Add helper method to determine the current application horizontal size classes.
[chromium-blink-merge.git] / media / base / container_names.cc
blob48523d11ee7805e17ba1a7577ecb4ae3cee6849e
1 // Copyright (c) 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/base/container_names.h"
7 #include <cctype>
8 #include <limits>
10 #include "base/logging.h"
11 #include "media/base/bit_reader.h"
13 namespace media {
15 namespace container_names {
17 #define TAG(a, b, c, d) \
18 ((static_cast<uint32>(static_cast<uint8>(a)) << 24) | \
19 (static_cast<uint32>(static_cast<uint8>(b)) << 16) | \
20 (static_cast<uint32>(static_cast<uint8>(c)) << 8) | \
21 (static_cast<uint32>(static_cast<uint8>(d))))
23 #define RCHECK(x) \
24 do { \
25 if (!(x)) \
26 return false; \
27 } while (0)
29 #define UTF8_BYTE_ORDER_MARK "\xef\xbb\xbf"
31 // Helper function to read 2 bytes (16 bits, big endian) from a buffer.
32 static int Read16(const uint8* p) {
33 return p[0] << 8 | p[1];
36 // Helper function to read 3 bytes (24 bits, big endian) from a buffer.
37 static uint32 Read24(const uint8* p) {
38 return p[0] << 16 | p[1] << 8 | p[2];
41 // Helper function to read 4 bytes (32 bits, big endian) from a buffer.
42 static uint32 Read32(const uint8* p) {
43 return p[0] << 24 | p[1] << 16 | p[2] << 8 | p[3];
46 // Helper function to read 4 bytes (32 bits, little endian) from a buffer.
47 static uint32 Read32LE(const uint8* p) {
48 return p[3] << 24 | p[2] << 16 | p[1] << 8 | p[0];
51 // Helper function to do buffer comparisons with a string without going off the
52 // end of the buffer.
53 static bool StartsWith(const uint8* buffer,
54 size_t buffer_size,
55 const char* prefix) {
56 size_t prefix_size = strlen(prefix);
57 return (prefix_size <= buffer_size &&
58 memcmp(buffer, prefix, prefix_size) == 0);
61 // Helper function to do buffer comparisons with another buffer (to allow for
62 // embedded \0 in the comparison) without going off the end of the buffer.
63 static bool StartsWith(const uint8* buffer,
64 size_t buffer_size,
65 const uint8* prefix,
66 size_t prefix_size) {
67 return (prefix_size <= buffer_size &&
68 memcmp(buffer, prefix, prefix_size) == 0);
71 // Helper function to read up to 64 bits from a bit stream.
72 static uint64 ReadBits(BitReader* reader, int num_bits) {
73 DCHECK_GE(reader->bits_available(), num_bits);
74 DCHECK((num_bits > 0) && (num_bits <= 64));
75 uint64 value;
76 reader->ReadBits(num_bits, &value);
77 return value;
80 const int kAc3FrameSizeTable[38][3] = {
81 { 128, 138, 192 }, { 128, 140, 192 }, { 160, 174, 240 }, { 160, 176, 240 },
82 { 192, 208, 288 }, { 192, 210, 288 }, { 224, 242, 336 }, { 224, 244, 336 },
83 { 256, 278, 384 }, { 256, 280, 384 }, { 320, 348, 480 }, { 320, 350, 480 },
84 { 384, 416, 576 }, { 384, 418, 576 }, { 448, 486, 672 }, { 448, 488, 672 },
85 { 512, 556, 768 }, { 512, 558, 768 }, { 640, 696, 960 }, { 640, 698, 960 },
86 { 768, 834, 1152 }, { 768, 836, 1152 }, { 896, 974, 1344 },
87 { 896, 976, 1344 }, { 1024, 1114, 1536 }, { 1024, 1116, 1536 },
88 { 1280, 1392, 1920 }, { 1280, 1394, 1920 }, { 1536, 1670, 2304 },
89 { 1536, 1672, 2304 }, { 1792, 1950, 2688 }, { 1792, 1952, 2688 },
90 { 2048, 2228, 3072 }, { 2048, 2230, 3072 }, { 2304, 2506, 3456 },
91 { 2304, 2508, 3456 }, { 2560, 2768, 3840 }, { 2560, 2770, 3840 }
94 // Checks for an ADTS AAC container.
95 static bool CheckAac(const uint8* buffer, int buffer_size) {
96 // Audio Data Transport Stream (ADTS) header is 7 or 9 bytes
97 // (from http://wiki.multimedia.cx/index.php?title=ADTS)
98 RCHECK(buffer_size > 6);
100 int offset = 0;
101 while (offset + 6 < buffer_size) {
102 BitReader reader(buffer + offset, 6);
104 // Syncword must be 0xfff.
105 RCHECK(ReadBits(&reader, 12) == 0xfff);
107 // Skip MPEG version.
108 reader.SkipBits(1);
110 // Layer is always 0.
111 RCHECK(ReadBits(&reader, 2) == 0);
113 // Skip protection + profile.
114 reader.SkipBits(1 + 2);
116 // Check sampling frequency index.
117 RCHECK(ReadBits(&reader, 4) != 15); // Forbidden.
119 // Skip private stream, channel configuration, originality, home,
120 // copyrighted stream, and copyright_start.
121 reader.SkipBits(1 + 3 + 1 + 1 + 1 + 1);
123 // Get frame length (includes header).
124 int size = ReadBits(&reader, 13);
125 RCHECK(size > 0);
126 offset += size;
128 return true;
131 const uint16 kAc3SyncWord = 0x0b77;
133 // Checks for an AC3 container.
134 static bool CheckAc3(const uint8* buffer, int buffer_size) {
135 // Reference: ATSC Standard: Digital Audio Compression (AC-3, E-AC-3)
136 // Doc. A/52:2012
137 // (http://www.atsc.org/cms/standards/A52-2012(12-17).pdf)
139 // AC3 container looks like syncinfo | bsi | audblk * 6 | aux | check.
140 RCHECK(buffer_size > 6);
142 int offset = 0;
143 while (offset + 6 < buffer_size) {
144 BitReader reader(buffer + offset, 6);
146 // Check syncinfo.
147 RCHECK(ReadBits(&reader, 16) == kAc3SyncWord);
149 // Skip crc1.
150 reader.SkipBits(16);
152 // Verify fscod.
153 int sample_rate_code = ReadBits(&reader, 2);
154 RCHECK(sample_rate_code != 3); // Reserved.
156 // Verify frmsizecod.
157 int frame_size_code = ReadBits(&reader, 6);
158 RCHECK(frame_size_code < 38); // Undefined.
160 // Verify bsid.
161 RCHECK(ReadBits(&reader, 5) < 10); // Normally 8 or 6, 16 used by EAC3.
163 offset += kAc3FrameSizeTable[frame_size_code][sample_rate_code];
165 return true;
168 // Checks for an EAC3 container (very similar to AC3)
169 static bool CheckEac3(const uint8* buffer, int buffer_size) {
170 // Reference: ATSC Standard: Digital Audio Compression (AC-3, E-AC-3)
171 // Doc. A/52:2012
172 // (http://www.atsc.org/cms/standards/A52-2012(12-17).pdf)
174 // EAC3 container looks like syncinfo | bsi | audfrm | audblk* | aux | check.
175 RCHECK(buffer_size > 6);
177 int offset = 0;
178 while (offset + 6 < buffer_size) {
179 BitReader reader(buffer + offset, 6);
181 // Check syncinfo.
182 RCHECK(ReadBits(&reader, 16) == kAc3SyncWord);
184 // Verify strmtyp.
