1 // Copyright 2013 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/mp3/mp3_stream_parser.h"
8 #include "base/callback_helpers.h"
9 #include "base/message_loop/message_loop.h"
10 #include "media/base/bit_reader.h"
11 #include "media/base/buffers.h"
12 #include "media/base/stream_parser_buffer.h"
13 #include "media/base/video_decoder_config.h"
14 #include "net/http/http_util.h"
18 static const uint32 kMP3StartCodeMask = 0xffe00000;
19 static const uint32 kICYStartCode = 0x49435920; // 'ICY '
21 // Arbitrary upper bound on the size of an IceCast header before it
23 static const int kMaxIcecastHeaderSize = 4096;
25 static const uint32 kID3StartCodeMask = 0xffffff00;
26 static const uint32 kID3v1StartCode = 0x54414700; // 'TAG\0'
27 static const int kID3v1Size = 128;
28 static const int kID3v1ExtendedSize = 227;
29 static const uint32 kID3v2StartCode = 0x49443300; // 'ID3\0'
31 // Map that determines which bitrate_index & channel_mode combinations
33 // Derived from: http://mpgedit.org/mpgedit/mpeg_format/MP3Format.html
34 static const bool kIsAllowed[17][4] = {
35 { true, true, true, true }, // free
36 { true, false, false, false }, // 32
37 { true, false, false, false }, // 48
38 { true, false, false, false }, // 56
39 { true, true, true, true }, // 64
40 { true, false, false, false }, // 80
41 { true, true, true, true }, // 96
42 { true, true, true, true }, // 112
43 { true, true, true, true }, // 128
44 { true, true, true, true }, // 160
45 { true, true, true, true }, // 192
46 { false, true, true, true }, // 224
47 { false, true, true, true }, // 256
48 { false, true, true, true }, // 320
49 { false, true, true, true }, // 384
50 { false, false, false, false } // bad
53 // Maps version and layer information in the frame header
54 // into an index for the |kBitrateMap|.
55 // Derived from: http://mpgedit.org/mpgedit/mpeg_format/MP3Format.html
56 static const int kVersionLayerMap[4][4] = {
57 // { reserved, L3, L2, L1 }
58 { 5, 4, 4, 3 }, // MPEG 2.5
59 { 5, 5, 5, 5 }, // reserved
60 { 5, 4, 4, 3 }, // MPEG 2
61 { 5, 2, 1, 0 } // MPEG 1
64 // Maps the bitrate index field in the header and an index
65 // from |kVersionLayerMap| to a frame bitrate.
66 // Derived from: http://mpgedit.org/mpgedit/mpeg_format/MP3Format.html
67 static const int kBitrateMap[16][6] = {
68 // { V1L1, V1L2, V1L3, V2L1, V2L2 & V2L3, reserved }
70 { 32, 32, 32, 32, 8, 0 },
71 { 64, 48, 40, 48, 16, 0 },
72 { 96, 56, 48, 56, 24, 0 },
73 { 128, 64, 56, 64, 32, 0 },
74 { 160, 80, 64, 80, 40, 0 },
75 { 192, 96, 80, 96, 48, 0 },
76 { 224, 112, 96, 112, 56, 0 },
77 { 256, 128, 112, 128, 64, 0 },
78 { 288, 160, 128, 144, 80, 0 },
79 { 320, 192, 160, 160, 96, 0 },
80 { 352, 224, 192, 176, 112, 0 },
81 { 384, 256, 224, 192, 128, 0 },
82 { 416, 320, 256, 224, 144, 0 },
83 { 448, 384, 320, 256, 160, 0 },
87 // Maps the sample rate index and version fields from the frame header
89 // Derived from: http://mpgedit.org/mpgedit/mpeg_format/MP3Format.html
90 static const int kSampleRateMap[4][4] = {
91 // { V2.5, reserved, V2, V1 }
92 { 11025, 0, 22050, 44100 },
93 { 12000, 0, 24000, 48000 },
94 { 8000, 0, 16000, 32000 },
98 // Frame header field constants.
