Upstream version 10.39.225.0
[platform/framework/web/crosswalk.git] / src / media / formats / mp4 / track_run_iterator.cc
1 // Copyright 2014 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4
5 #include "media/formats/mp4/track_run_iterator.h"
6
7 #include <algorithm>
8
9 #include "media/base/buffers.h"
10 #include "media/formats/mp4/rcheck.h"
11 #include "media/formats/mp4/sample_to_group_iterator.h"
12
13 namespace media {
14 namespace mp4 {
15
16 struct SampleInfo {
17   int size;
18   int duration;
19   int cts_offset;
20   bool is_keyframe;
21   bool is_random_access_point;
22   uint32 cenc_group_description_index;
23 };
24
25 struct TrackRunInfo {
26   uint32 track_id;
27   std::vector<SampleInfo> samples;
28   int64 timescale;
29   int64 start_dts;
30   int64 sample_start_offset;
31
32   bool is_audio;
33   const AudioSampleEntry* audio_description;
34   const VideoSampleEntry* video_description;
35
36   int64 aux_info_start_offset;  // Only valid if aux_info_total_size > 0.
37   int aux_info_default_size;
38   std::vector<uint8> aux_info_sizes;  // Populated if default_size == 0.
39   int aux_info_total_size;
40
41   std::vector<CencSampleEncryptionInfoEntry> sample_encryption_info;
42
43   TrackRunInfo();
44   ~TrackRunInfo();
45 };
46
47 TrackRunInfo::TrackRunInfo()
48     : track_id(0),
49       timescale(-1),
50       start_dts(-1),
51       sample_start_offset(-1),
52       is_audio(false),
53       aux_info_start_offset(-1),
54       aux_info_default_size(-1),
55       aux_info_total_size(-1) {
56 }
57 TrackRunInfo::~TrackRunInfo() {}
58
59 base::TimeDelta TimeDeltaFromRational(int64 numer, int64 denom) {
60   // To avoid overflow, split the following calculation:
61   // (numer * base::Time::kMicrosecondsPerSecond) / denom
62   // into:
63   //  (numer / denom) * base::Time::kMicrosecondsPerSecond +
64   // ((numer % denom) * base::Time::kMicrosecondsPerSecond) / denom
65   int64 a = numer / denom;
66   DCHECK_LE((a > 0 ? a : -a), kint64max / base::Time::kMicrosecondsPerSecond);
67   int64 timea_in_us = a * base::Time::kMicrosecondsPerSecond;
68
69   int64 b = numer % denom;
70   DCHECK_LE((b > 0 ? b : -b), kint64max / base::Time::kMicrosecondsPerSecond);
71   int64 timeb_in_us = (b * base::Time::kMicrosecondsPerSecond) / denom;
72
73   DCHECK((timeb_in_us < 0) || (timea_in_us <= kint64max - timeb_in_us));
74   DCHECK((timeb_in_us > 0) || (timea_in_us >= kint64min - timeb_in_us));
75   return base::TimeDelta::FromMicroseconds(timea_in_us + timeb_in_us);
76 }
77
78 DecodeTimestamp DecodeTimestampFromRational(int64 numer, int64 denom) {
79   return DecodeTimestamp::FromPresentationTime(
80       TimeDeltaFromRational(numer, denom));
81 }
82
83 TrackRunIterator::TrackRunIterator(const Movie* moov,
84                                    const LogCB& log_cb)
85     : moov_(moov), log_cb_(log_cb), sample_offset_(0) {
86   CHECK(moov);
87 }
88
89 TrackRunIterator::~TrackRunIterator() {}
90
91 static bool PopulateSampleInfo(const TrackExtends& trex,
92                                const TrackFragmentHeader& tfhd,
93                                const TrackFragmentRun& trun,
94                                const int64 edit_list_offset,
95                                const uint32 i,
96                                SampleInfo* sample_info,
