt32("isSync", 1);
2090             }
2091 
2092             if (mFormat == NULL) {
2093                 mFormat = new MetaData;
2094                 if (!MakeAVCCodecSpecificData(*mFormat,
2095                         accessUnit->data(),
2096                         accessUnit->size())) {
2097                     mFormat.clear();
2098                 }
2099             }
2100 
2101             if (mSampleDecryptor != NULL && shrunkBytes > 0) {
2102                 size_t adjustedSize = accessUnit->size() - shrunkBytes;
2103                 ALOGV("dequeueAccessUnitH264[%d]: AU size adjusted %zu -> %zu",
2104                         mAUIndex, accessUnit->size(), adjustedSize);
2105                 accessUnit->setRange(0, adjustedSize);
2106             }
2107 
2108             ALOGV("dequeueAccessUnitH264[%d]: AU %p(%zu) dstOffset:%zu, nals:%zu, totalSize:%zu ",
2109                     mAUIndex, accessUnit->data(), accessUnit->size(),
2110                     dstOffset, nals.size(), totalSize);
2111             mAUIndex++;
2112 
2113             return accessUnit;
2114         }
2115 
2116         NALPosition pos;
2117         pos.nalOffset = nalStart - mBuffer->data();
2118         pos.nalSize = nalSize;
2119 
2120         nals.push(pos);
2121 
2122         totalSize += nalSize;
2123     }
2124     if (err != (status_t)-EAGAIN) {
2125         ALOGE("Unexpeted err");
2126         return NULL;
2127     }
2128 
2129     return NULL;
2130 }
2131 
dequeueAccessUnitMPEGAudio()2132 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitMPEGAudio() {
2133     const uint8_t *data = mBuffer->data();
2134     size_t size = mBuffer->size();
2135 
2136     if (size < 4) {
2137         return NULL;
2138     }
2139 
2140     uint32_t header = U32_AT(data);
2141 
2142     size_t frameSize;
2143     int samplingRate, numChannels, bitrate, numSamples;
2144     if (!GetMPEGAudioFrameSize(
2145                 header, &frameSize, &samplingRate, &numChannels,
2146                 &bitrate, &numSamples)) {
2147         ALOGE("Failed to get audio frame size");
2148         mBuffer->setRange(0, 0);
2149         return NULL;
2150     }
2151 
2152     if (size < frameSize) {
2153         return NULL;
2154     }
2155 
2156     unsigned layer = 4 - ((header >> 17) & 3);
2157 
2158     sp<ABuffer> accessUnit = new ABuffer(frameSize);
2159     memcpy(accessUnit->data(), data, frameSize);
2160 
2161     memmove(mBuffer->data(),
2162             mBuffer->data() + frameSize,
2163             mBuffer->size() - frameSize);
2164 
2165     mBuffer->setRange(0, mBuffer->size() - frameSize);
2166 
2167     int64_t timeUs = fetchTimestamp(frameSize);
2168     if (timeUs < 0LL) {
2169         ALOGE("Negative timeUs");
2170         return NULL;
2171     }
2172 
2173     if (mFormat != NULL) {
2174         const char *mime;
2175         if (mFormat->findCString(kKeyMIMEType, &mime)) {
2176             if ((layer == 1) && strcmp (mime, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_I)) {
2177                 ALOGE("Audio layer is not MPEG_LAYER_I");
2178                 return NULL;
2179             } else if ((layer == 2) && strcmp (mime, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_II)) {
2180                 ALOGE("Audio layer is not MPEG_LAYER_II");
2181                 return NULL;
2182             } else if ((layer == 3) && strcmp (mime, MEDIA_MIMETYPE_AUDIO_MPEG)) {
2183                 ALOGE("Audio layer is not AUDIO_MPEG");
2184                 return NULL;
2185             }
2186         }
2187     }
2188 
2189     accessUnit->meta()->setInt64("timeUs", timeUs);
2190     accessUnit->meta()->setInt32("isSync", 1);
2191 
2192     if (mFormat == NULL) {
2193         mFormat = new MetaData;
2194 
2195         switch (layer) {
2196             case 1:
2197                 mFormat->setCString(
2198                         kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_I);
2199                 break;
2200             case 2:
2201                 mFormat->setCString(
2202                         kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_II);
2203                 break;
2204             case 3:
2205                 mFormat->setCString(
2206                         kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG);
2207                 break;
2208             default:
2209                 return NULL;
2210         }
2211 
2212         mFormat->setInt32(kKeySampleRate, samplingRate);
2213         mFormat->setInt32(kKeyChannelCount, numChannels);
2214     }
2215 
2216     return accessUnit;
2217 }
2218 
EncodeSize14(uint8_t ** _ptr,size_t size)2219 static void EncodeSize14(uint8_t **_ptr, size_t size) {
2220     if (size > 0x3fff) {
2221         ALOGE("Wrong size");
2222         return;
2223     }
2224 
2225     uint8_t *ptr = *_ptr;
2226 
2227     *ptr++ = 0x80 | (size >> 7);
2228     *ptr++ = size & 0x7f;
2229 
2230     *_ptr = ptr;
2231 }
2232 
MakeMPEGVideoESDS(const sp<ABuffer> & csd)2233 static sp<ABuffer> MakeMPEGVideoESDS(const sp<ABuffer> &csd) {
2234     sp<ABuffer> esds = new ABuffer(csd->size() + 25);
2235 
2236     uint8_t *ptr = esds->data();
2237     *ptr++ = 0x03;
2238     EncodeSize14(&ptr, 22 + csd->size());
2239 
2240     *ptr++ = 0x00;  // ES_ID
2241     *ptr++ = 0x00;
2242 
2243     *ptr++ = 0x00;  // streamDependenceFlag, URL_Flag, OCRstreamFlag
2244 
2245     *ptr++ = 0x04;
2246     EncodeSize14(&ptr, 16 + csd->size());
2247 
2248     *ptr++ = 0x40;  // Audio ISO/IEC 14496-3
2249 
2250     for (size_t i = 0; i < 12; ++i) {
2251         *ptr++ = 0x00;
2252     }
2253 
2254     *ptr++ = 0x05;
2255     EncodeSize14(&ptr, csd->size());
2256 
2257     memcpy(ptr, csd->data(), csd->size());
2258 
2259     return esds;
2260 }
2261 
dequeueAccessUnitMPEGVideo()2262 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitMPEGVideo() {
2263     const uint8_t *data = mBuffer->data();
2264     size_t size = mBuffer->size();
2265 
2266     Vector<size_t> userDataPositions;
2267 
2268     bool sawPictureStart = false;
2269     int pprevStartCode = -1;
2270     int prevStartCode = -1;
2271     int currentStartCode = -1;
2272     bool gopFound = false;
2273     bool isClosedGop = false;
2274     bool brokenLink = false;
2275 
2276     size_t offset = 0;
2277     while (offset + 3 < size) {
2278         if (memcmp(&data[offset], "\x00\x00\x01", 3)) {
2279             ++offset;
2280             continue;
2281         }
2282 
2283         pprevStartCode = prevStartCode;
2284         prevStartCode = currentStartCode;
2285         currentStartCode = data[offset + 3];
2286 
2287         if (currentStartCode == 0xb3 && mFormat == NULL) {
2288             memmove(mBuffer->data(), mBuffer->data() + offset, size - offset);
2289             size -= offset;
2290             (void)fetchTimestamp(offset);
2291             offset = 0;
2292             mBuffer->setRange(0, size);
2293         }
2294 
2295         if ((prevStartCode == 0xb3 && currentStartCode != 0xb5)
2296                 || (pprevStartCode == 0xb3 && prevStartCode == 0xb5)) {
2297             // seqHeader without/with extension
2298 
2299             if (mFormat == NULL) {
2300                 if (size < 7u) {
2301                     ALOGE("Size too small");
2302                     return NULL;
2303                 }
2304 
2305                 unsigned width =
2306                     (data[4] << 4) | data[5] >> 4;
2307 
2308                 unsigned height =
2309                     ((data[5] & 0x0f) << 8) | data[6];
2310 
2311                 mFormat = new MetaData;
