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/mpeg/mpeg_audio_stream_parser_base.h"
8 #include "base/callback_helpers.h"
9 #include "base/message_loop/message_loop.h"
10 #include "media/base/buffers.h"
11 #include "media/base/stream_parser_buffer.h"
12 #include "media/base/text_track_config.h"
13 #include "media/base/video_decoder_config.h"
17 static const uint32 kICYStartCode = 0x49435920; // 'ICY '
19 // Arbitrary upper bound on the size of an IceCast header before it
21 static const int kMaxIcecastHeaderSize = 4096;
23 static const uint32 kID3StartCodeMask = 0xffffff00;
24 static const uint32 kID3v1StartCode = 0x54414700; // 'TAG\0'
25 static const int kID3v1Size = 128;
26 static const int kID3v1ExtendedSize = 227;
27 static const uint32 kID3v2StartCode = 0x49443300; // 'ID3\0'
29 static int LocateEndOfHeaders(const uint8_t* buf, int buf_len, int i) {
32 for (; i < buf_len; ++i) {
38 } else if (c != '\r' || last_c != '\n') {
46 MPEGAudioStreamParserBase::MPEGAudioStreamParserBase(uint32 start_code_mask,
47 AudioCodec audio_codec)
48 : state_(UNINITIALIZED),
49 in_media_segment_(false),
50 start_code_mask_(start_code_mask),
51 audio_codec_(audio_codec) {}
53 MPEGAudioStreamParserBase::~MPEGAudioStreamParserBase() {}
55 void MPEGAudioStreamParserBase::Init(const InitCB& init_cb,
56 const NewConfigCB& config_cb,
57 const NewBuffersCB& new_buffers_cb,
58 bool ignore_text_tracks,
59 const NeedKeyCB& need_key_cb,
60 const NewMediaSegmentCB& new_segment_cb,
61 const base::Closure& end_of_segment_cb,
62 const LogCB& log_cb) {
63 DVLOG(1) << __FUNCTION__;
64 DCHECK_EQ(state_, UNINITIALIZED);
66 config_cb_ = config_cb;
67 new_buffers_cb_ = new_buffers_cb;
68 new_segment_cb_ = new_segment_cb;
69 end_of_segment_cb_ = end_of_segment_cb;
72 ChangeState(INITIALIZED);
75 void MPEGAudioStreamParserBase::Flush() {
76 DVLOG(1) << __FUNCTION__;
77 DCHECK_NE(state_, UNINITIALIZED);
79 timestamp_helper_->SetBaseTimestamp(base::TimeDelta());
80 in_media_segment_ = false;
83 bool MPEGAudioStreamParserBase::Parse(const uint8* buf, int size) {
84 DVLOG(1) << __FUNCTION__ << "(" << size << ")";
87 DCHECK_NE(state_, UNINITIALIZED);
89 if (state_ == PARSE_ERROR)
92 DCHECK_EQ(state_, INITIALIZED);
94 queue_.Push(buf, size);
96 bool end_of_segment = true;
101 queue_.Peek(&data, &data_size);
106 uint32 start_code = data[0] << 24 | data[1] << 16 | data[2] << 8 | data[3];
108 bool parsed_metadata = true;
109 if ((start_code & start_code_mask_) == start_code_mask_) {
110 bytes_read = ParseFrame(data, data_size, &buffers);
112 // Only allow the current segment to end if a full frame has been parsed.
113 end_of_segment = bytes_read > 0;
114 parsed_metadata = false;
115 } else if (start_code == kICYStartCode) {
116 bytes_read = ParseIcecastHeader(data, data_size);
117 } else if ((start_code & kID3StartCodeMask) == kID3v1StartCode) {
118 bytes_read = ParseID3v1(data, data_size);
119 } else if ((start_code & kID3StartCodeMask) == kID3v2StartCode) {
120 bytes_read = ParseID3v2(data, data_size);
122 bytes_read = FindNextValidStartCode(data, data_size);
124 if (bytes_read > 0) {
125 DVLOG(1) << "Unexpected start code 0x" << std::hex << start_code;
126 DVLOG(1) << "SKIPPING " << bytes_read << " bytes of garbage.";
130 CHECK_LE(bytes_read, data_size);
132 if (bytes_read < 0) {
133 ChangeState(PARSE_ERROR);
135 } else if (bytes_read == 0) {
140 // Send pending buffers if we have encountered metadata.
141 if (parsed_metadata && !buffers.empty() && !SendBuffers(&buffers, true))
144 queue_.Pop(bytes_read);
145 end_of_segment = true;
151 // Send buffers collected in this append that haven't been sent yet.
152 return SendBuffers(&buffers, end_of_segment);
155 void MPEGAudioStreamParserBase::ChangeState(State state) {
156 DVLOG(1) << __FUNCTION__ << "() : " << state_ << " -> " << state;
160 int MPEGAudioStreamParserBase::ParseFrame(const uint8* data,
162 BufferQueue* buffers) {
163 DVLOG(2) << __FUNCTION__ << "(" << size << ")";
166 ChannelLayout channel_layout;
169 int bytes_read = ParseFrameHeader(
170 data, size, &frame_size, &sample_rate, &channel_layout, &sample_count);
175 // Make sure data contains the entire frame.
176 if (size < frame_size)
179 DVLOG(2) << " sample_rate " << sample_rate
180 << " channel_layout " << channel_layout
181 << " frame_size " << frame_size
182 << " sample_count " << sample_count;
184 if (config_.IsValidConfig() &&
185 (config_.samples_per_second() != sample_rate ||
186 config_.channel_layout() != channel_layout)) {
187 // Clear config data so that a config change is initiated.