99 static const int kVersion2 = 2;
100 static const int kVersionReserved = 1;
101 static const int kVersion2_5 = 0;
102 static const int kLayerReserved = 0;
103 static const int kLayer1 = 3;
104 static const int kLayer2 = 2;
105 static const int kLayer3 = 1;
106 static const int kBitrateFree = 0;
107 static const int kBitrateBad = 0xf;
108 static const int kSampleRateReserved = 3;
110 MP3StreamParser::MP3StreamParser()
111 : state_(UNINITIALIZED),
112 in_media_segment_(false) {
115 MP3StreamParser::~MP3StreamParser() {}
117 void MP3StreamParser::Init(const InitCB& init_cb,
118 const NewConfigCB& config_cb,
119 const NewBuffersCB& new_buffers_cb,
120 const NewTextBuffersCB& text_cb,
121 const NeedKeyCB& need_key_cb,
122 const AddTextTrackCB& add_text_track_cb,
123 const NewMediaSegmentCB& new_segment_cb,
124 const base::Closure& end_of_segment_cb,
125 const LogCB& log_cb) {
126 DVLOG(1) << __FUNCTION__;
127 DCHECK_EQ(state_, UNINITIALIZED);
129 config_cb_ = config_cb;
130 new_buffers_cb_ = new_buffers_cb;
131 new_segment_cb_ = new_segment_cb;
132 end_of_segment_cb_ = end_of_segment_cb;
135 ChangeState(INITIALIZED);
138 void MP3StreamParser::Flush() {
139 DVLOG(1) << __FUNCTION__;
140 DCHECK_NE(state_, UNINITIALIZED);
142 timestamp_helper_->SetBaseTimestamp(base::TimeDelta());
143 in_media_segment_ = false;
146 bool MP3StreamParser::Parse(const uint8* buf, int size) {
147 DVLOG(1) << __FUNCTION__ << "(" << size << ")";
150 DCHECK_NE(state_, UNINITIALIZED);
152 if (state_ == PARSE_ERROR)
155 DCHECK_EQ(state_, INITIALIZED);
157 queue_.Push(buf, size);
159 bool end_of_segment = true;
164 queue_.Peek(&data, &data_size);
169 uint32 start_code = data[0] << 24 | data[1] << 16 | data[2] << 8 | data[3];
171 bool parsed_metadata = true;
172 if ((start_code & kMP3StartCodeMask) == kMP3StartCodeMask) {
173 bytes_read = ParseMP3Frame(data, data_size, &buffers);
175 // Only allow the current segment to end if a full frame has been parsed.
176 end_of_segment = bytes_read > 0;
177 parsed_metadata = false;
178 } else if (start_code == kICYStartCode) {
179 bytes_read = ParseIcecastHeader(data, data_size);
180 } else if ((start_code & kID3StartCodeMask) == kID3v1StartCode) {
181 bytes_read = ParseID3v1(data, data_size);
182 } else if ((start_code & kID3StartCodeMask) == kID3v2StartCode) {
183 bytes_read = ParseID3v2(data, data_size);
185 bytes_read = FindNextValidStartCode(data, data_size);
187 if (bytes_read > 0) {
188 DVLOG(1) << "Unexpected start code 0x" << std::hex << start_code;
189 DVLOG(1) << "SKIPPING " << bytes_read << " bytes of garbage.";
193 CHECK_LE(bytes_read, data_size);
195 if (bytes_read < 0) {
196 ChangeState(PARSE_ERROR);
198 } else if (bytes_read == 0) {
203 // Send pending buffers if we have encountered metadata.
204 if (parsed_metadata && !buffers.empty() && !SendBuffers(&buffers, true))
207 queue_.Pop(bytes_read);
208 end_of_segment = true;
214 // Send buffers collected in this append that haven't been sent yet.