97                                const SampleDependsOn sdtp_sample_depends_on,
98                                const LogCB& log_cb) {
99   if (i < trun.sample_sizes.size()) {
100     sample_info->size = trun.sample_sizes[i];
101   } else if (tfhd.default_sample_size > 0) {
102     sample_info->size = tfhd.default_sample_size;
103   } else {
104     sample_info->size = trex.default_sample_size;
105   }
106
107   if (i < trun.sample_durations.size()) {
108     sample_info->duration = trun.sample_durations[i];
109   } else if (tfhd.default_sample_duration > 0) {
110     sample_info->duration = tfhd.default_sample_duration;
111   } else {
112     sample_info->duration = trex.default_sample_duration;
113   }
114
115   if (i < trun.sample_composition_time_offsets.size()) {
116     sample_info->cts_offset = trun.sample_composition_time_offsets[i];
117   } else {
118     sample_info->cts_offset = 0;
119   }
120   sample_info->cts_offset += edit_list_offset;
121
122   uint32 flags;
123   if (i < trun.sample_flags.size()) {
124     flags = trun.sample_flags[i];
125   } else if (tfhd.has_default_sample_flags) {
126     flags = tfhd.default_sample_flags;
127   } else {
128     flags = trex.default_sample_flags;
129   }
130
131   SampleDependsOn sample_depends_on =
132       static_cast<SampleDependsOn>((flags >> 24) & 0x3);
133
134   if (sample_depends_on == kSampleDependsOnUnknown)
135     sample_depends_on = sdtp_sample_depends_on;
136
137   // ISO/IEC 14496-12  Section 8.8.3.1 : The negation of |sample_is_sync_sample|
138   // provides the same information as the sync sample table [8.6.2]. When
139   // |sample_is_sync_sample| is true for a sample, it is the same as if the
140   // sample were not in a movie fragment and marked with an entry in the sync
141   // sample table (or, if all samples are sync samples, the sync sample table
142   // were absent).
143   bool sample_is_sync_sample = !(flags & kSampleIsNonSyncSample);
144   sample_info->is_random_access_point = sample_is_sync_sample;
145
146   switch (sample_depends_on) {
147     case kSampleDependsOnUnknown:
148       sample_info->is_keyframe = sample_is_sync_sample;
149       break;
150
151     case kSampleDependsOnOthers:
152       sample_info->is_keyframe = false;
153       break;
154
155     case kSampleDependsOnNoOther:
156       sample_info->is_keyframe = true;
157       break;
158
159     case kSampleDependsOnReserved:
160       MEDIA_LOG(log_cb) << "Reserved value used in sample dependency info.";
161       return false;
162   }
163   return true;
164 }
165
166 // In well-structured encrypted media, each track run will be immediately
167 // preceded by its auxiliary information; this is the only optimal storage
168 // pattern in terms of minimum number of bytes from a serial stream needed to
169 // begin playback. It also allows us to optimize caching on memory-constrained
170 // architectures, because we can cache the relatively small auxiliary
171 // information for an entire run and then discard data from the input stream,
172 // instead of retaining the entire 'mdat' box.
173 //
174 // We optimize for this situation (with no loss of generality) by sorting track
175 // runs during iteration in order of their first data offset (either sample data
176 // or auxiliary data).