185 RCHECK(ReadBits(&reader, 2) != 3);
187 // Skip substreamid.
188 reader.SkipBits(3);
190 // Get frmsize. Include syncinfo size and convert to bytes.
191 int frame_size = (ReadBits(&reader, 11) + 1) * 2;
192 RCHECK(frame_size >= 7);
194 // Skip fscod, fscod2, acmod, and lfeon.
195 reader.SkipBits(2 + 2 + 3 + 1);
197 // Verify bsid.
198 int bit_stream_id = ReadBits(&reader, 5);
199 RCHECK(bit_stream_id >= 11 && bit_stream_id <= 16);
201 offset += frame_size;
203 return true;
206 // Additional checks for a BINK container.
207 static bool CheckBink(const uint8* buffer, int buffer_size) {
208 // Reference: http://wiki.multimedia.cx/index.php?title=Bink_Container
209 RCHECK(buffer_size >= 44);
211 // Verify number of frames specified.
212 RCHECK(Read32LE(buffer + 8) > 0);
214 // Verify width in range.
215 int width = Read32LE(buffer + 20);
216 RCHECK(width > 0 && width <= 32767);
218 // Verify height in range.
219 int height = Read32LE(buffer + 24);
220 RCHECK(height > 0 && height <= 32767);
222 // Verify frames per second specified.
223 RCHECK(Read32LE(buffer + 28) > 0);
225 // Verify video frames per second specified.
226 RCHECK(Read32LE(buffer + 32) > 0);
228 // Number of audio tracks must be 256 or less.
229 return (Read32LE(buffer + 40) <= 256);
232 // Additional checks for a CAF container.
233 static bool CheckCaf(const uint8* buffer, int buffer_size) {
234 // Reference: Apple Core Audio Format Specification 1.0
235 // (https://developer.apple.com/library/mac/#documentation/MusicAudio/Reference/CAFSpec/CAF_spec/CAF_spec.html)
236 RCHECK(buffer_size >= 52);
237 BitReader reader(buffer, buffer_size);
239 // mFileType should be "caff".
240 RCHECK(ReadBits(&reader, 32) == TAG('c', 'a', 'f', 'f'));
242 // mFileVersion should be 1.
243 RCHECK(ReadBits(&reader, 16) == 1);
245 // Skip mFileFlags.
246 reader.SkipBits(16);
248 // First chunk should be Audio Description chunk, size 32l.
249 RCHECK(ReadBits(&reader, 32) == TAG('d', 'e', 's', 'c'));
250 RCHECK(ReadBits(&reader, 64) == 32);
252 // CAFAudioFormat.mSampleRate(float64) not 0
253 RCHECK(ReadBits(&reader, 64) != 0);
255 // CAFAudioFormat.mFormatID not 0
256 RCHECK(ReadBits(&reader, 32) != 0);
258 // Skip CAFAudioFormat.mBytesPerPacket and mFramesPerPacket.
259 reader.SkipBits(32 + 32);
261 // CAFAudioFormat.mChannelsPerFrame not 0
262 RCHECK(ReadBits(&reader, 32) != 0);
263 return true;
266 static bool kSamplingFrequencyValid[16] = { false, true, true, true, false,
267 false, true, true, true, false,
268 false, true, true, true, false,
269 false };
270 static bool kExtAudioIdValid[8] = { true, false, true, false, false, false,
271 true, false };
273 // Additional checks for a DTS container.
274 static bool CheckDts(const uint8* buffer, int buffer_size) {
275 // Reference: ETSI TS 102 114 V1.3.1 (2011-08)
276 // (http://www.etsi.org/deliver/etsi_ts/102100_102199/102114/01.03.01_60/ts_102114v010301p.pdf)
277 RCHECK(buffer_size > 11);
279 int offset = 0;
280 while (offset + 11 < buffer_size) {
281 BitReader reader(buffer + offset, 11);
283 // Verify sync word.
284 RCHECK(ReadBits(&reader, 32) == 0x7ffe8001);
286 // Skip frame type and deficit sample count.
287 reader.SkipBits(1 + 5);
289 // Verify CRC present flag.
290 RCHECK(ReadBits(&reader, 1) == 0); // CPF must be 0.
292 // Verify number of PCM sample blocks.
293 RCHECK(ReadBits(&reader, 7) >= 5);
295 // Verify primary frame byte size.
296 int frame_size = ReadBits(&reader, 14);
297 RCHECK(frame_size >= 95);
299 // Skip audio channel arrangement.
300 reader.SkipBits(6);
302 // Verify core audio sampling frequency is an allowed value.
303 RCHECK(kSamplingFrequencyValid[ReadBits(&reader, 4)]);
305 // Verify transmission bit rate is valid.
306 RCHECK(ReadBits(&reader, 5) <= 25);
308 // Verify reserved field is 0.
309 RCHECK(ReadBits(&reader, 1) == 0);
311 // Skip dynamic range flag, time stamp flag, auxiliary data flag, and HDCD.
312 reader.SkipBits(1 + 1 + 1 + 1);
314 // Verify extension audio descriptor flag is an allowed value.
315 RCHECK(kExtAudioIdValid[ReadBits(&reader, 3)]);
317 // Skip extended coding flag and audio sync word insertion flag.
318 reader.SkipBits(1 + 1);
320 // Verify low frequency effects flag is an allowed value.
321 RCHECK(ReadBits(&reader, 2) != 3);
323 offset += frame_size + 1;
325 return true;
328 // Checks for a DV container.
329 static bool CheckDV(const uint8* buffer, int buffer_size) {
330 // Reference: SMPTE 314M (Annex A has differences with IEC 61834).
331 // (http://standards.smpte.org/content/978-1-61482-454-1/st-314-2005/SEC1.body.pdf)
332 RCHECK(buffer_size > 11);
334 int offset = 0;
335 int current_sequence_number = -1;
336 int last_block_number[6] = {0};
337 while (offset + 11 < buffer_size) {
338 BitReader reader(buffer + offset, 11);
340 // Decode ID data. Sections 5, 6, and 7 are reserved.
341 int section = ReadBits(&reader, 3);
342 RCHECK(section < 5);
344 // Next bit must be 1.
345 RCHECK(ReadBits(&reader, 1) == 1);
347 // Skip arbitrary bits.
348 reader.SkipBits(4);
350 int sequence_number = ReadBits(&reader, 4);
352 // Skip FSC.
353 reader.SkipBits(1);
355 // Next 3 bits must be 1.
356 RCHECK(ReadBits(&reader, 3) == 7);
358 int block_number = ReadBits(&reader, 8);
360 if (section == 0) { // Header.
361 // Validate the reserved bits in the next 8 bytes.
362 reader.SkipBits(1);
363 RCHECK(ReadBits(&reader, 1) == 0);
364 RCHECK(ReadBits(&reader, 11) == 0x7ff);
365 reader.SkipBits(4);
366 RCHECK(ReadBits(&reader, 4) == 0xf);
367 reader.SkipBits(4);
368 RCHECK(ReadBits(&reader, 4) == 0xf);
369 reader.SkipBits(4);
370 RCHECK(ReadBits(&reader, 4) == 0xf);
371 reader.SkipBits(3);
372 RCHECK(ReadBits(&reader, 24) == 0xffffff);
373 current_sequence_number = sequence_number;
374 for (size_t i = 0; i < arraysize(last_block_number); ++i)
375 last_block_number[i] = -1;
376 } else {
377 // Sequence number must match (this will also fail if no header seen).
378 RCHECK(sequence_number == current_sequence_number);
379 // Block number should be increasing.
380 RCHECK(block_number > last_block_number[section]);
381 last_block_number[section] = block_number;
384 // Move to next block.
