Revert "Omit calls to set composing region when pasting image."
[chromium-blink-merge.git] / media / formats / mp4 / mp4_stream_parser.cc
blob18698b36c4ab2426d00f1832743e7561df5eb8c1
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.
5 #include "media/formats/mp4/mp4_stream_parser.h"
7 #include "base/callback_helpers.h"
8 #include "base/logging.h"
9 #include "base/time/time.h"
10 #include "media/base/audio_decoder_config.h"
11 #include "media/base/stream_parser_buffer.h"
12 #include "media/base/text_track_config.h"
13 #include "media/base/timestamp_constants.h"
14 #include "media/base/video_decoder_config.h"
15 #include "media/base/video_util.h"
16 #include "media/formats/mp4/box_definitions.h"
17 #include "media/formats/mp4/box_reader.h"
18 #include "media/formats/mp4/es_descriptor.h"
19 #include "media/formats/mp4/rcheck.h"
20 #include "media/formats/mpeg/adts_constants.h"
22 namespace media {
23 namespace mp4 {
25 MP4StreamParser::MP4StreamParser(const std::set<int>& audio_object_types,
26 bool has_sbr)
27 : state_(kWaitingForInit),
28 moof_head_(0),
29 mdat_tail_(0),
30 highest_end_offset_(0),
31 has_audio_(false),
32 has_video_(false),
33 audio_track_id_(0),
34 video_track_id_(0),
35 audio_object_types_(audio_object_types),
36 has_sbr_(has_sbr),
37 is_audio_track_encrypted_(false),
38 is_video_track_encrypted_(false),
39 num_top_level_box_skipped_(0) {
42 MP4StreamParser::~MP4StreamParser() {}
44 void MP4StreamParser::Init(
45 const InitCB& init_cb,
46 const NewConfigCB& config_cb,
47 const NewBuffersCB& new_buffers_cb,
48 bool /* ignore_text_tracks */,
49 const EncryptedMediaInitDataCB& encrypted_media_init_data_cb,
50 const NewMediaSegmentCB& new_segment_cb,
51 const base::Closure& end_of_segment_cb,
52 const scoped_refptr<MediaLog>& media_log) {
53 DCHECK_EQ(state_, kWaitingForInit);
54 DCHECK(init_cb_.is_null());
55 DCHECK(!init_cb.is_null());
56 DCHECK(!config_cb.is_null());
57 DCHECK(!new_buffers_cb.is_null());
58 DCHECK(!encrypted_media_init_data_cb.is_null());
59 DCHECK(!end_of_segment_cb.is_null());
61 ChangeState(kParsingBoxes);
62 init_cb_ = init_cb;
63 config_cb_ = config_cb;
64 new_buffers_cb_ = new_buffers_cb;
65 encrypted_media_init_data_cb_ = encrypted_media_init_data_cb;
66 new_segment_cb_ = new_segment_cb;
67 end_of_segment_cb_ = end_of_segment_cb;
68 media_log_ = media_log;
71 void MP4StreamParser::Reset() {
72 queue_.Reset();
73 runs_.reset();
74 moof_head_ = 0;
75 mdat_tail_ = 0;
78 void MP4StreamParser::Flush() {
79 DCHECK_NE(state_, kWaitingForInit);
80 Reset();
81 ChangeState(kParsingBoxes);
84 bool MP4StreamParser::Parse(const uint8* buf, int size) {
85 DCHECK_NE(state_, kWaitingForInit);
87 if (state_ == kError)
88 return false;
90 queue_.Push(buf, size);
92 BufferQueue audio_buffers;
93 BufferQueue video_buffers;
95 bool result = false;
96 bool err = false;
98 do {
99 switch (state_) {
100 case kWaitingForInit:
101 case kError:
102 NOTREACHED();
103 return false;
105 case kParsingBoxes:
106 result = ParseBox(&err);
107 break;
109 case kWaitingForSampleData:
110 result = HaveEnoughDataToEnqueueSamples();
111 if (result)
112 ChangeState(kEmittingSamples);
113 break;
115 case kEmittingSamples:
116 result = EnqueueSample(&audio_buffers, &video_buffers, &err);
117 if (result) {
118 int64 max_clear = runs_->GetMaxClearOffset() + moof_head_;
119 err = !ReadAndDiscardMDATsUntil(max_clear);
121 break;
123 } while (result && !err);
125 if (!err)
126 err = !SendAndFlushSamples(&audio_buffers, &video_buffers);
128 if (err) {
129 DLOG(ERROR) << "Error while parsing MP4";
130 moov_.reset();
131 Reset();
132 ChangeState(kError);
133 return false;
136 return true;
139 bool MP4StreamParser::ParseBox(bool* err) {
140 const uint8* buf;
141 int size;
142 queue_.Peek(&buf, &size);
143 if (!size) return false;
145 scoped_ptr<BoxReader> reader(
146 BoxReader::ReadTopLevelBox(buf, size, media_log_, err));
147 if (reader.get() == NULL) return false;
149 if (reader->type() == FOURCC_MOOV) {
150 *err = !ParseMoov(reader.get());
151 } else if (reader->type() == FOURCC_MOOF) {
152 moof_head_ = queue_.head();
153 *err = !ParseMoof(reader.get());
155 // Set up first mdat offset for ReadMDATsUntil().