177 class CompareMinTrackRunDataOffset {
178  public:
179   bool operator()(const TrackRunInfo& a, const TrackRunInfo& b) {
180     int64 a_aux = a.aux_info_total_size ? a.aux_info_start_offset : kint64max;
181     int64 b_aux = b.aux_info_total_size ? b.aux_info_start_offset : kint64max;
182
183     int64 a_lesser = std::min(a_aux, a.sample_start_offset);
184     int64 a_greater = std::max(a_aux, a.sample_start_offset);
185     int64 b_lesser = std::min(b_aux, b.sample_start_offset);
186     int64 b_greater = std::max(b_aux, b.sample_start_offset);
187
188     if (a_lesser == b_lesser) return a_greater < b_greater;
189     return a_lesser < b_lesser;
190   }
191 };
192
193 bool TrackRunIterator::Init(const MovieFragment& moof) {
194   runs_.clear();
195
196   for (size_t i = 0; i < moof.tracks.size(); i++) {
197     const TrackFragment& traf = moof.tracks[i];
198
199     const Track* trak = NULL;
200     for (size_t t = 0; t < moov_->tracks.size(); t++) {
201       if (moov_->tracks[t].header.track_id == traf.header.track_id)
202         trak = &moov_->tracks[t];
203     }
204     RCHECK(trak);
205
206     const TrackExtends* trex = NULL;
207     for (size_t t = 0; t < moov_->extends.tracks.size(); t++) {
208       if (moov_->extends.tracks[t].track_id == traf.header.track_id)
209         trex = &moov_->extends.tracks[t];
210     }
211     RCHECK(trex);
212
213     const SampleDescription& stsd =
214         trak->media.information.sample_table.description;
215     if (stsd.type != kAudio && stsd.type != kVideo) {
216       DVLOG(1) << "Skipping unhandled track type";
217       continue;
218     }
219     size_t desc_idx = traf.header.sample_description_index;
220     if (!desc_idx) desc_idx = trex->default_sample_description_index;
221     RCHECK(desc_idx > 0);  // Descriptions are one-indexed in the file
222     desc_idx -= 1;
223
224     // Process edit list to remove CTS offset introduced in the presence of
225     // B-frames (those that contain a single edit with a nonnegative media
226     // time). Other uses of edit lists are not supported, as they are
227     // both uncommon and better served by higher-level protocols.
228     int64 edit_list_offset = 0;
229     const std::vector<EditListEntry>& edits = trak->edit.list.edits;
230     if (!edits.empty()) {
231       if (edits.size() > 1)
232         DVLOG(1) << "Multi-entry edit box detected; some components ignored.";
233
234       if (edits[0].media_time < 0) {
235         DVLOG(1) << "Empty edit list entry ignored.";
236       } else {
237         edit_list_offset = -edits[0].media_time;
238       }
239     }
240
241     SampleToGroupIterator sample_to_group_itr(traf.sample_to_group);
242     bool is_sample_to_group_valid = sample_to_group_itr.IsValid();
243
244     int64 run_start_dts = traf.decode_time.decode_time;
245     int sample_count_sum = 0;
246     for (size_t j = 0; j < traf.runs.size(); j++) {
247       const TrackFragmentRun& trun = traf.runs[j];
248       TrackRunInfo tri;
249       tri.track_id = traf.header.track_id;
250       tri.timescale = trak->media.header.timescale;
251       tri.start_dts = run_start_dts;
252       tri.sample_start_offset = trun.data_offset;
253       tri.sample_encryption_info = traf.sample_group_description.entries;
254
255       tri.is_audio = (stsd.type == kAudio);
256       if (tri.is_audio) {
257         RCHECK(!stsd.audio_entries.empty());
258         if (desc_idx > stsd.audio_entries.size())
259           desc_idx = 0;
260         tri.audio_description = &stsd.audio_entries[desc_idx];
261       } else {
262         RCHECK(!stsd.video_entries.empty());
263         if (desc_idx > stsd.video_entries.size())
264           desc_idx = 0;
265         tri.video_description = &stsd.video_entries[desc_idx];
266       }
267
268       // Collect information from the auxiliary_offset entry with the same index
269       // in the 'saiz' container as the current run's index in the 'trun'
270       // container, if it is present.
271       if (traf.auxiliary_offset.offsets.size() > j) {
272         // There should be an auxiliary info entry corresponding to each sample
273         // in the auxiliary offset entry's corresponding track run.
274         RCHECK(traf.auxiliary_size.sample_count >=
275                sample_count_sum + trun.sample_count);
276         tri.aux_info_start_offset = traf.auxiliary_offset.offsets[j];
277         tri.aux_info_default_size =
278             traf.auxiliary_size.default_sample_info_size;
279         if (tri.aux_info_default_size == 0) {
280           const std::vector<uint8>& sizes =
281               traf.auxiliary_size.sample_info_sizes;
282           tri.aux_info_sizes.insert(tri.aux_info_sizes.begin(),
283               sizes.begin() + sample_count_sum,
284               sizes.begin() + sample_count_sum + trun.sample_count);
285         }
286
287         // If the default info size is positive, find the total size of the aux
288         // info block from it, otherwise sum over the individual sizes of each
289         // aux info entry in the aux_offset entry.