2312                 mFormat->setCString(kKeyMIMEType, MEDIA_MIMETYPE_VIDEO_MPEG2);
2313                 mFormat->setInt32(kKeyWidth, width);
2314                 mFormat->setInt32(kKeyHeight, height);
2315 
2316                 ALOGI("found MPEG2 video codec config (%d x %d)", width, height);
2317 
2318                 sp<ABuffer> csd = new ABuffer(offset);
2319                 memcpy(csd->data(), data, offset);
2320 
2321                 memmove(mBuffer->data(),
2322                         mBuffer->data() + offset,
2323                         mBuffer->size() - offset);
2324 
2325                 mBuffer->setRange(0, mBuffer->size() - offset);
2326                 size -= offset;
2327                 (void)fetchTimestamp(offset);
2328                 offset = 0;
2329 
2330                 // hexdump(csd->data(), csd->size());
2331 
2332                 sp<ABuffer> esds = MakeMPEGVideoESDS(csd);
2333                 mFormat->setData(
2334                         kKeyESDS, kTypeESDS, esds->data(), esds->size());
2335 
2336                 return NULL;
2337             }
2338         }
2339 
2340         if (mFormat != NULL && currentStartCode == 0xb8) {
2341             // GOP layer
2342             if (offset + 7 >= size) {
2343                 ALOGE("Size too small");
2344                 return NULL;
2345             }
2346             gopFound = true;
2347             isClosedGop = (data[offset + 7] & 0x40) != 0;
2348             brokenLink = (data[offset + 7] & 0x20) != 0;
2349         }
2350 
2351         if (mFormat != NULL && currentStartCode == 0xb2) {
2352             userDataPositions.add(offset);
2353         }
2354 
2355         if (mFormat != NULL && currentStartCode == 0x00) {
2356             // Picture start
2357 
2358             if (!sawPictureStart) {
2359                 sawPictureStart = true;
2360             } else {
2361                 sp<ABuffer> accessUnit = new ABuffer(offset);
2362                 memcpy(accessUnit->data(), data, offset);
2363 
2364                 memmove(mBuffer->data(),
2365                         mBuffer->data() + offset,
2366                         mBuffer->size() - offset);
2367 
2368                 mBuffer->setRange(0, mBuffer->size() - offset);
2369 
2370                 int64_t timeUs = fetchTimestamp(offset);
2371                 if (timeUs < 0LL) {
2372                     ALOGE("Negative timeUs");
2373                     return NULL;
2374                 }
2375 
2376                 offset = 0;
2377 
2378                 accessUnit->meta()->setInt64("timeUs", timeUs);
2379                 if (gopFound && (!brokenLink || isClosedGop)) {
2380                     accessUnit->meta()->setInt32("isSync", 1);
2381                 }
2382 
2383                 ALOGV("returning MPEG video access unit at time %" PRId64 " us",
2384                       timeUs);
2385 
2386                 // hexdump(accessUnit->data(), accessUnit->size());
2387 
2388                 if (userDataPositions.size() > 0) {
2389                     sp<ABuffer> mpegUserData =
2390                         new ABuffer(userDataPositions.size() * sizeof(size_t));
2391                     if (mpegUserData != NULL && mpegUserData->data() != NULL) {
2392                         for (size_t i = 0; i < userDataPositions.size(); ++i) {
2393                             memcpy(
2394                                     mpegUserData->data() + i * sizeof(size_t),
2395                                     &userDataPositions[i], sizeof(size_t));
2396                         }
2397                         accessUnit->meta()->setBuffer("mpeg-user-data", mpegUserData);
2398                     }
2399                 }
2400 
2401                 return accessUnit;
2402             }
2403         }
2404 
2405         ++offset;
2406     }
2407 
2408     return NULL;
2409 }
