1 /*
2 * Copyright (C) 2009 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 "MP3Extractor"
19 #include <utils/Log.h>
20
21 #include "include/MP3Extractor.h"
22
23 #include "include/avc_utils.h"
24 #include "include/ID3.h"
25 #include "include/VBRISeeker.h"
26 #include "include/XINGSeeker.h"
27
28 #include <media/stagefright/foundation/ADebug.h>
29 #include <media/stagefright/foundation/AMessage.h>
30 #include <media/stagefright/DataSource.h>
31 #include <media/stagefright/MediaBuffer.h>
32 #include <media/stagefright/MediaBufferGroup.h>
33 #include <media/stagefright/MediaDefs.h>
34 #include <media/stagefright/MediaErrors.h>
35 #include <media/stagefright/MediaSource.h>
36 #include <media/stagefright/MetaData.h>
37 #include <media/stagefright/Utils.h>
38 #include <utils/String8.h>
39
40 namespace android {
41
42 // Everything must match except for
43 // protection, bitrate, padding, private bits, mode, mode extension,
44 // copyright bit, original bit and emphasis.
45 // Yes ... there are things that must indeed match...
46 static const uint32_t kMask = 0xfffe0c00;
47
Resync(const sp<DataSource> & source,uint32_t match_header,off64_t * inout_pos,off64_t * post_id3_pos,uint32_t * out_header)48 static bool Resync(
49 const sp<DataSource> &source, uint32_t match_header,
50 off64_t *inout_pos, off64_t *post_id3_pos, uint32_t *out_header) {
51 if (post_id3_pos != NULL) {
52 *post_id3_pos = 0;
53 }
54
55 if (*inout_pos == 0) {
56 // Skip an optional ID3 header if syncing at the very beginning
57 // of the datasource.
58
59 for (;;) {
60 uint8_t id3header[10];
61 if (source->readAt(*inout_pos, id3header, sizeof(id3header))
62 < (ssize_t)sizeof(id3header)) {
63 // If we can't even read these 10 bytes, we might as well bail
64 // out, even if there _were_ 10 bytes of valid mp3 audio data...
65 return false;
66 }
67
68 if (memcmp("ID3", id3header, 3)) {
69 break;
70 }
71
72 // Skip the ID3v2 header.
73
74 size_t len =
75 ((id3header[6] & 0x7f) << 21)
76 | ((id3header[7] & 0x7f) << 14)
77 | ((id3header[8] & 0x7f) << 7)
78 | (id3header[9] & 0x7f);
79
80 len += 10;
81
82 *inout_pos += len;
83
84 ALOGV("skipped ID3 tag, new starting offset is %lld (0x%016llx)",
85 *inout_pos, *inout_pos);
86 }
87
88 if (post_id3_pos != NULL) {
89 *post_id3_pos = *inout_pos;
90 }
91 }
92
93 off64_t pos = *inout_pos;
94 bool valid = false;
95
96 const size_t kMaxReadBytes = 1024;
97 const size_t kMaxBytesChecked = 128 * 1024;
98 uint8_t buf[kMaxReadBytes];
99 ssize_t bytesToRead = kMaxReadBytes;
100 ssize_t totalBytesRead = 0;
101 ssize_t remainingBytes = 0;
102 bool reachEOS = false;
103 uint8_t *tmp = buf;
104
105 do {
106 if (pos >= *inout_pos + kMaxBytesChecked) {
107 // Don't scan forever.
108 ALOGV("giving up at offset %lld", pos);
109 break;
110 }
111
112 if (remainingBytes < 4) {
113 if (reachEOS) {
114 break;
115 } else {
116 memcpy(buf, tmp, remainingBytes);
117 bytesToRead = kMaxReadBytes - remainingBytes;
118
119 /*
120 * The next read position should start from the end of
121 * the last buffer, and thus should include the remaining
122 * bytes in the buffer.
123 */
124 totalBytesRead = source->readAt(pos + remainingBytes,
125 buf + remainingBytes,
126 bytesToRead);
127 if (totalBytesRead <= 0) {
128 break;
129 }
130 reachEOS = (totalBytesRead != bytesToRead);
131 totalBytesRead += remainingBytes;
132 remainingBytes = totalBytesRead;
133 tmp = buf;
134 continue;
135 }
136 }
137
138 uint32_t header = U32_AT(tmp);
139
140 if (match_header != 0 && (header & kMask) != (match_header & kMask)) {
141 ++pos;
142 ++tmp;
143 --remainingBytes;
144 continue;
145 }
146
147 size_t frame_size;
148 int sample_rate, num_channels, bitrate;
149 if (!GetMPEGAudioFrameSize(
150 header, &frame_size,
151 &sample_rate, &num_channels, &bitrate)) {
152 ++pos;
153 ++tmp;
154 --remainingBytes;
155 continue;
156 }
157
158 ALOGV("found possible 1st frame at %lld (header = 0x%08x)", pos, header);
159
160 // We found what looks like a valid frame,
161 // now find its successors.