385 offset += 80;
387 return true;
391 // Checks for a GSM container.
392 static bool CheckGsm(const uint8* buffer, int buffer_size) {
393 // Reference: ETSI EN 300 961 V8.1.1
394 // (http://www.etsi.org/deliver/etsi_en/300900_300999/300961/08.01.01_60/en_300961v080101p.pdf)
395 // also http://tools.ietf.org/html/rfc3551#page-24
396 // GSM files have a 33 byte block, only first 4 bits are fixed.
397 RCHECK(buffer_size >= 1024); // Need enough data to do a decent check.
399 int offset = 0;
400 while (offset < buffer_size) {
401 // First 4 bits of each block are xD.
402 RCHECK((buffer[offset] & 0xf0) == 0xd0);
403 offset += 33;
405 return true;
408 // Advance to the first set of |num_bits| bits that match |start_code|. |offset|
409 // is the current location in the buffer, and is updated. |bytes_needed| is the
410 // number of bytes that must remain in the buffer when |start_code| is found.
411 // Returns true if start_code found (and enough space in the buffer after it),
412 // false otherwise.
413 static bool AdvanceToStartCode(const uint8* buffer,
414 int buffer_size,
415 int* offset,
416 int bytes_needed,
417 int num_bits,
418 uint32 start_code) {
419 DCHECK_GE(bytes_needed, 3);
420 DCHECK_LE(num_bits, 24); // Only supports up to 24 bits.
422 // Create a mask to isolate |num_bits| bits, once shifted over.
423 uint32 bits_to_shift = 24 - num_bits;
424 uint32 mask = (1 << num_bits) - 1;
425 while (*offset + bytes_needed < buffer_size) {
426 uint32 next = Read24(buffer + *offset);
427 if (((next >> bits_to_shift) & mask) == start_code)
428 return true;
429 ++(*offset);
431 return false;
434 // Checks for an H.261 container.
435 static bool CheckH261(const uint8* buffer, int buffer_size) {
436 // Reference: ITU-T Recommendation H.261 (03/1993)
437 // (http://www.itu.int/rec/T-REC-H.261-199303-I/en)
438 RCHECK(buffer_size > 16);
440 int offset = 0;
441 bool seen_start_code = false;
442 while (true) {
443 // Advance to picture_start_code, if there is one.
444 if (!AdvanceToStartCode(buffer, buffer_size, &offset, 4, 20, 0x10)) {
445 // No start code found (or off end of buffer), so success if
446 // there was at least one valid header.
447 return seen_start_code;
450 // Now verify the block. AdvanceToStartCode() made sure that there are
451 // at least 4 bytes remaining in the buffer.
452 BitReader reader(buffer + offset, buffer_size - offset);
453 RCHECK(ReadBits(&reader, 20) == 0x10);
455 // Skip the temporal reference and PTYPE.
456 reader.SkipBits(5 + 6);
458 // Skip any extra insertion information. Since this is open-ended, if we run
459 // out of bits assume that the buffer is correctly formatted.
460 int extra = ReadBits(&reader, 1);
461 while (extra == 1) {
462 if (!reader.SkipBits(8))
463 return seen_start_code;
464 if (!reader.ReadBits(1, &extra))
465 return seen_start_code;
468 // Next should be a Group of Blocks start code. Again, if we run out of
469 // bits, then assume that the buffer up to here is correct, and the buffer
470 // just happened to end in the middle of a header.
471 int next;
472 if (!reader.ReadBits(16, &next))
473 return seen_start_code;
474 RCHECK(next == 1);
476 // Move to the next block.
477 seen_start_code = true;
478 offset += 4;
482 // Checks for an H.263 container.
483 static bool CheckH263(const uint8* buffer, int buffer_size) {
484 // Reference: ITU-T Recommendation H.263 (01/2005)
485 // (http://www.itu.int/rec/T-REC-H.263-200501-I/en)
486 // header is PSC(22b) + TR(8b) + PTYPE(8+b).
487 RCHECK(buffer_size > 16);
489 int offset = 0;
490 bool seen_start_code = false;
491 while (true) {
492 // Advance to picture_start_code, if there is one.
493 if (!AdvanceToStartCode(buffer, buffer_size, &offset, 9, 22, 0x20)) {
494 // No start code found (or off end of buffer), so success if
495 // there was at least one valid header.
496 return seen_start_code;
499 // Now verify the block. AdvanceToStartCode() made sure that there are
500 // at least 9 bytes remaining in the buffer.
501 BitReader reader(buffer + offset, 9);
502 RCHECK(ReadBits(&reader, 22) == 0x20);
504 // Skip the temporal reference.
505 reader.SkipBits(8);
507 // Verify that the first 2 bits of PTYPE are 10b.
508 RCHECK(ReadBits(&reader, 2) == 2);
510 // Skip the split screen indicator, document camera indicator, and full
511 // picture freeze release.
512 reader.SkipBits(1 + 1 + 1);
514 // Verify Source Format.
515 int format = ReadBits(&reader, 3);
516 RCHECK(format != 0 && format != 6); // Forbidden or reserved.
518 if (format == 7) {
519 // Verify full extended PTYPE.
520 int ufep = ReadBits(&reader, 3);
521 if (ufep == 1) {
522 // Verify the optional part of PLUSPTYPE.
523 format = ReadBits(&reader, 3);
524 RCHECK(format != 0 && format != 7); // Reserved.
525 reader.SkipBits(11);
526 // Next 4 bits should be b1000.
527 RCHECK(ReadBits(&reader, 4) == 8); // Not allowed.
528 } else {
529 RCHECK(ufep == 0); // Only 0 and 1 allowed.
532 // Verify picture type code is not a reserved value.
533 int picture_type_code = ReadBits(&reader, 3);
534 RCHECK(picture_type_code != 6 && picture_type_code != 7); // Reserved.
536 // Skip picture resampling mode, reduced resolution mode,
537 // and rounding type.
538 reader.SkipBits(1 + 1 + 1);
540 // Next 3 bits should be b001.
541 RCHECK(ReadBits(&reader, 3) == 1); // Not allowed.
544 // Move to the next block.
545 seen_start_code = true;
546 offset += 9;
550 // Checks for an H.264 container.
551 static bool CheckH264(const uint8* buffer, int buffer_size) {
552 // Reference: ITU-T Recommendation H.264 (01/2012)
553 // (http://www.itu.int/rec/T-REC-H.264)
554 // Section B.1: Byte stream NAL unit syntax and semantics.
555 RCHECK(buffer_size > 4);
557 int offset = 0;
558 int parameter_count = 0;
559 while (true) {
560 // Advance to picture_start_code, if there is one.
561 if (!AdvanceToStartCode(buffer, buffer_size, &offset, 4, 24, 1)) {
562 // No start code found (or off end of buffer), so success if
563 // there was at least one valid header.
564 return parameter_count > 0;
567 // Now verify the block. AdvanceToStartCode() made sure that there are
568 // at least 4 bytes remaining in the buffer.
569 BitReader reader(buffer + offset, 4);
570 RCHECK(ReadBits(&reader, 24) == 1);
572 // Verify forbidden_zero_bit.
573 RCHECK(ReadBits(&reader, 1) == 0);
575 // Extract nal_ref_idc and nal_unit_type.
576 int nal_ref_idc = ReadBits(&reader, 2);
577 int nal_unit_type = ReadBits(&reader, 5);
579 switch (nal_unit_type) {
580 case 5: // Coded slice of an IDR picture.
581 RCHECK(nal_ref_idc != 0);
582 break;
583 case 6: // Supplemental enhancement information (SEI).
584 case 9: // Access unit delimiter.
585 case 10: // End of sequence.