290         if (tri.aux_info_default_size) {
291           tri.aux_info_total_size =
292               tri.aux_info_default_size * trun.sample_count;
293         } else {
294           tri.aux_info_total_size = 0;
295           for (size_t k = 0; k < trun.sample_count; k++) {
296             tri.aux_info_total_size += tri.aux_info_sizes[k];
297           }
298         }
299       } else {
300         tri.aux_info_start_offset = -1;
301         tri.aux_info_total_size = 0;
302       }
303
304       tri.samples.resize(trun.sample_count);
305       for (size_t k = 0; k < trun.sample_count; k++) {
306         if (!PopulateSampleInfo(*trex, traf.header, trun, edit_list_offset,
307                                 k, &tri.samples[k],
308                                 traf.sdtp.sample_depends_on(k),
309                                 log_cb_)) {
310           return false;
311         }
312
313         run_start_dts += tri.samples[k].duration;
314
315         if (!is_sample_to_group_valid) {
316           // Set group description index to 0 to read encryption information
317           // from TrackEncryption Box.
318           tri.samples[k].cenc_group_description_index = 0;
319           continue;
320         }
321
322         // ISO-14496-12 Section 8.9.2.3 and 8.9.4 : group description index
323         // (1) ranges from 1 to the number of sample group entries in the track
324         // level SampleGroupDescription Box, or (2) takes the value 0 to
325         // indicate that this sample is a member of no group, in this case, the
326         // sample is associated with the default values specified in
327         // TrackEncryption Box, or (3) starts at 0x10001, i.e. the index value
328         // 1, with the value 1 in the top 16 bits, to reference fragment-local
329         // SampleGroupDescription Box.
330         // Case (1) is not supported currently. We might not need it either as
331         // the same functionality can be better achieved using (2).
332         uint32 index = sample_to_group_itr.group_description_index();
333         if (index >= SampleToGroupEntry::kFragmentGroupDescriptionIndexBase) {
334           index -= SampleToGroupEntry::kFragmentGroupDescriptionIndexBase;
335           RCHECK(index != 0 && index <= tri.sample_encryption_info.size());
336         } else if (index != 0) {
337           NOTIMPLEMENTED() << "'sgpd' box in 'moov' is not supported.";
338           return false;
339         }
340         tri.samples[k].cenc_group_description_index = index;
341         is_sample_to_group_valid = sample_to_group_itr.Advance();
342       }
343       runs_.push_back(tri);
344       sample_count_sum += trun.sample_count;
345     }
346
347     // We should have iterated through all samples in SampleToGroup Box.
348     RCHECK(!sample_to_group_itr.IsValid());
349   }
350
351   std::sort(runs_.begin(), runs_.end(), CompareMinTrackRunDataOffset());
352   run_itr_ = runs_.begin();
353   ResetRun();
354   return true;
355 }
356
357 void TrackRunIterator::AdvanceRun() {
358   ++run_itr_;
359   ResetRun();
360 }
361
362 void TrackRunIterator::ResetRun() {
363   if (!IsRunValid()) return;
364   sample_dts_ = run_itr_->start_dts;
365   sample_offset_ = run_itr_->sample_start_offset;
366   sample_itr_ = run_itr_->samples.begin();
367   cenc_info_.clear();
368 }
369
370 void TrackRunIterator::AdvanceSample() {
371   DCHECK(IsSampleValid());
372   sample_dts_ += sample_itr_->duration;
373   sample_offset_ += sample_itr_->size;
374   ++sample_itr_;
375 }
376
377 // This implementation only indicates a need for caching if CENC auxiliary
378 // info is available in the stream.
379 bool TrackRunIterator::AuxInfoNeedsToBeCached() {
380   DCHECK(IsRunValid());
381   return aux_info_size() > 0 && cenc_info_.size() == 0;
382 }
383
384 // This implementation currently only caches CENC auxiliary info.