215 return SendBuffers(&buffers, end_of_segment);
218 void MP3StreamParser::ChangeState(State state) {
219 DVLOG(1) << __FUNCTION__ << "() : " << state_ << " -> " << state;
223 int MP3StreamParser::ParseFrameHeader(const uint8* data, int size,
226 ChannelLayout* channel_layout,
227 int* sample_count) const {
235 BitReader reader(data, size);
241 int sample_rate_index;
247 if (!reader.ReadBits(11, &sync) ||
248 !reader.ReadBits(2, &version) ||
249 !reader.ReadBits(2, &layer) ||
250 !reader.ReadBits(1, &is_protected) ||
251 !reader.ReadBits(4, &bitrate_index) ||
252 !reader.ReadBits(2, &sample_rate_index) ||
253 !reader.ReadBits(1, &has_padding) ||
254 !reader.ReadBits(1, &is_private) ||
255 !reader.ReadBits(2, &channel_mode) ||
256 !reader.ReadBits(6, &other_flags)) {
260 DVLOG(2) << "Header data :" << std::hex
261 << " sync 0x" << sync
262 << " version 0x" << version
263 << " layer 0x" << layer
264 << " bitrate_index 0x" << bitrate_index
265 << " sample_rate_index 0x" << sample_rate_index
266 << " channel_mode 0x" << channel_mode;
269 version == kVersionReserved ||
270 layer == kLayerReserved ||
271 bitrate_index == kBitrateFree || bitrate_index == kBitrateBad ||
272 sample_rate_index == kSampleRateReserved) {
273 MEDIA_LOG(log_cb_) << "Invalid header data :" << std::hex
274 << " sync 0x" << sync
275 << " version 0x" << version
276 << " layer 0x" << layer
277 << " bitrate_index 0x" << bitrate_index
278 << " sample_rate_index 0x" << sample_rate_index
279 << " channel_mode 0x" << channel_mode;
283 if (layer == kLayer2 && kIsAllowed[bitrate_index][channel_mode]) {
284 MEDIA_LOG(log_cb_) << "Invalid (bitrate_index, channel_mode) combination :"
286 << " bitrate_index " << bitrate_index
287 << " channel_mode " << channel_mode;
291 int bitrate = kBitrateMap[bitrate_index][kVersionLayerMap[version][layer]];
294 MEDIA_LOG(log_cb_) << "Invalid bitrate :" << std::hex
295 << " version " << version
296 << " layer " << layer
297 << " bitrate_index " << bitrate_index;
301 DVLOG(2) << " bitrate " << bitrate;
303 int frame_sample_rate = kSampleRateMap[sample_rate_index][version];
304 if (frame_sample_rate == 0) {
305 MEDIA_LOG(log_cb_) << "Invalid sample rate :" << std::hex
306 << " version " << version
307 << " sample_rate_index " << sample_rate_index;
312 *sample_rate = frame_sample_rate;
314 // http://teslabs.com/openplayer/docs/docs/specs/mp3_structure2.pdf
316 int samples_per_frame;
319 samples_per_frame = 384;
323 samples_per_frame = 1152;
327 if (version == kVersion2 || version == kVersion2_5)
328 samples_per_frame = 576;
330 samples_per_frame = 1152;
338 *sample_count = samples_per_frame;
340 // http://teslabs.com/openplayer/docs/docs/specs/mp3_structure2.pdf
341 // Text just below Table 2.1.5.
342 if (layer == kLayer1) {
343 // This formulation is a slight variation on the equation below,
344 // but has slightly different truncation characteristics to deal
345 // with the fact that Layer 1 has 4 byte "slots" instead of single
347 *frame_size = 4 * (12 * bitrate * 1000 / frame_sample_rate);
350 ((samples_per_frame / 8) * bitrate * 1000) / frame_sample_rate;
354 *frame_size += (layer == kLayer1) ? 4 : 1;
356 if (channel_layout) {
357 // Map Stereo(0), Joint Stereo(1), and Dual Channel (2) to
358 // CHANNEL_LAYOUT_STEREO and Single Channel (3) to CHANNEL_LAYOUT_MONO.
360 (channel_mode == 3) ? CHANNEL_LAYOUT_MONO : CHANNEL_LAYOUT_STEREO;
366 int MP3StreamParser::ParseMP3Frame(const uint8* data,
368 BufferQueue* buffers) {
369 DVLOG(2) << __FUNCTION__ << "(" << size << ")";
372 ChannelLayout channel_layout;
375 int bytes_read = ParseFrameHeader(
376 data, size, &frame_size, &sample_rate, &channel_layout, &sample_count);
381 // Make sure data contains the entire frame.
382 if (size < frame_size)
385 DVLOG(2) << " sample_rate " << sample_rate
386 << " channel_layout " << channel_layout
387 << " frame_size " << frame_size;
389 if (config_.IsValidConfig() &&
390 (config_.samples_per_second() != sample_rate ||
391 config_.channel_layout() != channel_layout)) {
392 // Clear config data so that a config change is initiated.
393 config_ = AudioDecoderConfig();
395 // Send all buffers associated with the previous config.