586 case 11: // End of stream.
587 case 12: // Filler data.
588 RCHECK(nal_ref_idc == 0);
589 break;
590 case 7: // Sequence parameter set.
591 case 8: // Picture parameter set.
592 ++parameter_count;
593 break;
596 // Skip the current start_code_prefix and move to the next.
597 offset += 4;
601 static const char kHlsSignature[] = "#EXTM3U";
602 static const char kHls1[] = "#EXT-X-STREAM-INF:";
603 static const char kHls2[] = "#EXT-X-TARGETDURATION:";
604 static const char kHls3[] = "#EXT-X-MEDIA-SEQUENCE:";
606 // Additional checks for a HLS container.
607 static bool CheckHls(const uint8* buffer, int buffer_size) {
608 // HLS is simply a play list used for Apple HTTP Live Streaming.
609 // Reference: Apple HTTP Live Streaming Overview
610 // (http://goo.gl/MIwxj)
612 if (StartsWith(buffer, buffer_size, kHlsSignature)) {
613 // Need to find "#EXT-X-STREAM-INF:", "#EXT-X-TARGETDURATION:", or
614 // "#EXT-X-MEDIA-SEQUENCE:" somewhere in the buffer. Other playlists (like
615 // WinAmp) only have additional lines with #EXTINF
616 // (http://en.wikipedia.org/wiki/M3U).
617 int offset = strlen(kHlsSignature);
618 while (offset < buffer_size) {
619 if (buffer[offset] == '#') {
620 if (StartsWith(buffer + offset, buffer_size - offset, kHls1) ||
621 StartsWith(buffer + offset, buffer_size - offset, kHls2) ||
622 StartsWith(buffer + offset, buffer_size - offset, kHls3)) {
623 return true;
626 ++offset;
629 return false;
632 // Checks for a MJPEG stream.
633 static bool CheckMJpeg(const uint8* buffer, int buffer_size) {
634 // Reference: ISO/IEC 10918-1 : 1993(E), Annex B
635 // (http://www.w3.org/Graphics/JPEG/itu-t81.pdf)
636 RCHECK(buffer_size >= 16);
638 int offset = 0;
639 int last_restart = -1;
640 int num_codes = 0;
641 while (offset + 5 < buffer_size) {
642 // Marker codes are always a two byte code with the first byte xFF.
643 RCHECK(buffer[offset] == 0xff);
644 uint8 code = buffer[offset + 1];
645 RCHECK(code >= 0xc0 || code == 1);
647 // Skip sequences of xFF.
648 if (code == 0xff) {
649 ++offset;
650 continue;
653 // Success if the next marker code is EOI (end of image)
654 if (code == 0xd9)
655 return true;
657 // Check remaining codes.
658 if (code == 0xd8 || code == 1) {
659 // SOI (start of image) / TEM (private use). No other data with header.
660 offset += 2;
661 } else if (code >= 0xd0 && code <= 0xd7) {
662 // RST (restart) codes must be in sequence. No other data with header.
663 int restart = code & 0x07;
664 if (last_restart >= 0)
665 RCHECK(restart == (last_restart + 1) % 8);
666 last_restart = restart;
667 offset += 2;
668 } else {
669 // All remaining marker codes are followed by a length of the header.
670 int length = Read16(buffer + offset + 2) + 2;
672 // Special handling of SOS (start of scan) marker since the entropy
673 // coded data follows the SOS. Any xFF byte in the data block must be
674 // followed by x00 in the data.
675 if (code == 0xda) {
676 int number_components = buffer[offset + 4];
677 RCHECK(length == 8 + 2 * number_components);
679 // Advance to the next marker.
680 offset += length;
681 while (offset + 2 < buffer_size) {
682 if (buffer[offset] == 0xff && buffer[offset + 1] != 0)
683 break;
684 ++offset;
686 } else {
687 // Skip over the marker data for the other marker codes.
688 offset += length;
691 ++num_codes;
693 return (num_codes > 1);
696 enum Mpeg2StartCodes {
697 PROGRAM_END_CODE = 0xb9,
698 PACK_START_CODE = 0xba
701 // Checks for a MPEG2 Program Stream.
702 static bool CheckMpeg2ProgramStream(const uint8* buffer, int buffer_size) {
703 // Reference: ISO/IEC 13818-1 : 2000 (E) / ITU-T Rec. H.222.0 (2000 E).
704 RCHECK(buffer_size > 14);
706 int offset = 0;
707 while (offset + 14 < buffer_size) {
708 BitReader reader(buffer + offset, 14);
710 // Must start with pack_start_code.
711 RCHECK(ReadBits(&reader, 24) == 1);
712 RCHECK(ReadBits(&reader, 8) == PACK_START_CODE);
714 // Determine MPEG version (MPEG1 has b0010, while MPEG2 has b01).
715 int mpeg_version = ReadBits(&reader, 2);
716 if (mpeg_version == 0) {
717 // MPEG1, 10 byte header
718 // Validate rest of version code
719 RCHECK(ReadBits(&reader, 2) == 2);
720 } else {
721 RCHECK(mpeg_version == 1);
724 // Skip system_clock_reference_base [32..30].
725 reader.SkipBits(3);
727 // Verify marker bit.
728 RCHECK(ReadBits(&reader, 1) == 1);
730 // Skip system_clock_reference_base [29..15].
731 reader.SkipBits(15);
733 // Verify next marker bit.
734 RCHECK(ReadBits(&reader, 1) == 1);
736 // Skip system_clock_reference_base [14..0].
737 reader.SkipBits(15);
739 // Verify next marker bit.
740 RCHECK(ReadBits(&reader, 1) == 1);
742 if (mpeg_version == 0) {
743 // Verify second marker bit.
744 RCHECK(ReadBits(&reader, 1) == 1);
746 // Skip mux_rate.
747 reader.SkipBits(22);
749 // Verify next marker bit.
750 RCHECK(ReadBits(&reader, 1) == 1);
752 // Update offset to be after this header.
753 offset += 12;
754 } else {
755 // Must be MPEG2.
756 // Skip program_mux_rate.
757 reader.SkipBits(22);
759 // Verify pair of marker bits.
760 RCHECK(ReadBits(&reader, 2) == 3);
762 // Skip reserved.
763 reader.SkipBits(5);
765 // Update offset to be after this header.
766 int pack_stuffing_length = ReadBits(&reader, 3);
767 offset += 14 + pack_stuffing_length;
770 // Check for system headers and PES_packets.
771 while (offset + 6 < buffer_size && Read24(buffer + offset) == 1) {
772 // Next 8 bits determine stream type.
773 int stream_id = buffer[offset + 3];
775 // Some stream types are reserved and shouldn't occur.
776 if (mpeg_version == 0)
777 RCHECK(stream_id != 0xbc && stream_id < 0xf0);
778 else
779 RCHECK(stream_id != 0xfc && stream_id != 0xfd && stream_id != 0xfe);
781 // Some stream types are used for pack headers.
782 if (stream_id == PACK_START_CODE) // back to outer loop.
783 break;
784 if (stream_id == PROGRAM_END_CODE) // end of stream.
785 return true;
787 int pes_length = Read16(buffer + offset + 4);
788 RCHECK(pes_length > 0);
789 offset = offset + 6 + pes_length;
792 // Success as we are off the end of the buffer and liked everything
793 // in the buffer.
794 return true;
797 const uint8 kMpeg2SyncWord = 0x47;
799 // Checks for a MPEG2 Transport Stream.
800 static bool CheckMpeg2TransportStream(const uint8* buffer, int buffer_size) {
801 // Spec: ISO/IEC 13818-1 : 2000 (E) / ITU-T Rec. H.222.0 (2000 E).