188 config_ = AudioDecoderConfig();
190 // Send all buffers associated with the previous config.
191 if (!buffers->empty() && !SendBuffers(buffers, true))
195 if (!config_.IsValidConfig()) {
196 config_.Initialize(audio_codec_, kSampleFormatF32, channel_layout,
197 sample_rate, NULL, 0, false, false,
198 base::TimeDelta(), base::TimeDelta());
200 base::TimeDelta base_timestamp;
201 if (timestamp_helper_)
202 base_timestamp = timestamp_helper_->GetTimestamp();
204 timestamp_helper_.reset(new AudioTimestampHelper(sample_rate));
205 timestamp_helper_->SetBaseTimestamp(base_timestamp);
207 VideoDecoderConfig video_config;
208 bool success = config_cb_.Run(config_, video_config, TextTrackConfigMap());
210 if (!init_cb_.is_null())
211 base::ResetAndReturn(&init_cb_).Run(success, kInfiniteDuration());
217 // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
218 // type and allow multiple audio tracks, if applicable. See
219 // https://crbug.com/341581.
220 scoped_refptr<StreamParserBuffer> buffer =
221 StreamParserBuffer::CopyFrom(data, frame_size, true,
222 DemuxerStream::AUDIO, 0);
223 buffer->set_timestamp(timestamp_helper_->GetTimestamp());
224 buffer->set_duration(timestamp_helper_->GetFrameDuration(sample_count));
225 buffers->push_back(buffer);
227 timestamp_helper_->AddFrames(sample_count);
232 int MPEGAudioStreamParserBase::ParseIcecastHeader(const uint8* data, int size) {
233 DVLOG(1) << __FUNCTION__ << "(" << size << ")";
238 if (memcmp("ICY ", data, 4))
241 int locate_size = std::min(size, kMaxIcecastHeaderSize);
242 int offset = LocateEndOfHeaders(data, locate_size, 4);
244 if (locate_size == kMaxIcecastHeaderSize) {
245 MEDIA_LOG(log_cb_) << "Icecast header is too large.";
255 int MPEGAudioStreamParserBase::ParseID3v1(const uint8* data, int size) {
256 DVLOG(1) << __FUNCTION__ << "(" << size << ")";
258 if (size < kID3v1Size)
261 // TODO(acolwell): Add code to actually validate ID3v1 data and
262 // expose it as a metadata text track.
263 return !memcmp(data, "TAG+", 4) ? kID3v1ExtendedSize : kID3v1Size;
266 int MPEGAudioStreamParserBase::ParseID3v2(const uint8* data, int size) {
267 DVLOG(1) << __FUNCTION__ << "(" << size << ")";
272 BitReader reader(data, size);
278 if (!reader.ReadBits(24, &id) ||
279 !reader.ReadBits(16, &version) ||
280 !reader.ReadBits(8, &flags) ||
281 !ParseSyncSafeInt(&reader, &id3_size)) {
285 int32 actual_tag_size = 10 + id3_size;
287 // Increment size if 'Footer present' flag is set.
289 actual_tag_size += 10;
291 // Make sure we have the entire tag.
292 if (size < actual_tag_size)
295 // TODO(acolwell): Add code to actually validate ID3v2 data and
296 // expose it as a metadata text track.
297 return actual_tag_size;
300 bool MPEGAudioStreamParserBase::ParseSyncSafeInt(BitReader* reader,
303 for (int i = 0; i < 4; ++i) {
305 if (!reader->ReadBits(1, &tmp) || tmp != 0) {
306 MEDIA_LOG(log_cb_) << "ID3 syncsafe integer byte MSb is not 0!";
310 if (!reader->ReadBits(7, &tmp))
320 int MPEGAudioStreamParserBase::FindNextValidStartCode(const uint8* data,
322 const uint8* start = data;
323 const uint8* end = data + size;
325 while (start < end) {
326 int bytes_left = end - start;
327 const uint8* candidate_start_code =
328 static_cast<const uint8*>(memchr(start, 0xff, bytes_left));
330 if (!candidate_start_code)
333 bool parse_header_failed = false;
334 const uint8* sync = candidate_start_code;
335 // Try to find 3 valid frames in a row. 3 was selected to decrease
336 // the probability of false positives.
337 for (int i = 0; i < 3; ++i) {
338 int sync_size = end - sync;
340 int sync_bytes = ParseFrameHeader(
341 sync, sync_size, &frame_size, NULL, NULL, NULL);
346 if (sync_bytes > 0) {
347 DCHECK_LT(sync_bytes, sync_size);
349 // Skip over this frame so we can check the next one.
352 // Make sure the next frame starts inside the buffer.
356 DVLOG(1) << "ParseFrameHeader() " << i << " failed @" << (sync - data);
357 parse_header_failed = true;
362 if (parse_header_failed) {
363 // One of the frame header parses failed so |candidate_start_code|
364 // did not point to the start of a real frame. Move |start| forward
365 // so we can find the next candidate.
366 start = candidate_start_code + 1;
370 return candidate_start_code - data;
376 bool MPEGAudioStreamParserBase::SendBuffers(BufferQueue* buffers,
377 bool end_of_segment) {
378 DCHECK(!buffers->empty());
380 if (!in_media_segment_) {
381 in_media_segment_ = true;
382 new_segment_cb_.Run();
385 BufferQueue empty_video_buffers;
386 TextBufferQueueMap empty_text_map;
387 if (!new_buffers_cb_.Run(*buffers, empty_video_buffers, empty_text_map))
391 if (end_of_segment) {
392 in_media_segment_ = false;
393 end_of_segment_cb_.Run();