396 if (!buffers->empty() && !SendBuffers(buffers, true))
400 if (!config_.IsValidConfig()) {
401 config_.Initialize(kCodecMP3, kSampleFormatF32, channel_layout,
402 sample_rate, NULL, 0, false, false,
403 base::TimeDelta(), base::TimeDelta());
405 base::TimeDelta base_timestamp;
406 if (timestamp_helper_)
407 base_timestamp = timestamp_helper_->GetTimestamp();
409 timestamp_helper_.reset(new AudioTimestampHelper(sample_rate));
410 timestamp_helper_->SetBaseTimestamp(base_timestamp);
412 VideoDecoderConfig video_config;
413 bool success = config_cb_.Run(config_, video_config);
415 if (!init_cb_.is_null())
416 base::ResetAndReturn(&init_cb_).Run(success, kInfiniteDuration());
422 scoped_refptr<StreamParserBuffer> buffer =
423 StreamParserBuffer::CopyFrom(data, frame_size, true);
424 buffer->set_timestamp(timestamp_helper_->GetTimestamp());
425 buffer->set_duration(timestamp_helper_->GetFrameDuration(sample_count));
426 buffers->push_back(buffer);
428 timestamp_helper_->AddFrames(sample_count);
433 int MP3StreamParser::ParseIcecastHeader(const uint8* data, int size) {
434 DVLOG(1) << __FUNCTION__ << "(" << size << ")";
439 if (memcmp("ICY ", data, 4))
442 int locate_size = std::min(size, kMaxIcecastHeaderSize);
443 int offset = net::HttpUtil::LocateEndOfHeaders(
444 reinterpret_cast<const char*>(data), locate_size, 4);
446 if (locate_size == kMaxIcecastHeaderSize) {
447 MEDIA_LOG(log_cb_) << "Icecast header is too large.";
457 int MP3StreamParser::ParseID3v1(const uint8* data, int size) {
458 DVLOG(1) << __FUNCTION__ << "(" << size << ")";
460 if (size < kID3v1Size)
463 // TODO(acolwell): Add code to actually validate ID3v1 data and
464 // expose it as a metadata text track.
465 return !memcmp(data, "TAG+", 4) ? kID3v1ExtendedSize : kID3v1Size;
468 int MP3StreamParser::ParseID3v2(const uint8* data, int size) {
469 DVLOG(1) << __FUNCTION__ << "(" << size << ")";
474 BitReader reader(data, size);
480 if (!reader.ReadBits(24, &id) ||
481 !reader.ReadBits(16, &version) ||
482 !reader.ReadBits(8, &flags) ||
483 !ParseSyncSafeInt(&reader, &id3_size)) {
487 int32 actual_tag_size = 10 + id3_size;
489 // Increment size if 'Footer present' flag is set.
491 actual_tag_size += 10;
493 // Make sure we have the entire tag.
494 if (size < actual_tag_size)
497 // TODO(acolwell): Add code to actually validate ID3v2 data and
498 // expose it as a metadata text track.
499 return actual_tag_size;
502 bool MP3StreamParser::ParseSyncSafeInt(BitReader* reader, int32* value) {
504 for (int i = 0; i < 4; ++i) {
506 if (!reader->ReadBits(1, &tmp) || tmp != 0) {
507 MEDIA_LOG(log_cb_) << "ID3 syncsafe integer byte MSb is not 0!";
511 if (!reader->ReadBits(7, &tmp))
521 int MP3StreamParser::FindNextValidStartCode(const uint8* data, int size) const {
522 const uint8* start = data;
523 const uint8* end = data + size;
525 while (start < end) {
526 int bytes_left = end - start;
527 const uint8* candidate_start_code =
528 static_cast<const uint8*>(memchr(start, 0xff, bytes_left));
530 if (!candidate_start_code)
533 bool parse_header_failed = false;
534 const uint8* sync = candidate_start_code;
535 // Try to find 3 valid frames in a row. 3 was selected to decrease
536 // the probability of false positives.
537 for (int i = 0; i < 3; ++i) {
538 int sync_size = end - sync;
540 int sync_bytes = ParseFrameHeader(
541 sync, sync_size, &frame_size, NULL, NULL, NULL);
546 if (sync_bytes > 0) {
547 DCHECK_LT(sync_bytes, sync_size);
549 // Skip over this frame so we can check the next one.
552 // Make sure the next frame starts inside the buffer.
556 DVLOG(1) << "ParseFrameHeader() " << i << " failed @" << (sync - data);
557 parse_header_failed = true;
562 if (parse_header_failed) {
563 // One of the frame header parses failed so |candidate_start_code|
564 // did not point to the start of a real frame. Move |start| forward
565 // so we can find the next candidate.
566 start = candidate_start_code + 1;
570 return candidate_start_code - data;
576 bool MP3StreamParser::SendBuffers(BufferQueue* buffers, bool end_of_segment) {
577 DCHECK(!buffers->empty());
579 if (!in_media_segment_) {
580 in_media_segment_ = true;
581 new_segment_cb_.Run();
584 BufferQueue empty_video_buffers;
585 if (!new_buffers_cb_.Run(*buffers, empty_video_buffers))
589 if (end_of_segment) {
590 in_media_segment_ = false;
591 end_of_segment_cb_.Run();