2410 
getNextChunkSize(const uint8_t * data,size_t size)2411 static ssize_t getNextChunkSize(
2412         const uint8_t *data, size_t size) {
2413     static const char kStartCode[] = "\x00\x00\x01";
2414 
2415     // per ISO/IEC 14496-2 6.2.1, a chunk has a 3-byte prefix + 1-byte start code
2416     // we need at least <prefix><start><next prefix> to successfully scan
2417     if (size < 3 + 1 + 3) {
2418         return -EAGAIN;
2419     }
2420 
2421     if (memcmp(kStartCode, data, 3)) {
2422         return -EAGAIN;
2423     }
2424 
2425     size_t offset = 4;
2426     while (offset + 2 < size) {
2427         if (!memcmp(&data[offset], kStartCode, 3)) {
2428             return offset;
2429         }
2430 
2431         ++offset;
2432     }
2433 
2434     return -EAGAIN;
2435 }
2436 
dequeueAccessUnitMPEG4Video()2437 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitMPEG4Video() {
2438     uint8_t *data = mBuffer->data();
2439     size_t size = mBuffer->size();
2440 
2441     enum {
2442         SKIP_TO_VISUAL_OBJECT_SEQ_START,
2443         EXPECT_VISUAL_OBJECT_START,
2444         EXPECT_VO_START,
2445         EXPECT_VOL_START,
2446         WAIT_FOR_VOP_START,
2447         SKIP_TO_VOP_START,
2448 
2449     } state;
2450 
2451     if (mFormat == NULL) {
2452         state = SKIP_TO_VISUAL_OBJECT_SEQ_START;
2453     } else {
2454         state = SKIP_TO_VOP_START;
2455     }
2456 
2457     int32_t width = -1, height = -1;
2458 
2459     size_t offset = 0;
2460     ssize_t chunkSize;
2461     while ((chunkSize = getNextChunkSize(
2462                     &data[offset], size - offset)) > 0) {
2463         bool discard = false;
2464 
2465         unsigned chunkType = data[offset + 3];
2466 
2467         switch (state) {
2468             case SKIP_TO_VISUAL_OBJECT_SEQ_START:
2469             {
2470                 if (chunkType == 0xb0) {
2471                     // Discard anything before this marker.
2472 
2473                     state = EXPECT_VISUAL_OBJECT_START;
2474                 } else {
2475                     discard = true;
2476                     offset += chunkSize;
2477                     ALOGW("b/74114680, advance to next chunk");
2478                     android_errorWriteLog(0x534e4554, "74114680");
2479                 }
2480                 break;
2481             }
2482 
2483             case EXPECT_VISUAL_OBJECT_START:
2484             {
2485                 if (chunkType != 0xb5) {
2486                     ALOGE("Unexpected chunkType");
2487                     return NULL;
2488                 }
2489                 state = EXPECT_VO_START;
2490                 break;
2491             }
2492 
2493             case EXPECT_VO_START:
2494             {
2495                 if (chunkType > 0x1f) {
2496                     ALOGE("Unexpected chunkType");
2497                     return NULL;
2498                 }
2499                 state = EXPECT_VOL_START;
2500                 break;
2501             }
2502 
2503             case EXPECT_VOL_START:
2504             {
2505                 if ((chunkType & 0xf0) != 0x20) {
2506                     ALOGE("Wrong chunkType");
2507                     return NULL;
2508                 }
2509 
2510                 if (!ExtractDimensionsFromVOLHeader(
2511                             &data[offset], chunkSize,
2512                             &width, &height)) {
2513                     ALOGE("Failed to get dimension");
2514                     return NULL;
2515                 }
2516 
2517                 state = WAIT_FOR_VOP_START;
2518                 break;
2519             }
2520 
2521             case WAIT_FOR_VOP_START:
2522             {
2523                 if (chunkType == 0xb3 || chunkType == 0xb6) {
2524                     // group of VOP or VOP start.