802 // Normal packet size is 188 bytes. However, some systems add various error
803 // correction data at the end, resulting in packet of length 192/204/208
804 // (https://en.wikipedia.org/wiki/MPEG_transport_stream). Determine the
805 // length with the first packet.
806 RCHECK(buffer_size >= 250); // Want more than 1 packet to check.
808 int offset = 0;
809 int packet_length = -1;
810 while (buffer[offset] != kMpeg2SyncWord && offset < 20) {
811 // Skip over any header in the first 20 bytes.
812 ++offset;
815 while (offset + 6 < buffer_size) {
816 BitReader reader(buffer + offset, 6);
818 // Must start with sync byte.
819 RCHECK(ReadBits(&reader, 8) == kMpeg2SyncWord);
821 // Skip transport_error_indicator, payload_unit_start_indicator, and
822 // transport_priority.
823 reader.SkipBits(1 + 1 + 1);
825 // Verify the pid is not a reserved value.
826 int pid = ReadBits(&reader, 13);
827 RCHECK(pid < 3 || pid > 15);
829 // Skip transport_scrambling_control.
830 reader.SkipBits(2);
832 // Adaptation_field_control can not be 0.
833 int adaptation_field_control = ReadBits(&reader, 2);
834 RCHECK(adaptation_field_control != 0);
836 // If there is an adaptation_field, verify it.
837 if (adaptation_field_control >= 2) {
838 // Skip continuity_counter.
839 reader.SkipBits(4);
841 // Get adaptation_field_length and verify it.
842 int adaptation_field_length = ReadBits(&reader, 8);
843 if (adaptation_field_control == 2)
844 RCHECK(adaptation_field_length == 183);
845 else
846 RCHECK(adaptation_field_length <= 182);
849 // Attempt to determine the packet length on the first packet.
850 if (packet_length < 0) {
851 if (buffer[offset + 188] == kMpeg2SyncWord)
852 packet_length = 188;
853 else if (buffer[offset + 192] == kMpeg2SyncWord)
854 packet_length = 192;
855 else if (buffer[offset + 204] == kMpeg2SyncWord)
856 packet_length = 204;
857 else
858 packet_length = 208;
860 offset += packet_length;
862 return true;
865 enum Mpeg4StartCodes {
866 VISUAL_OBJECT_SEQUENCE_START_CODE = 0xb0,
867 VISUAL_OBJECT_SEQUENCE_END_CODE = 0xb1,
868 VISUAL_OBJECT_START_CODE = 0xb5,
869 VOP_START_CODE = 0xb6
872 // Checks for a raw MPEG4 bitstream container.
873 static bool CheckMpeg4BitStream(const uint8* buffer, int buffer_size) {
874 // Defined in ISO/IEC 14496-2:2001.
875 // However, no length ... simply scan for start code values.
876 // Note tags are very similar to H.264.
877 RCHECK(buffer_size > 4);
879 int offset = 0;
880 int sequence_start_count = 0;
881 int sequence_end_count = 0;
882 int visual_object_count = 0;
883 int vop_count = 0;
884 while (true) {
885 // Advance to start_code, if there is one.
886 if (!AdvanceToStartCode(buffer, buffer_size, &offset, 6, 24, 1)) {
887 // Not a complete sequence in memory, so return true if we've seen a
888 // visual_object_sequence_start_code and a visual_object_start_code.
889 return (sequence_start_count > 0 && visual_object_count > 0);
892 // Now verify the block. AdvanceToStartCode() made sure that there are
893 // at least 6 bytes remaining in the buffer.
894 BitReader reader(buffer + offset, 6);
895 RCHECK(ReadBits(&reader, 24) == 1);
897 int start_code = ReadBits(&reader, 8);
898 RCHECK(start_code < 0x30 || start_code > 0xaf); // 30..AF and
899 RCHECK(start_code < 0xb7 || start_code > 0xb9); // B7..B9 reserved
901 switch (start_code) {
902 case VISUAL_OBJECT_SEQUENCE_START_CODE: {
903 ++sequence_start_count;
904 // Verify profile in not one of many reserved values.
905 int profile = ReadBits(&reader, 8);
906 RCHECK(profile > 0);
907 RCHECK(profile < 0x04 || profile > 0x10);
908 RCHECK(profile < 0x13 || profile > 0x20);
909 RCHECK(profile < 0x23 || profile > 0x31);
910 RCHECK(profile < 0x35 || profile > 0x41);
911 RCHECK(profile < 0x43 || profile > 0x60);
912 RCHECK(profile < 0x65 || profile > 0x70);
913 RCHECK(profile < 0x73 || profile > 0x80);
914 RCHECK(profile < 0x83 || profile > 0x90);
915 RCHECK(profile < 0x95 || profile > 0xa0);
916 RCHECK(profile < 0xa4 || profile > 0xb0);
917 RCHECK(profile < 0xb5 || profile > 0xc0);
918 RCHECK(profile < 0xc3 || profile > 0xd0);
919 RCHECK(profile < 0xe4);
920 break;
923 case VISUAL_OBJECT_SEQUENCE_END_CODE:
924 RCHECK(++sequence_end_count == sequence_start_count);
925 break;
927 case VISUAL_OBJECT_START_CODE: {
928 ++visual_object_count;
929 if (ReadBits(&reader, 1) == 1) {
930 int visual_object_verid = ReadBits(&reader, 4);
931 RCHECK(visual_object_verid > 0 && visual_object_verid < 3);
932 RCHECK(ReadBits(&reader, 3) != 0);
934 int visual_object_type = ReadBits(&reader, 4);
935 RCHECK(visual_object_type > 0 && visual_object_type < 6);
936 break;
939 case VOP_START_CODE:
940 RCHECK(++vop_count <= visual_object_count);
941 break;
943 // Skip this block.
944 offset += 6;
948 // Additional checks for a MOV/QuickTime/MPEG4 container.
949 static bool CheckMov(const uint8* buffer, int buffer_size) {
950 // Reference: ISO/IEC 14496-12:2005(E).
951 // (http://standards.iso.org/ittf/PubliclyAvailableStandards/c061988_ISO_IEC_14496-12_2012.zip)
952 RCHECK(buffer_size > 8);
954 int offset = 0;
955 while (offset + 8 < buffer_size) {
956 uint32 atomsize = Read32(buffer + offset);
957 uint32 atomtype = Read32(buffer + offset + 4);
958 // Only need to check for ones that are valid at the top level.
959 switch (atomtype) {
960 case TAG('f','t','y','p'):
961 case TAG('p','d','i','n'):
962 case TAG('m','o','o','v'):
963 case TAG('m','o','o','f'):
964 case TAG('m','f','r','a'):
965 case TAG('m','d','a','t'):
966 case TAG('f','r','e','e'):
967 case TAG('s','k','i','p'):
968 case TAG('m','e','t','a'):
969 case TAG('m','e','c','o'):
970 case TAG('s','t','y','p'):
971 case TAG('s','i','d','x'):
972 case TAG('s','s','i','x'):
973 case TAG('p','r','f','t'):
974 case TAG('b','l','o','c'):
975 break;
976 default:
977 return false;
979 if (atomsize == 1) {
980 // Indicates that the length is the next 64bits.
981 if (offset + 16 > buffer_size)
982 break;
983 if (Read32(buffer + offset + 8) != 0)
984 break; // Offset is way past buffer size.
985 atomsize = Read32(buffer + offset + 12);
987 if (atomsize == 0 || atomsize > static_cast<size_t>(buffer_size))
988 break; // Indicates the last atom or length too big.