385 bool TrackRunIterator::CacheAuxInfo(const uint8* buf, int buf_size) {
386   RCHECK(AuxInfoNeedsToBeCached() && buf_size >= aux_info_size());
387
388   cenc_info_.resize(run_itr_->samples.size());
389   int64 pos = 0;
390   for (size_t i = 0; i < run_itr_->samples.size(); i++) {
391     int info_size = run_itr_->aux_info_default_size;
392     if (!info_size)
393       info_size = run_itr_->aux_info_sizes[i];
394
395     if (IsSampleEncrypted(i)) {
396       BufferReader reader(buf + pos, info_size);
397       RCHECK(cenc_info_[i].Parse(GetIvSize(i), &reader));
398     }
399     pos += info_size;
400   }
401
402   return true;
403 }
404
405 bool TrackRunIterator::IsRunValid() const {
406   return run_itr_ != runs_.end();
407 }
408
409 bool TrackRunIterator::IsSampleValid() const {
410   return IsRunValid() && (sample_itr_ != run_itr_->samples.end());
411 }
412
413 // Because tracks are in sorted order and auxiliary information is cached when
414 // returning samples, it is guaranteed that no data will be required before the
415 // lesser of the minimum data offset of this track and the next in sequence.
416 // (The stronger condition - that no data is required before the minimum data
417 // offset of this track alone - is not guaranteed, because the BMFF spec does
418 // not have any inter-run ordering restrictions.)
419 int64 TrackRunIterator::GetMaxClearOffset() {
420   int64 offset = kint64max;
421
422   if (IsSampleValid()) {
423     offset = std::min(offset, sample_offset_);
424     if (AuxInfoNeedsToBeCached())
425       offset = std::min(offset, aux_info_offset());
426   }
427   if (run_itr_ != runs_.end()) {
428     std::vector<TrackRunInfo>::const_iterator next_run = run_itr_ + 1;
429     if (next_run != runs_.end()) {
430       offset = std::min(offset, next_run->sample_start_offset);
431       if (next_run->aux_info_total_size)
432         offset = std::min(offset, next_run->aux_info_start_offset);
433     }
434   }
435   if (offset == kint64max) return 0;
436   return offset;
437 }
438
439 uint32 TrackRunIterator::track_id() const {
440   DCHECK(IsRunValid());
441   return run_itr_->track_id;
442 }
443
444 bool TrackRunIterator::is_encrypted() const {
445   DCHECK(IsSampleValid());
446   return IsSampleEncrypted(sample_itr_ - run_itr_->samples.begin());
447 }
448
449 int64 TrackRunIterator::aux_info_offset() const {
450   return run_itr_->aux_info_start_offset;
451 }
452
453 int TrackRunIterator::aux_info_size() const {
454   return run_itr_->aux_info_total_size;
455 }
456
457 bool TrackRunIterator::is_audio() const {
458   DCHECK(IsRunValid());
459   return run_itr_->is_audio;
460 }
461
462 const AudioSampleEntry& TrackRunIterator::audio_description() const {
463   DCHECK(is_audio());
464   DCHECK(run_itr_->audio_description);
465   return *run_itr_->audio_description;
466 }
467
468 const VideoSampleEntry& TrackRunIterator::video_description() const {
469   DCHECK(!is_audio());
470   DCHECK(run_itr_->video_description);
471   return *run_itr_->video_description;
472 }
473
474 int64 TrackRunIterator::sample_offset() const {
475   DCHECK(IsSampleValid());
476   return sample_offset_;
477 }
478
479 int TrackRunIterator::sample_size() const {
480   DCHECK(IsSampleValid());
481   return sample_itr_->size;
482 }
483
484 DecodeTimestamp TrackRunIterator::dts() const {
485   DCHECK(IsSampleValid());
486   return DecodeTimestampFromRational(sample_dts_, run_itr_->timescale);
487 }
488
489 base::TimeDelta TrackRunIterator::cts() const {
490   DCHECK(IsSampleValid());