2525 
2526                     mFormat = new MetaData;
2527                     mFormat->setCString(
2528                             kKeyMIMEType, MEDIA_MIMETYPE_VIDEO_MPEG4);
2529 
2530                     mFormat->setInt32(kKeyWidth, width);
2531                     mFormat->setInt32(kKeyHeight, height);
2532 
2533                     ALOGI("found MPEG4 video codec config (%d x %d)",
2534                          width, height);
2535 
2536                     sp<ABuffer> csd = new ABuffer(offset);
2537                     memcpy(csd->data(), data, offset);
2538 
2539                     // hexdump(csd->data(), csd->size());
2540 
2541                     sp<ABuffer> esds = MakeMPEGVideoESDS(csd);
2542                     mFormat->setData(
2543                             kKeyESDS, kTypeESDS,
2544                             esds->data(), esds->size());
2545 
2546                     discard = true;
2547                     state = SKIP_TO_VOP_START;
2548                 }
2549 
2550                 break;
2551             }
2552 
2553             case SKIP_TO_VOP_START:
2554             {
2555                 if (chunkType == 0xb6) {
2556                     int vopCodingType = (data[offset + 4] & 0xc0) >> 6;
2557 
2558                     offset += chunkSize;
2559 
2560                     sp<ABuffer> accessUnit = new ABuffer(offset);
2561                     memcpy(accessUnit->data(), data, offset);
2562 
2563                     memmove(data, &data[offset], size - offset);
2564                     size -= offset;
2565                     mBuffer->setRange(0, size);
2566 
2567                     int64_t timeUs = fetchTimestamp(offset);
2568                     if (timeUs < 0LL) {
2569                         ALOGE("Negative timeus");
2570                         return NULL;
2571                     }
2572 
2573                     offset = 0;
2574 
2575                     accessUnit->meta()->setInt64("timeUs", timeUs);
2576                     if (vopCodingType == 0) {  // intra-coded VOP
2577                         accessUnit->meta()->setInt32("isSync", 1);
2578                     }
2579 
2580                     ALOGV("returning MPEG4 video access unit at time %" PRId64 " us",
2581                          timeUs);
2582 
2583                     // hexdump(accessUnit->data(), accessUnit->size());
2584 
2585                     return accessUnit;
2586                 } else if (chunkType != 0xb3) {
2587                     offset += chunkSize;
2588                     discard = true;
2589                 }
2590 
2591                 break;
2592             }
2593 
2594             default:
2595                 ALOGE("Unknown state: %d", state);
2596                 return NULL;
2597         }
2598 
2599         if (discard) {
2600             (void)fetchTimestamp(offset);
2601             memmove(data, &data[offset], size - offset);
2602             size -= offset;
2603             offset = 0;
2604             mBuffer->setRange(0, size);
2605         } else {
2606             offset += chunkSize;
2607         }
2608     }
2609 
2610     return NULL;
2611 }
2612 
signalEOS()2613 void ElementaryStreamQueue::signalEOS() {
2614     if (!mEOSReached) {
2615         if (mMode == MPEG_VIDEO) {
2616             const char *theEnd = "\x00\x00\x01\x00";
2617             appendData(theEnd, 4, 0);
2618         }
2619         mEOSReached = true;
2620     } else {
2621         ALOGW("EOS already signaled");
2622     }
2623 }
2624 
dequeueAccessUnitMetadata()2625 sp<ABuffer> ElementaryStreamQueue::dequeueAccessUnitMetadata() {
2626     size_t size = mBuffer->size();
2627     if (!size) {
2628         return NULL;
2629     }
2630 
2631     sp<ABuffer> accessUnit = new ABuffer(size);
2632     int64_t timeUs = fetchTimestamp(size);
2633     accessUnit->meta()->setInt64("timeUs", timeUs);
2634 
2635     memcpy(accessUnit->data(), mBuffer->data(), size);
2636     mBuffer->setRange(0, 0);
2637 
2638     if (mFormat == NULL) {
2639         mFormat = new MetaData;
2640         mFormat->setCString(kKeyMIMEType, MEDIA_MIMETYPE_DATA_TIMED_ID3);
2641     }
2642 
2643     return accessUnit;
2644 }
2645 
signalNewSampleAesKey(const sp<AMessage> & keyItem)2646 void ElementaryStreamQueue::signalNewSampleAesKey(const sp<AMessage> &keyItem) {
2647     if (mSampleDecryptor == NULL) {
2648         ALOGE("signalNewSampleAesKey: Stream %x is not encrypted; keyItem: %p",
2649                 mMode, keyItem.get());
2650         return;
2651     }
2652 
2653     mSampleDecryptor->signalNewSampleAesKey(keyItem);
2654 }
2655 
2656 
2657 }  // namespace android
2658