156 mdat_tail_ = queue_.head() + reader->size();
158 // Return early to avoid evicting 'moof' data from queue. Auxiliary info may
159 // be located anywhere in the file, including inside the 'moof' itself.
160 // (Since 'default-base-is-moof' is mandated, no data references can come
161 // before the head of the 'moof', so keeping this box around is sufficient.)
162 return !(*err);
163 } else {
164 // TODO(wolenetz,chcunningham): Enforce more strict adherence to MSE byte
165 // stream spec for ftyp and styp. See http://crbug.com/504514.
166 DVLOG(2) << "Skipping unrecognized top-level box: "
167 << FourCCToString(reader->type());
170 queue_.Pop(reader->size());
171 return !(*err);
174 bool MP4StreamParser::ParseMoov(BoxReader* reader) {
175 moov_.reset(new Movie);
176 RCHECK(moov_->Parse(reader));
177 runs_.reset();
179 has_audio_ = false;
180 has_video_ = false;
182 AudioDecoderConfig audio_config;
183 VideoDecoderConfig video_config;
185 for (std::vector<Track>::const_iterator track = moov_->tracks.begin();
186 track != moov_->tracks.end(); ++track) {
187 // TODO(strobe): Only the first audio and video track present in a file are
188 // used. (Track selection is better accomplished via Source IDs, though, so
189 // adding support for track selection within a stream is low-priority.)
190 const SampleDescription& samp_descr =
191 track->media.information.sample_table.description;
193 // TODO(strobe): When codec reconfigurations are supported, detect and send
194 // a codec reconfiguration for fragments using a sample description index
195 // different from the previous one
196 size_t desc_idx = 0;
197 for (size_t t = 0; t < moov_->extends.tracks.size(); t++) {
198 const TrackExtends& trex = moov_->extends.tracks[t];
199 if (trex.track_id == track->header.track_id) {
200 desc_idx = trex.default_sample_description_index;
201 break;
204 RCHECK(desc_idx > 0);
205 desc_idx -= 1; // BMFF descriptor index is one-based
207 if (track->media.handler.type == kAudio && !audio_config.IsValidConfig()) {
208 RCHECK(!samp_descr.audio_entries.empty());
210 // It is not uncommon to find otherwise-valid files with incorrect sample
211 // description indices, so we fail gracefully in that case.