162
163 off64_t test_pos = pos + frame_size;
164
165 valid = true;
166 for (int j = 0; j < 3; ++j) {
167 uint8_t tmp[4];
168 if (source->readAt(test_pos, tmp, 4) < 4) {
169 valid = false;
170 break;
171 }
172
173 uint32_t test_header = U32_AT(tmp);
174
175 ALOGV("subsequent header is %08x", test_header);
176
177 if ((test_header & kMask) != (header & kMask)) {
178 valid = false;
179 break;
180 }
181
182 size_t test_frame_size;
183 if (!GetMPEGAudioFrameSize(
184 test_header, &test_frame_size)) {
185 valid = false;
186 break;
187 }
188
189 ALOGV("found subsequent frame #%d at %lld", j + 2, test_pos);
190
191 test_pos += test_frame_size;
192 }
193
194 if (valid) {
195 *inout_pos = pos;
196
197 if (out_header != NULL) {
198 *out_header = header;
199 }
200 } else {
201 ALOGV("no dice, no valid sequence of frames found.");
202 }
203
204 ++pos;
205 ++tmp;
206 --remainingBytes;
207 } while (!valid);
208
209 return valid;
210 }
211
212 class MP3Source : public MediaSource {
213 public:
214 MP3Source(
215 const sp<MetaData> &meta, const sp<DataSource> &source,
216 off64_t first_frame_pos, uint32_t fixed_header,
217 const sp<MP3Seeker> &seeker);
218
219 virtual status_t start(MetaData *params = NULL);
220 virtual status_t stop();
221
222 virtual sp<MetaData> getFormat();
223
224 virtual status_t read(
225 MediaBuffer **buffer, const ReadOptions *options = NULL);
226
227 protected:
228 virtual ~MP3Source();
229
230 private:
231 static const size_t kMaxFrameSize;
232 sp<MetaData> mMeta;
233 sp<DataSource> mDataSource;
234 off64_t mFirstFramePos;
235 uint32_t mFixedHeader;
236 off64_t mCurrentPos;
237 int64_t mCurrentTimeUs;
238 bool mStarted;
239 sp<MP3Seeker> mSeeker;
240 MediaBufferGroup *mGroup;
241
242 int64_t mBasisTimeUs;
243 int64_t mSamplesRead;
244
245 MP3Source(const MP3Source &);
246 MP3Source &operator=(const MP3Source &);
247 };
248
MP3Extractor(const sp<DataSource> & source,const sp<AMessage> & meta)249 MP3Extractor::MP3Extractor(
250 const sp<DataSource> &source, const sp<AMessage> &meta)
251 : mInitCheck(NO_INIT),
252 mDataSource(source),
253 mFirstFramePos(-1),
254 mFixedHeader(0) {
255 off64_t pos = 0;
256 off64_t post_id3_pos;
257 uint32_t header;
258 bool success;
259
260 int64_t meta_offset;
261 uint32_t meta_header;
262 int64_t meta_post_id3_offset;
263 if (meta != NULL
264 && meta->findInt64("offset", &meta_offset)
265 && meta->findInt32("header", (int32_t *)&meta_header)
266 && meta->findInt64("post-id3-offset", &meta_post_id3_offset)) {
267 // The sniffer has already done all the hard work for us, simply
268 // accept its judgement.