989 offset += atomsize;
991 return true;
994 enum MPEGVersion {
995 VERSION_25 = 0,
996 VERSION_RESERVED,
997 VERSION_2,
998 VERSION_1
1000 enum MPEGLayer {
1001 L_RESERVED = 0,
1002 LAYER_3,
1003 LAYER_2,
1004 LAYER_1
1007 static int kSampleRateTable[4][4] = { { 11025, 12000, 8000, 0 }, // v2.5
1008 { 0, 0, 0, 0 }, // not used
1009 { 22050, 24000, 16000, 0 }, // v2
1010 { 44100, 48000, 32000, 0 } // v1
1013 static int kBitRateTableV1L1[16] = { 0, 32, 64, 96, 128, 160, 192, 224, 256,
1014 288, 320, 352, 384, 416, 448, 0 };
1015 static int kBitRateTableV1L2[16] = { 0, 32, 48, 56, 64, 80, 96, 112, 128, 160,
1016 192, 224, 256, 320, 384, 0 };
1017 static int kBitRateTableV1L3[16] = { 0, 32, 40, 48, 56, 64, 80, 96, 112, 128,
1018 160, 192, 224, 256, 320, 0 };
1019 static int kBitRateTableV2L1[16] = { 0, 32, 48, 56, 64, 80, 96, 112, 128, 144,
1020 160, 176, 192, 224, 256, 0 };
1021 static int kBitRateTableV2L23[16] = { 0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96,
1022 112, 128, 144, 160, 0 };
1024 static bool ValidMpegAudioFrameHeader(const uint8* header,
1025 int header_size,
1026 int* framesize) {
1027 // Reference: http://mpgedit.org/mpgedit/mpeg_format/mpeghdr.htm.
1028 DCHECK_GE(header_size, 4);
1029 *framesize = 0;
1030 BitReader reader(header, 4); // Header can only be 4 bytes long.
1032 // Verify frame sync (11 bits) are all set.
1033 RCHECK(ReadBits(&reader, 11) == 0x7ff);
1035 // Verify MPEG audio version id.
1036 int version = ReadBits(&reader, 2);
1037 RCHECK(version != 1); // Reserved.
1039 // Verify layer.
1040 int layer = ReadBits(&reader, 2);
1041 RCHECK(layer != 0);
1043 // Skip protection bit.
1044 reader.SkipBits(1);
1046 // Verify bitrate index.
1047 int bitrate_index = ReadBits(&reader, 4);
1048 RCHECK(bitrate_index != 0xf);
1050 // Verify sampling rate frequency index.
1051 int sampling_index = ReadBits(&reader, 2);
1052 RCHECK(sampling_index != 3);
1054 // Get padding bit.
1055 int padding = ReadBits(&reader, 1);
1057 // Frame size:
1058 // For Layer I files = (12 * BitRate / SampleRate + Padding) * 4
1059 // For others = 144 * BitRate / SampleRate + Padding
1060 // Unfortunately, BitRate and SampleRate are coded.
1061 int sampling_rate = kSampleRateTable[version][sampling_index];
1062 int bitrate;
1063 if (version == VERSION_1) {
1064 if (layer == LAYER_1)
1065 bitrate = kBitRateTableV1L1[bitrate_index];
1066 else if (layer == LAYER_2)
1067 bitrate = kBitRateTableV1L2[bitrate_index];
1068 else
1069 bitrate = kBitRateTableV1L3[bitrate_index];
1070 } else {
1071 if (layer == LAYER_1)
1072 bitrate = kBitRateTableV2L1[bitrate_index];
1073 else
1074 bitrate = kBitRateTableV2L23[bitrate_index];
1076 if (layer == LAYER_1)
1077 *framesize = ((12000 * bitrate) / sampling_rate + padding) * 4;
1078 else
1079 *framesize = (144000 * bitrate) / sampling_rate + padding;
1080 return (bitrate > 0 && sampling_rate > 0);
1083 // Extract a size encoded the MP3 way.
1084 static int GetMp3HeaderSize(const uint8* buffer, int buffer_size) {
1085 DCHECK_GE(buffer_size, 9);
1086 int size = ((buffer[6] & 0x7f) << 21) + ((buffer[7] & 0x7f) << 14) +
1087 ((buffer[8] & 0x7f) << 7) + (buffer[9] & 0x7f) + 10;
1088 if (buffer[5] & 0x10) // Footer added?
1089 size += 10;
1090 return size;
1093 // Additional checks for a MP3 container.
1094 static bool CheckMp3(const uint8* buffer, int buffer_size, bool seenHeader) {
1095 RCHECK(buffer_size >= 10); // Must be enough to read the initial header.
1097 int framesize;
1098 int numSeen = 0;
1099 int offset = 0;
1100 if (seenHeader) {
1101 offset = GetMp3HeaderSize(buffer, buffer_size);
1102 } else {
1103 // Skip over leading 0's.
1104 while (offset < buffer_size && buffer[offset] == 0)
1105 ++offset;
1108 while (offset + 3 < buffer_size) {
1109 RCHECK(ValidMpegAudioFrameHeader(
1110 buffer + offset, buffer_size - offset, &framesize));
1112 // Have we seen enough valid headers?
1113 if (++numSeen > 10)
1114 return true;
1115 offset += framesize;
1117 // Off the end of the buffer, return success if a few valid headers seen.
1118 return numSeen > 2;
1121 // Check that the next characters in |buffer| represent a number. The format
1122 // accepted is optional whitespace followed by 1 or more digits. |max_digits|
1123 // specifies the maximum number of digits to process. Returns true if a valid
1124 // number is found, false otherwise.
1125 static bool VerifyNumber(const uint8* buffer,
1126 int buffer_size,
1127 int* offset,
1128 int max_digits) {
1129 RCHECK(*offset < buffer_size);
1131 // Skip over any leading space.
1132 while (isspace(buffer[*offset])) {
1133 ++(*offset);
1134 RCHECK(*offset < buffer_size);
1137 // Need to process up to max_digits digits.
1138 int numSeen = 0;
1139 while (--max_digits >= 0 && isdigit(buffer[*offset])) {
1140 ++numSeen;
1141 ++(*offset);
1142 if (*offset >= buffer_size)
1143 return true; // Out of space but seen a digit.
1146 // Success if at least one digit seen.
1147 return (numSeen > 0);
1150 // Check that the next character in |buffer| is one of |c1| or |c2|. |c2| is
1151 // optional. Returns true if there is a match, false if no match or out of
1152 // space.
1153 static inline bool VerifyCharacters(const uint8* buffer,
1154 int buffer_size,
1155 int* offset,
1156 char c1,
1157 char c2) {
1158 RCHECK(*offset < buffer_size);
1159 char c = static_cast<char>(buffer[(*offset)++]);
1160 return (c == c1 || (c == c2 && c2 != 0));
1163 // Checks for a SRT container.
1164 static bool CheckSrt(const uint8* buffer, int buffer_size) {
1165 // Reference: http://en.wikipedia.org/wiki/SubRip
1166 RCHECK(buffer_size > 20);
1168 // First line should just be the subtitle sequence number.
1169 int offset = StartsWith(buffer, buffer_size, UTF8_BYTE_ORDER_MARK) ? 3 : 0;
1170 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 100));
1171 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, '\n', '\r'));
1173 // Skip any additional \n\r.
1174 while (VerifyCharacters(buffer, buffer_size, &offset, '\n', '\r')) {}
1175 --offset; // Since VerifyCharacters() gobbled up the next non-CR/LF.
1177 // Second line should look like the following:
1178 // 00:00:10,500 --> 00:00:13,000
1179 // Units separator can be , or .