212 if (desc_idx >= samp_descr.audio_entries.size())
213 desc_idx = 0;
214 const AudioSampleEntry& entry = samp_descr.audio_entries[desc_idx];
215 const AAC& aac = entry.esds.aac;
217 if (!(entry.format == FOURCC_MP4A ||
218 (entry.format == FOURCC_ENCA &&
219 entry.sinf.format.format == FOURCC_MP4A))) {
220 MEDIA_LOG(ERROR, media_log_) << "Unsupported audio format 0x"
221 << std::hex << entry.format
222 << " in stsd box.";
223 return false;
226 uint8 audio_type = entry.esds.object_type;
227 DVLOG(1) << "audio_type " << std::hex << static_cast<int>(audio_type);
228 if (audio_object_types_.find(audio_type) == audio_object_types_.end()) {
229 MEDIA_LOG(ERROR, media_log_)
230 << "audio object type 0x" << std::hex << audio_type
231 << " does not match what is specified in the"
232 << " mimetype.";
233 return false;
236 AudioCodec codec = kUnknownAudioCodec;
237 ChannelLayout channel_layout = CHANNEL_LAYOUT_NONE;
238 int sample_per_second = 0;
239 std::vector<uint8> extra_data;
240 // Check if it is MPEG4 AAC defined in ISO 14496 Part 3 or
241 // supported MPEG2 AAC varients.
242 if (ESDescriptor::IsAAC(audio_type)) {
243 codec = kCodecAAC;
244 channel_layout = aac.GetChannelLayout(has_sbr_);
245 sample_per_second = aac.GetOutputSamplesPerSecond(has_sbr_);
246 #if defined(OS_ANDROID)
247 extra_data = aac.codec_specific_data();
248 #endif
249 } else {
250 MEDIA_LOG(ERROR, media_log_) << "Unsupported audio object type 0x"
251 << std::hex << audio_type << " in esds.";
252 return false;
255 SampleFormat sample_format;
256 if (entry.samplesize == 8) {
257 sample_format = kSampleFormatU8;
258 } else if (entry.samplesize == 16) {
259 sample_format = kSampleFormatS16;
260 } else if (entry.samplesize == 32) {
261 sample_format = kSampleFormatS32;
262 } else {
263 LOG(ERROR) << "Unsupported sample size.";
264 return false;
267 is_audio_track_encrypted_ = entry.sinf.info.track_encryption.is_encrypted;
268 DVLOG(1) << "is_audio_track_encrypted_: " << is_audio_track_encrypted_;
269 audio_config.Initialize(
270 codec, sample_format, channel_layout, sample_per_second,
271 extra_data.size() ? &extra_data[0] : NULL, extra_data.size(),
272 is_audio_track_encrypted_, base::TimeDelta(), 0);
273 has_audio_ = true;
274 audio_track_id_ = track->header.track_id;
276 if (track->media.handler.type == kVideo && !video_config.IsValidConfig()) {
277 RCHECK(!samp_descr.video_entries.empty());
278 if (desc_idx >= samp_descr.video_entries.size())
279 desc_idx = 0;
280 const VideoSampleEntry& entry = samp_descr.video_entries[desc_idx];
282 if (!entry.IsFormatValid()) {
283 MEDIA_LOG(ERROR, media_log_) << "Unsupported video format 0x"
284 << std::hex << entry.format
285 << " in stsd box.";
286 return false;
289 // TODO(strobe): Recover correct crop box
290 gfx::Size coded_size(entry.width, entry.height);
291 gfx::Rect visible_rect(coded_size);
293 // If PASP is available, use the coded size and PASP to calculate the
294 // natural size. Otherwise, use the size in track header for natural size.