269 pos = (off64_t)meta_offset;
270 header = meta_header;
271 post_id3_pos = (off64_t)meta_post_id3_offset;
272
273 success = true;
274 } else {
275 success = Resync(mDataSource, 0, &pos, &post_id3_pos, &header);
276 }
277
278 if (!success) {
279 // mInitCheck will remain NO_INIT
280 return;
281 }
282
283 mFirstFramePos = pos;
284 mFixedHeader = header;
285
286 size_t frame_size;
287 int sample_rate;
288 int num_channels;
289 int bitrate;
290 GetMPEGAudioFrameSize(
291 header, &frame_size, &sample_rate, &num_channels, &bitrate);
292
293 unsigned layer = 4 - ((header >> 17) & 3);
294
295 mMeta = new MetaData;
296
297 switch (layer) {
298 case 1:
299 mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_I);
300 break;
301 case 2:
302 mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_II);
303 break;
304 case 3:
305 mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG);
306 break;
307 default:
308 TRESPASS();
309 }
310
311 mMeta->setInt32(kKeySampleRate, sample_rate);
312 mMeta->setInt32(kKeyBitRate, bitrate * 1000);
313 mMeta->setInt32(kKeyChannelCount, num_channels);
314
315 sp<XINGSeeker> seeker = XINGSeeker::CreateFromSource(mDataSource, mFirstFramePos);
316
317 if (seeker == NULL) {
318 mSeeker = VBRISeeker::CreateFromSource(mDataSource, post_id3_pos);
319 } else {
320 mSeeker = seeker;
321 int encd = seeker->getEncoderDelay();
322 int encp = seeker->getEncoderPadding();
323 if (encd != 0 || encp != 0) {
324 mMeta->setInt32(kKeyEncoderDelay, encd);
325 mMeta->setInt32(kKeyEncoderPadding, encp);
326 }
327 }
328
329 if (mSeeker != NULL) {
330 // While it is safe to send the XING/VBRI frame to the decoder, this will
331 // result in an extra 1152 samples being output. The real first frame to
332 // decode is after the XING/VBRI frame, so skip there.
333 mFirstFramePos += frame_size;
334 }
335
336 int64_t durationUs;
337
338 if (mSeeker == NULL || !mSeeker->getDuration(&durationUs)) {
339 off64_t fileSize;
340 if (mDataSource->getSize(&fileSize) == OK) {
341 durationUs = 8000LL * (fileSize - mFirstFramePos) / bitrate;
342 } else {
343 durationUs = -1;
344 }
345 }
346
347 if (durationUs >= 0) {
348 mMeta->setInt64(kKeyDuration, durationUs);
349 }
350
351 mInitCheck = OK;
352
353 // Get iTunes-style gapless info if present.
354 // When getting the id3 tag, skip the V1 tags to prevent the source cache
355 // from being iterated to the end of the file.
356 ID3 id3(mDataSource, true);
357 if (id3.isValid()) {
358 ID3::Iterator *com = new ID3::Iterator(id3, "COM");
359 if (com->done()) {
360 delete com;
361 com = new ID3::Iterator(id3, "COMM");
362 }
363 while(!com->done()) {
364 String8 commentdesc;
365 String8 commentvalue;
366 com->getString(&commentdesc, &commentvalue);
367 const char * desc = commentdesc.string();
368 const char * value = commentvalue.string();
369
370 // first 3 characters are the language, which we don't care about
371 if(strlen(desc) > 3 && strcmp(desc + 3, "iTunSMPB") == 0) {
372
373 int32_t delay, padding;
374 if (sscanf(value, " %*x %x %x %*x", &delay, &padding) == 2) {
375 mMeta->setInt32(kKeyEncoderDelay, delay);
376 mMeta->setInt32(kKeyEncoderPadding, padding);
377 }
378 break;
379 }
380 com->next();
381 }
382 delete com;
383 com = NULL;
384 }
385 }
386
countTracks()387 size_t MP3Extractor::countTracks() {
388 return mInitCheck != OK ? 0 : 1;
389 }
390
getTrack(size_t index)391 sp<MediaSource> MP3Extractor::getTrack(size_t index) {
392 if (mInitCheck != OK || index != 0) {
393 return NULL;
394 }
395
396 return new MP3Source(
397 mMeta, mDataSource, mFirstFramePos, mFixedHeader,
398 mSeeker);
399 }
400
getTrackMetaData(size_t index,uint32_t flags)401 sp<MetaData> MP3Extractor::getTrackMetaData(size_t index, uint32_t flags) {
402 if (mInitCheck != OK || index != 0) {
403 return NULL;
404 }
405
406 return mMeta;
407 }
408
409 ////////////////////////////////////////////////////////////////////////////////
410
411 // The theoretical maximum frame size for an MPEG audio stream should occur
412 // while playing a Layer 2, MPEGv2.5 audio stream at 160kbps (with padding).
413 // The size of this frame should be...
414 // ((1152 samples/frame * 160000 bits/sec) /
415 // (8000 samples/sec * 8 bits/byte)) + 1 padding byte/frame = 2881 bytes/frame.