491   return TimeDeltaFromRational(sample_dts_ + sample_itr_->cts_offset,
492                                run_itr_->timescale);
493 }
494
495 base::TimeDelta TrackRunIterator::duration() const {
496   DCHECK(IsSampleValid());
497   return TimeDeltaFromRational(sample_itr_->duration, run_itr_->timescale);
498 }
499
500 bool TrackRunIterator::is_keyframe() const {
501   DCHECK(IsSampleValid());
502   return sample_itr_->is_keyframe;
503 }
504
505 bool TrackRunIterator::is_random_access_point() const {
506   DCHECK(IsSampleValid());
507   return sample_itr_->is_random_access_point;
508 }
509
510 const TrackEncryption& TrackRunIterator::track_encryption() const {
511   if (is_audio())
512     return audio_description().sinf.info.track_encryption;
513   return video_description().sinf.info.track_encryption;
514 }
515
516 scoped_ptr<DecryptConfig> TrackRunIterator::GetDecryptConfig() {
517   DCHECK(is_encrypted());
518
519   if (cenc_info_.empty()) {
520     DCHECK_EQ(0, aux_info_size());
521     MEDIA_LOG(log_cb_) << "Aux Info is not available.";
522     return scoped_ptr<DecryptConfig>();
523   }
524
525   size_t sample_idx = sample_itr_ - run_itr_->samples.begin();
526   DCHECK_LT(sample_idx, cenc_info_.size());
527   const FrameCENCInfo& cenc_info = cenc_info_[sample_idx];
528
529   size_t total_size = 0;
530   if (!cenc_info.subsamples.empty() &&
531       (!cenc_info.GetTotalSizeOfSubsamples(&total_size) ||
532        total_size != static_cast<size_t>(sample_size()))) {
533     MEDIA_LOG(log_cb_) << "Incorrect CENC subsample size.";
534     return scoped_ptr<DecryptConfig>();
535   }
536
537   const std::vector<uint8>& kid = GetKeyId(sample_idx);
538   return scoped_ptr<DecryptConfig>(new DecryptConfig(
539       std::string(reinterpret_cast<const char*>(&kid[0]), kid.size()),
540       std::string(reinterpret_cast<const char*>(cenc_info.iv),
541                   arraysize(cenc_info.iv)),
542       cenc_info.subsamples));
543 }
544
545 uint32 TrackRunIterator::GetGroupDescriptionIndex(uint32 sample_index) const {
546   DCHECK(IsRunValid());
547   DCHECK_LT(sample_index, run_itr_->samples.size());
548   return run_itr_->samples[sample_index].cenc_group_description_index;
549 }
550
551 const CencSampleEncryptionInfoEntry&
552 TrackRunIterator::GetSampleEncryptionInfoEntry(
553     uint32 group_description_index) const {
554   DCHECK(IsRunValid());
555   DCHECK_NE(group_description_index, 0u);
556   DCHECK_LE(group_description_index, run_itr_->sample_encryption_info.size());
557   // |group_description_index| is 1-based. Subtract by 1 to index the vector.
558   return run_itr_->sample_encryption_info[group_description_index - 1];
559 }
560
561 bool TrackRunIterator::IsSampleEncrypted(size_t sample_index) const {
562   uint32 index = GetGroupDescriptionIndex(sample_index);
563   return (index == 0) ? track_encryption().is_encrypted
564                       : GetSampleEncryptionInfoEntry(index).is_encrypted;
565 }
566
567 const std::vector<uint8>& TrackRunIterator::GetKeyId(
568     size_t sample_index) const {
569   uint32 index = GetGroupDescriptionIndex(sample_index);
570   return (index == 0) ? track_encryption().default_kid
571                       : GetSampleEncryptionInfoEntry(index).key_id;
572 }
573
574 uint8 TrackRunIterator::GetIvSize(size_t sample_index) const {
575   uint32 index = GetGroupDescriptionIndex(sample_index);
576   return (index == 0) ? track_encryption().default_iv_size
577                       : GetSampleEncryptionInfoEntry(index).iv_size;
578 }
579
580 }  // namespace mp4
581 }  // namespace media