295 gfx::Size natural_size(visible_rect.size());
296 if (entry.pixel_aspect.h_spacing != 1 ||
297 entry.pixel_aspect.v_spacing != 1) {
298 natural_size =
299 GetNaturalSize(visible_rect.size(), entry.pixel_aspect.h_spacing,
300 entry.pixel_aspect.v_spacing);
301 } else if (track->header.width && track->header.height) {
302 natural_size =
303 gfx::Size(track->header.width, track->header.height);
306 is_video_track_encrypted_ = entry.sinf.info.track_encryption.is_encrypted;
307 DVLOG(1) << "is_video_track_encrypted_: " << is_video_track_encrypted_;
308 video_config.Initialize(entry.video_codec, entry.video_codec_profile,
309 PIXEL_FORMAT_YV12, COLOR_SPACE_HD_REC709,
310 coded_size, visible_rect, natural_size,
311 // No decoder-specific buffer needed for AVC;
312 // SPS/PPS are embedded in the video stream
313 NULL, 0, is_video_track_encrypted_);
314 has_video_ = true;
315 video_track_id_ = track->header.track_id;
319 RCHECK(config_cb_.Run(audio_config, video_config, TextTrackConfigMap()));
321 StreamParser::InitParameters params(kInfiniteDuration());
322 if (moov_->extends.header.fragment_duration > 0) {
323 params.duration = TimeDeltaFromRational(
324 moov_->extends.header.fragment_duration, moov_->header.timescale);
325 params.liveness = DemuxerStream::LIVENESS_RECORDED;
326 } else if (moov_->header.duration > 0 &&
327 moov_->header.duration != kuint64max) {
328 params.duration =
329 TimeDeltaFromRational(moov_->header.duration, moov_->header.timescale);
330 params.liveness = DemuxerStream::LIVENESS_RECORDED;
331 } else {
332 // In ISO/IEC 14496-12:2005(E), 8.30.2: ".. If an MP4 file is created in
333 // real-time, such as used in live streaming, it is not likely that the
334 // fragment_duration is known in advance and this (mehd) box may be
335 // omitted."
336 // TODO(wolenetz): Investigate gating liveness detection on timeline_offset
337 // when it's populated. See http://crbug.com/312699
338 params.liveness = DemuxerStream::LIVENESS_LIVE;
341 DVLOG(1) << "liveness: " << params.liveness;
343 if (!init_cb_.is_null())
344 base::ResetAndReturn(&init_cb_).Run(params);
346 if (!moov_->pssh.empty())
347 OnEncryptedMediaInitData(moov_->pssh);
349 return true;
352 bool MP4StreamParser::ParseMoof(BoxReader* reader) {
353 RCHECK(moov_.get()); // Must already have initialization segment
354 MovieFragment moof;
355 RCHECK(moof.Parse(reader));
356 if (!runs_)
357 runs_.reset(new TrackRunIterator(moov_.get(), media_log_));
358 RCHECK(runs_->Init(moof));
359 RCHECK(ComputeHighestEndOffset(moof));
361 if (!moof.pssh.empty())
362 OnEncryptedMediaInitData(moof.pssh);
364 new_segment_cb_.Run();
365 ChangeState(kWaitingForSampleData);
366 return true;
369 void MP4StreamParser::OnEncryptedMediaInitData(
370 const std::vector<ProtectionSystemSpecificHeader>& headers) {
371 // TODO(strobe): ensure that the value of init_data (all PSSH headers
372 // concatenated in arbitrary order) matches the EME spec.
373 // See https://www.w3.org/Bugs/Public/show_bug.cgi?id=17673.
374 size_t total_size = 0;
375 for (size_t i = 0; i < headers.size(); i++)
376 total_size += headers[i].raw_box.size();
378 std::vector<uint8> init_data(total_size);
379 size_t pos = 0;
380 for (size_t i = 0; i < headers.size(); i++) {
381 memcpy(&init_data[pos], &headers[i].raw_box[0],
382 headers[i].raw_box.size());
383 pos += headers[i].raw_box.size();
385 encrypted_media_init_data_cb_.Run(EmeInitDataType::CENC, init_data);
388 bool MP4StreamParser::PrepareAACBuffer(
389 const AAC& aac_config, std::vector<uint8>* frame_buf,
390 std::vector<SubsampleEntry>* subsamples) const {
391 // Append an ADTS header to every audio sample.
392 RCHECK(aac_config.ConvertEsdsToADTS(frame_buf));
394 // As above, adjust subsample information to account for the headers. AAC is
395 // not required to use subsample encryption, so we may need to add an entry.
396 if (subsamples->empty()) {
397 subsamples->push_back(SubsampleEntry(
398 kADTSHeaderMinSize, frame_buf->size() - kADTSHeaderMinSize));
399 } else {
400 (*subsamples)[0].clear_bytes += kADTSHeaderMinSize;
402 return true;
405 bool MP4StreamParser::EnqueueSample(BufferQueue* audio_buffers,
406 BufferQueue* video_buffers,
407 bool* err) {
408 DCHECK_EQ(state_, kEmittingSamples);
410 if (!runs_->IsRunValid()) {
411 // Flush any buffers we've gotten in this chunk so that buffers don't
412 // cross NewSegment() calls
413 *err = !SendAndFlushSamples(audio_buffers, video_buffers);
414 if (*err)
415 return false;
417 // Remain in kEmittingSamples state, discarding data, until the end of
418 // the current 'mdat' box has been appended to the queue.