416 // Set our max frame size to the nearest power of 2 above this size (aka, 4kB)
417 const size_t MP3Source::kMaxFrameSize = (1 << 12); /* 4096 bytes */
MP3Source(const sp<MetaData> & meta,const sp<DataSource> & source,off64_t first_frame_pos,uint32_t fixed_header,const sp<MP3Seeker> & seeker)418 MP3Source::MP3Source(
419 const sp<MetaData> &meta, const sp<DataSource> &source,
420 off64_t first_frame_pos, uint32_t fixed_header,
421 const sp<MP3Seeker> &seeker)
422 : mMeta(meta),
423 mDataSource(source),
424 mFirstFramePos(first_frame_pos),
425 mFixedHeader(fixed_header),
426 mCurrentPos(0),
427 mCurrentTimeUs(0),
428 mStarted(false),
429 mSeeker(seeker),
430 mGroup(NULL),
431 mBasisTimeUs(0),
432 mSamplesRead(0) {
433 }
434
~MP3Source()435 MP3Source::~MP3Source() {
436 if (mStarted) {
437 stop();
438 }
439 }
440
start(MetaData *)441 status_t MP3Source::start(MetaData *) {
442 CHECK(!mStarted);
443
444 mGroup = new MediaBufferGroup;
445
446 mGroup->add_buffer(new MediaBuffer(kMaxFrameSize));
447
448 mCurrentPos = mFirstFramePos;
449 mCurrentTimeUs = 0;
450
451 mBasisTimeUs = mCurrentTimeUs;
452 mSamplesRead = 0;
453
454 mStarted = true;
455
456 return OK;
457 }
458
stop()459 status_t MP3Source::stop() {
460 CHECK(mStarted);
461
462 delete mGroup;
463 mGroup = NULL;
464
465 mStarted = false;
466
467 return OK;
468 }
469
getFormat()470 sp<MetaData> MP3Source::getFormat() {
471 return mMeta;
472 }
473
read(MediaBuffer ** out,const ReadOptions * options)474 status_t MP3Source::read(
475 MediaBuffer **out, const ReadOptions *options) {
476 *out = NULL;
477
478 int64_t seekTimeUs;
479 ReadOptions::SeekMode mode;
480 bool seekCBR = false;
481
482 if (options != NULL && options->getSeekTo(&seekTimeUs, &mode)) {
483 int64_t actualSeekTimeUs = seekTimeUs;
484 if (mSeeker == NULL
485 || !mSeeker->getOffsetForTime(&actualSeekTimeUs, &mCurrentPos)) {
486 int32_t bitrate;
487 if (!mMeta->findInt32(kKeyBitRate, &bitrate)) {
488 // bitrate is in bits/sec.
489 ALOGI("no bitrate");
490
491 return ERROR_UNSUPPORTED;
492 }
493
494 mCurrentTimeUs = seekTimeUs;
495 mCurrentPos = mFirstFramePos + seekTimeUs * bitrate / 8000000;
496 seekCBR = true;
497 } else {
498 mCurrentTimeUs = actualSeekTimeUs;
499 }
500
501 mBasisTimeUs = mCurrentTimeUs;
502 mSamplesRead = 0;
503 }
504
505 MediaBuffer *buffer;
506 status_t err = mGroup->acquire_buffer(&buffer);
507 if (err != OK) {
508 return err;
509 }
510
511 size_t frame_size;
512 int bitrate;
513 int num_samples;
514 int sample_rate;
515 for (;;) {
516 ssize_t n = mDataSource->readAt(mCurrentPos, buffer->data(), 4);
517 if (n < 4) {
518 buffer->release();
519 buffer = NULL;
520
521 return ERROR_END_OF_STREAM;
522 }
523
524 uint32_t header = U32_AT((const uint8_t *)buffer->data());
525
526 if ((header & kMask) == (mFixedHeader & kMask)
527 && GetMPEGAudioFrameSize(
528 header, &frame_size, &sample_rate, NULL,
529 &bitrate, &num_samples)) {
530
531 // re-calculate mCurrentTimeUs because we might have called Resync()
532 if (seekCBR) {
533 mCurrentTimeUs = (mCurrentPos - mFirstFramePos) * 8000 / bitrate;
534 mBasisTimeUs = mCurrentTimeUs;
535 }
536
537 break;
538 }
539
540 // Lost sync.