1180 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 100));
1181 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, ':', 0));
1182 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 2));
1183 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, ':', 0));
1184 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 2));
1185 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, ',', '.'));
1186 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 3));
1187 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, ' ', 0));
1188 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, '-', 0));
1189 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, '-', 0));
1190 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, '>', 0));
1191 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, ' ', 0));
1192 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 100));
1193 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, ':', 0));
1194 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 2));
1195 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, ':', 0));
1196 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 2));
1197 RCHECK(VerifyCharacters(buffer, buffer_size, &offset, ',', '.'));
1198 RCHECK(VerifyNumber(buffer, buffer_size, &offset, 3));
1199 return true;
1202 // Read a Matroska Element Id.
1203 static int GetElementId(BitReader* reader) {
1204 // Element ID is coded with the leading zero bits (max 3) determining size.
1205 // If it is an invalid encoding or the end of the buffer is reached,
1206 // return -1 as a tag that won't be expected.
1207 if (reader->bits_available() >= 8) {
1208 int num_bits_to_read = 0;
1209 static int prefix[] = { 0x80, 0x4000, 0x200000, 0x10000000 };
1210 for (int i = 0; i < 4; ++i) {
1211 num_bits_to_read += 7;
1212 if (ReadBits(reader, 1) == 1) {
1213 if (reader->bits_available() < num_bits_to_read)
1214 break;
1215 // prefix[] adds back the bits read individually.
1216 return ReadBits(reader, num_bits_to_read) | prefix[i];
1220 // Invalid encoding, return something not expected.
1221 return -1;
1224 // Read a Matroska Unsigned Integer (VINT).
1225 static uint64 GetVint(BitReader* reader) {
1226 // Values are coded with the leading zero bits (max 7) determining size.
1227 // If it is an invalid coding or the end of the buffer is reached,
1228 // return something that will go off the end of the buffer.
1229 if (reader->bits_available() >= 8) {
1230 int num_bits_to_read = 0;
1231 for (int i = 0; i < 8; ++i) {
1232 num_bits_to_read += 7;
1233 if (ReadBits(reader, 1) == 1) {
1234 if (reader->bits_available() < num_bits_to_read)
1235 break;
1236 return ReadBits(reader, num_bits_to_read);
1240 // Incorrect format (more than 7 leading 0's) or off the end of the buffer.
1241 // Since the return value is used as a byte size, return a value that will
1242 // cause a failure when used.
1243 return (reader->bits_available() / 8) + 2;
1246 // Additional checks for a WEBM container.
1247 static bool CheckWebm(const uint8* buffer, int buffer_size) {
1248 // Reference: http://www.matroska.org/technical/specs/index.html
1249 RCHECK(buffer_size > 12);
1251 BitReader reader(buffer, buffer_size);
1253 // Verify starting Element Id.
1254 RCHECK(GetElementId(&reader) == 0x1a45dfa3);
1256 // Get the header size, and ensure there are enough bits to check.
1257 int header_size = GetVint(&reader);
1258 RCHECK(reader.bits_available() / 8 >= header_size);
1260 // Loop through the header.
1261 while (reader.bits_available() > 0) {
1262 int tag = GetElementId(&reader);
1263 int tagsize = GetVint(&reader);
1264 switch (tag) {
1265 case 0x4286: // EBMLVersion
1266 case 0x42f7: // EBMLReadVersion
1267 case 0x42f2: // EBMLMaxIdLength
1268 case 0x42f3: // EBMLMaxSizeLength
1269 case 0x4287: // DocTypeVersion
1270 case 0x4285: // DocTypeReadVersion
1271 case 0xec: // void
1272 case 0xbf: // CRC32
1273 RCHECK(reader.SkipBits(tagsize * 8));
1274 break;
1276 case 0x4282: // EBMLDocType
1277 // Need to see "webm" or "matroska" next.
1278 switch (ReadBits(&reader, 32)) {
1279 case TAG('w', 'e', 'b', 'm') :
1280 return true;
1281 case TAG('m', 'a', 't', 'r') :
1282 return (ReadBits(&reader, 32) == TAG('o', 's', 'k', 'a'));
1284 return false;
1286 default: // Unrecognized tag
1287 return false;
1290 return false;
1293 enum VC1StartCodes {
1294 VC1_FRAME_START_CODE = 0x0d,
1295 VC1_ENTRY_POINT_START_CODE = 0x0e,
1296 VC1_SEQUENCE_START_CODE = 0x0f
1299 // Checks for a VC1 bitstream container.
1300 static bool CheckVC1(const uint8* buffer, int buffer_size) {
1301 // Reference: SMPTE 421M
1302 // (http://standards.smpte.org/content/978-1-61482-555-5/st-421-2006/SEC1.body.pdf)
1303 // However, no length ... simply scan for start code values.
1304 // Expect to see SEQ | [ [ ENTRY ] PIC* ]*
1305 // Note tags are very similar to H.264.
1307 RCHECK(buffer_size >= 24);
1309 // First check for Bitstream Metadata Serialization (Annex L)
1310 if (buffer[0] == 0xc5 &&
1311 Read32(buffer + 4) == 0x04 &&
1312 Read32(buffer + 20) == 0x0c) {
1313 // Verify settings in STRUCT_C and STRUCT_A
1314 BitReader reader(buffer + 8, 12);
1316 int profile = ReadBits(&reader, 4);
1317 if (profile == 0 || profile == 4) { // simple or main
1318 // Skip FRMRTQ_POSTPROC, BITRTQ_POSTPROC, and LOOPFILTER.
1319 reader.SkipBits(3 + 5 + 1);
1321 // Next bit must be 0.
1322 RCHECK(ReadBits(&reader, 1) == 0);
1324 // Skip MULTIRES.
1325 reader.SkipBits(1);
1327 // Next bit must be 1.
1328 RCHECK(ReadBits(&reader, 1) == 1);
1330 // Skip FASTUVMC, EXTENDED_MV, DQUANT, and VSTRANSFORM.
1331 reader.SkipBits(1 + 1 + 2 + 1);
1333 // Next bit must be 0.
1334 RCHECK(ReadBits(&reader, 1) == 0);
1336 // Skip OVERLAP, SYNCMARKER, RANGERED, MAXBFRAMES, QUANTIZER, and
1337 // FINTERPFLAG.
1338 reader.SkipBits(1 + 1 + 1 + 3 + 2 + 1);
1340 // Next bit must be 1.
1341 RCHECK(ReadBits(&reader, 1) == 1);
1343 } else {
1344 RCHECK(profile == 12); // Other profile values not allowed.
1345 RCHECK(ReadBits(&reader, 28) == 0);
1348 // Now check HORIZ_SIZE and VERT_SIZE, which must be 8192 or less.
1349 RCHECK(ReadBits(&reader, 32) <= 8192);
1350 RCHECK(ReadBits(&reader, 32) <= 8192);
1351 return true;
1354 // Buffer isn't Bitstream Metadata, so scan for start codes.
1355 int offset = 0;
1356 int sequence_start_code = 0;
1357 int frame_start_code = 0;
1358 while (true) {
1359 // Advance to start_code, if there is one.
1360 if (!AdvanceToStartCode(buffer, buffer_size, &offset, 5, 24, 1)) {
1361 // Not a complete sequence in memory, so return true if we've seen a
1362 // sequence start and a frame start (not checking entry points since
1363 // they only occur in advanced profiles).
1364 return (sequence_start_code > 0 && frame_start_code > 0);
1367 // Now verify the block. AdvanceToStartCode() made sure that there are
1368 // at least 5 bytes remaining in the buffer.
1369 BitReader reader(buffer + offset, 5);
1370 RCHECK(ReadBits(&reader, 24) == 1);
1372 // Keep track of the number of certain types received.