419 if (!queue_.Trim(mdat_tail_))
420 return false;
422 ChangeState(kParsingBoxes);
423 end_of_segment_cb_.Run();
424 return true;
427 if (!runs_->IsSampleValid()) {
428 runs_->AdvanceRun();
429 return true;
432 DCHECK(!(*err));
434 const uint8* buf;
435 int buf_size;
436 queue_.Peek(&buf, &buf_size);
437 if (!buf_size) return false;
439 bool audio = has_audio_ && audio_track_id_ == runs_->track_id();
440 bool video = has_video_ && video_track_id_ == runs_->track_id();
442 // Skip this entire track if it's not one we're interested in
443 if (!audio && !video) {
444 runs_->AdvanceRun();
445 return true;
448 // Attempt to cache the auxiliary information first. Aux info is usually
449 // placed in a contiguous block before the sample data, rather than being
450 // interleaved. If we didn't cache it, this would require that we retain the
451 // start of the segment buffer while reading samples. Aux info is typically
452 // quite small compared to sample data, so this pattern is useful on
453 // memory-constrained devices where the source buffer consumes a substantial
454 // portion of the total system memory.
455 if (runs_->AuxInfoNeedsToBeCached()) {
456 queue_.PeekAt(runs_->aux_info_offset() + moof_head_, &buf, &buf_size);
457 if (buf_size < runs_->aux_info_size()) return false;
458 *err = !runs_->CacheAuxInfo(buf, buf_size);
459 return !*err;
462 queue_.PeekAt(runs_->sample_offset() + moof_head_, &buf, &buf_size);
463 if (buf_size < runs_->sample_size()) return false;
465 scoped_ptr<DecryptConfig> decrypt_config;
466 std::vector<SubsampleEntry> subsamples;
467 if (runs_->is_encrypted()) {
468 decrypt_config = runs_->GetDecryptConfig();
469 if (!decrypt_config) {
470 *err = true;
471 return false;
473 subsamples = decrypt_config->subsamples();
476 std::vector<uint8> frame_buf(buf, buf + runs_->sample_size());
477 if (video) {
478 DCHECK(runs_->video_description().frame_bitstream_converter);
479 if (!runs_->video_description().frame_bitstream_converter->ConvertFrame(
480 &frame_buf, runs_->is_keyframe(), &subsamples)) {
481 MEDIA_LOG(ERROR, media_log_)
482 << "Failed to prepare video sample for decode";
483 *err = true;
484 return false;
488 if (audio) {
489 if (ESDescriptor::IsAAC(runs_->audio_description().esds.object_type) &&
490 !PrepareAACBuffer(runs_->audio_description().esds.aac,
491 &frame_buf, &subsamples)) {
492 MEDIA_LOG(ERROR, media_log_) << "Failed to prepare AAC sample for decode";
493 *err = true;
494 return false;
498 if (decrypt_config) {
499 if (!subsamples.empty()) {
500 // Create a new config with the updated subsamples.
501 decrypt_config.reset(new DecryptConfig(
502 decrypt_config->key_id(),
503 decrypt_config->iv(),
504 subsamples));
506 // else, use the existing config.
507 } else if ((audio && is_audio_track_encrypted_) ||
508 (video && is_video_track_encrypted_)) {
509 // The media pipeline requires a DecryptConfig with an empty |iv|.
510 // TODO(ddorwin): Refactor so we do not need a fake key ID ("1");
511 decrypt_config.reset(
512 new DecryptConfig("1", "", std::vector<SubsampleEntry>()));
515 StreamParserBuffer::Type buffer_type = audio ? DemuxerStream::AUDIO :
516 DemuxerStream::VIDEO;
518 // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
519 // type and allow multiple tracks for same media type, if applicable. See
520 // https://crbug.com/341581.