541 ALOGV("lost sync! header = 0x%08x, old header = 0x%08x\n", header, mFixedHeader);
542
543 off64_t pos = mCurrentPos;
544 if (!Resync(mDataSource, mFixedHeader, &pos, NULL, NULL)) {
545 ALOGE("Unable to resync. Signalling end of stream.");
546
547 buffer->release();
548 buffer = NULL;
549
550 return ERROR_END_OF_STREAM;
551 }
552
553 mCurrentPos = pos;
554
555 // Try again with the new position.
556 }
557
558 CHECK(frame_size <= buffer->size());
559
560 ssize_t n = mDataSource->readAt(mCurrentPos, buffer->data(), frame_size);
561 if (n < (ssize_t)frame_size) {
562 buffer->release();
563 buffer = NULL;
564
565 return ERROR_END_OF_STREAM;
566 }
567
568 buffer->set_range(0, frame_size);
569
570 buffer->meta_data()->setInt64(kKeyTime, mCurrentTimeUs);
571 buffer->meta_data()->setInt32(kKeyIsSyncFrame, 1);
572
573 mCurrentPos += frame_size;
574
575 mSamplesRead += num_samples;
576 mCurrentTimeUs = mBasisTimeUs + ((mSamplesRead * 1000000) / sample_rate);
577
578 *out = buffer;
579
580 return OK;
581 }
582
getMetaData()583 sp<MetaData> MP3Extractor::getMetaData() {
584 sp<MetaData> meta = new MetaData;
585
586 if (mInitCheck != OK) {
587 return meta;
588 }
589
590 meta->setCString(kKeyMIMEType, "audio/mpeg");
591
592 ID3 id3(mDataSource);
593
594 if (!id3.isValid()) {
595 return meta;
596 }
597
598 struct Map {
599 int key;
600 const char *tag1;
601 const char *tag2;
602 };
603 static const Map kMap[] = {
604 { kKeyAlbum, "TALB", "TAL" },
605 { kKeyArtist, "TPE1", "TP1" },
606 { kKeyAlbumArtist, "TPE2", "TP2" },
607 { kKeyComposer, "TCOM", "TCM" },
608 { kKeyGenre, "TCON", "TCO" },
609 { kKeyTitle, "TIT2", "TT2" },
610 { kKeyYear, "TYE", "TYER" },
611 { kKeyAuthor, "TXT", "TEXT" },
612 { kKeyCDTrackNumber, "TRK", "TRCK" },
613 { kKeyDiscNumber, "TPA", "TPOS" },
614 { kKeyCompilation, "TCP", "TCMP" },
615 };
616 static const size_t kNumMapEntries = sizeof(kMap) / sizeof(kMap[0]);
617
618 for (size_t i = 0; i < kNumMapEntries; ++i) {
619 ID3::Iterator *it = new ID3::Iterator(id3, kMap[i].tag1);
620 if (it->done()) {
621 delete it;
622 it = new ID3::Iterator(id3, kMap[i].tag2);
623 }
624
625 if (it->done()) {
626 delete it;
627 continue;
628 }
629
630 String8 s;
631 it->getString(&s);
632 delete it;
633
634 meta->setCString(kMap[i].key, s);
635 }
636
637 size_t dataSize;
638 String8 mime;
639 const void *data = id3.getAlbumArt(&dataSize, &mime);
640
641 if (data) {
642 meta->setData(kKeyAlbumArt, MetaData::TYPE_NONE, data, dataSize);
643 meta->setCString(kKeyAlbumArtMIME, mime.string());
644 }
645
646 return meta;
647 }
648
SniffMP3(const sp<DataSource> & source,String8 * mimeType,float * confidence,sp<AMessage> * meta)649 bool SniffMP3(
650 const sp<DataSource> &source, String8 *mimeType,
651 float *confidence, sp<AMessage> *meta) {
652 off64_t pos = 0;
653 off64_t post_id3_pos;
654 uint32_t header;
655 if (!Resync(source, 0, &pos, &post_id3_pos, &header)) {
656 return false;
657 }
658
659 *meta = new AMessage;
660 (*meta)->setInt64("offset", pos);
661 (*meta)->setInt32("header", header);
662 (*meta)->setInt64("post-id3-offset", post_id3_pos);
663
664 *mimeType = MEDIA_MIMETYPE_AUDIO_MPEG;
665 *confidence = 0.2f;
666
667 return true;
668 }
669
670 } // namespace android
671