521 scoped_refptr<StreamParserBuffer> stream_buf =
522 StreamParserBuffer::CopyFrom(&frame_buf[0], frame_buf.size(),
523 runs_->is_keyframe(),
524 buffer_type, 0);
526 if (decrypt_config)
527 stream_buf->set_decrypt_config(decrypt_config.Pass());
529 stream_buf->set_duration(runs_->duration());
530 stream_buf->set_timestamp(runs_->cts());
531 stream_buf->SetDecodeTimestamp(runs_->dts());
533 DVLOG(3) << "Pushing frame: aud=" << audio
534 << ", key=" << runs_->is_keyframe()
535 << ", dur=" << runs_->duration().InMilliseconds()
536 << ", dts=" << runs_->dts().InMilliseconds()
537 << ", cts=" << runs_->cts().InMilliseconds()
538 << ", size=" << runs_->sample_size();
540 if (audio) {
541 audio_buffers->push_back(stream_buf);
542 } else {
543 video_buffers->push_back(stream_buf);
546 runs_->AdvanceSample();
547 return true;
550 bool MP4StreamParser::SendAndFlushSamples(BufferQueue* audio_buffers,
551 BufferQueue* video_buffers) {
552 if (audio_buffers->empty() && video_buffers->empty())
553 return true;
555 TextBufferQueueMap empty_text_map;
556 bool success = new_buffers_cb_.Run(*audio_buffers,
557 *video_buffers,
558 empty_text_map);
559 audio_buffers->clear();
560 video_buffers->clear();
561 return success;
564 bool MP4StreamParser::ReadAndDiscardMDATsUntil(int64 max_clear_offset) {
565 bool err = false;
566 int64 upper_bound = std::min(max_clear_offset, queue_.tail());
567 while (mdat_tail_ < upper_bound) {
568 const uint8* buf = NULL;
569 int size = 0;
570 queue_.PeekAt(mdat_tail_, &buf, &size);
572 FourCC type;
573 int box_sz;
574 if (!BoxReader::StartTopLevelBox(buf, size, media_log_, &type, &box_sz,
575 &err))
576 break;
578 if (type != FOURCC_MDAT) {
579 MEDIA_LOG(DEBUG, media_log_)
580 << "Unexpected box type while parsing MDATs: "
581 << FourCCToString(type);
583 mdat_tail_ += box_sz;
585 queue_.Trim(std::min(mdat_tail_, upper_bound));
586 return !err;
589 void MP4StreamParser::ChangeState(State new_state) {
590 DVLOG(2) << "Changing state: " << new_state;
591 state_ = new_state;
594 bool MP4StreamParser::HaveEnoughDataToEnqueueSamples() {
595 DCHECK_EQ(state_, kWaitingForSampleData);
596 // For muxed content, make sure we have data up to |highest_end_offset_|
597 // so we can ensure proper enqueuing behavior. Otherwise assume we have enough
598 // data and allow per sample offset checks to meter sample enqueuing.
599 // TODO(acolwell): Fix trun box handling so we don't have to special case
600 // muxed content.
601 return !(has_audio_ && has_video_ &&
602 queue_.tail() < highest_end_offset_ + moof_head_);
605 bool MP4StreamParser::ComputeHighestEndOffset(const MovieFragment& moof) {
606 highest_end_offset_ = 0;
608 TrackRunIterator runs(moov_.get(), media_log_);
609 RCHECK(runs.Init(moof));
611 while (runs.IsRunValid()) {
612 int64 aux_info_end_offset = runs.aux_info_offset() + runs.aux_info_size();
613 if (aux_info_end_offset > highest_end_offset_)
614 highest_end_offset_ = aux_info_end_offset;
616 while (runs.IsSampleValid()) {
617 int64 sample_end_offset = runs.sample_offset() + runs.sample_size();
618 if (sample_end_offset > highest_end_offset_)
619 highest_end_offset_ = sample_end_offset;
621 runs.AdvanceSample();
623 runs.AdvanceRun();
626 return true;
629 } // namespace mp4
630 } // namespace media