1373 switch (ReadBits(&reader, 8)) {
1374 case VC1_SEQUENCE_START_CODE: {
1375 ++sequence_start_code;
1376 switch (ReadBits(&reader, 2)) {
1377 case 0: // simple
1378 case 1: // main
1379 RCHECK(ReadBits(&reader, 2) == 0);
1380 break;
1381 case 2: // complex
1382 return false;
1383 case 3: // advanced
1384 RCHECK(ReadBits(&reader, 3) <= 4); // Verify level = 0..4
1385 RCHECK(ReadBits(&reader, 2) == 1); // Verify colordiff_format = 1
1386 break;
1388 break;
1391 case VC1_ENTRY_POINT_START_CODE:
1392 // No fields in entry data to check. However, it must occur after
1393 // sequence header.
1394 RCHECK(sequence_start_code > 0);
1395 break;
1397 case VC1_FRAME_START_CODE:
1398 ++frame_start_code;
1399 break;
1401 offset += 5;
1405 // For some formats the signature is a bunch of characters. They are defined
1406 // below. Note that the first 4 characters of the string may be used as a TAG
1407 // in LookupContainerByFirst4. For signatures that contain embedded \0, use
1408 // uint8[].
1409 static const char kAmrSignature[] = "#!AMR";
1410 static const uint8 kAsfSignature[] = { 0x30, 0x26, 0xb2, 0x75, 0x8e, 0x66, 0xcf,
1411 0x11, 0xa6, 0xd9, 0x00, 0xaa, 0x00, 0x62,
1412 0xce, 0x6c };
1413 static const char kAssSignature[] = "[Script Info]";
1414 static const char kAssBomSignature[] = UTF8_BYTE_ORDER_MARK "[Script Info]";
1415 static const uint8 kWtvSignature[] = { 0xb7, 0xd8, 0x00, 0x20, 0x37, 0x49, 0xda,
1416 0x11, 0xa6, 0x4e, 0x00, 0x07, 0xe9, 0x5e,
1417 0xad, 0x8d };
1419 // Attempt to determine the container type from the buffer provided. This is
1420 // a simple pass, that uses the first 4 bytes of the buffer as an index to get
1421 // a rough idea of the container format.
1422 static MediaContainerName LookupContainerByFirst4(const uint8* buffer,
1423 int buffer_size) {
1424 // Minimum size that the code expects to exist without checking size.
1425 if (buffer_size < 12)
1426 return CONTAINER_UNKNOWN;
1428 uint32 first4 = Read32(buffer);
1429 switch (first4) {
1430 case 0x1a45dfa3:
1431 if (CheckWebm(buffer, buffer_size))
1432 return CONTAINER_WEBM;
1433 break;
1435 case 0x3026b275:
1436 if (StartsWith(buffer,
1437 buffer_size,
1438 kAsfSignature,
1439 sizeof(kAsfSignature))) {
1440 return CONTAINER_ASF;
1442 break;
1444 case TAG('#','!','A','M'):
1445 if (StartsWith(buffer, buffer_size, kAmrSignature))
1446 return CONTAINER_AMR;
1447 break;
1449 case TAG('#','E','X','T'):
1450 if (CheckHls(buffer, buffer_size))
1451 return CONTAINER_HLS;
1452 break;
1454 case TAG('.','R','M','F'):
1455 if (buffer[4] == 0 && buffer[5] == 0)
1456 return CONTAINER_RM;
1457 break;
1459 case TAG('.','r','a','\xfd'):
1460 return CONTAINER_RM;
1462 case TAG('B','I','K','b'):
1463 case TAG('B','I','K','d'):
1464 case TAG('B','I','K','f'):
1465 case TAG('B','I','K','g'):
1466 case TAG('B','I','K','h'):
1467 case TAG('B','I','K','i'):
1468 if (CheckBink(buffer, buffer_size))
1469 return CONTAINER_BINK;
1470 break;
1472 case TAG('c','a','f','f'):
1473 if (CheckCaf(buffer, buffer_size))
1474 return CONTAINER_CAF;
1475 break;
1477 case TAG('D','E','X','A'):
1478 if (buffer_size > 15 &&
1479 Read16(buffer + 11) <= 2048 &&
1480 Read16(buffer + 13) <= 2048) {
1481 return CONTAINER_DXA;
1483 break;
1485 case TAG('D','T','S','H'):
1486 if (Read32(buffer + 4) == TAG('D','H','D','R'))
1487 return CONTAINER_DTSHD;
1488 break;
1490 case 0x64a30100:
1491 case 0x64a30200:
1492 case 0x64a30300:
1493 case 0x64a30400:
1494 case 0x0001a364:
1495 case 0x0002a364:
1496 case 0x0003a364:
1497 if (Read32(buffer + 4) != 0 && Read32(buffer + 8) != 0)
1498 return CONTAINER_IRCAM;
1499 break;
1501 case TAG('f','L','a','C'):
1502 return CONTAINER_FLAC;
1504 case TAG('F','L','V',0):
1505 case TAG('F','L','V',1):
1506 case TAG('F','L','V',2):
1507 case TAG('F','L','V',3):
1508 case TAG('F','L','V',4):
1509 if (buffer[5] == 0 && Read32(buffer + 5) > 8)
1510 return CONTAINER_FLV;
1511 break;
1513 case TAG('F','O','R','M'):
1514 switch (Read32(buffer + 8)) {
1515 case TAG('A','I','F','F'):
1516 case TAG('A','I','F','C'):
1517 return CONTAINER_AIFF;
1519 break;
1521 case TAG('M','A','C',' '):
1522 return CONTAINER_APE;
1524 case TAG('O','N','2',' '):
1525 if (Read32(buffer + 8) == TAG('O','N','2','f'))
1526 return CONTAINER_AVI;
1527 break;
1529 case TAG('O','g','g','S'):
1530 if (buffer[5] <= 7)
1531 return CONTAINER_OGG;
1532 break;
1534 case TAG('R','F','6','4'):
1535 if (buffer_size > 16 && Read32(buffer + 12) == TAG('d','s','6','4'))
1536 return CONTAINER_WAV;
1537 break;
1539 case TAG('R','I','F','F'):
1540 switch (Read32(buffer + 8)) {
1541 case TAG('A','V','I',' '):
1542 case TAG('A','V','I','X'):
1543 case TAG('A','V','I','\x19'):
1544 case TAG('A','M','V',' '):
1545 return CONTAINER_AVI;
1546 case TAG('W','A','V','E'):
1547 return CONTAINER_WAV;
1549 break;
1551 case TAG('[','S','c','r'):
1552 if (StartsWith(buffer, buffer_size, kAssSignature))
1553 return CONTAINER_ASS;
1554 break;
1556 case TAG('\xef','\xbb','\xbf','['):
1557 if (StartsWith(buffer, buffer_size, kAssBomSignature))
1558 return CONTAINER_ASS;
1559 break;
1561 case 0x7ffe8001:
1562 case 0xfe7f0180:
1563 case 0x1fffe800:
1564 case 0xff1f00e8:
1565 if (CheckDts(buffer, buffer_size))
1566 return CONTAINER_DTS;
1567 break;
1569 case 0xb7d80020:
1570 if (StartsWith(buffer,
1571 buffer_size,
1572 kWtvSignature,
1573 sizeof(kWtvSignature))) {
1574 return CONTAINER_WTV;
1576 break;
1579 // Now try a few different ones that look at something other
1580 // than the first 4 bytes.
1581 uint32 first3 = first4 & 0xffffff00;
1582 switch (first3) {
1583 case TAG('C','W','S',0):
1584 case TAG('F','W','S',0):
1585 return CONTAINER_SWF;
1587 case TAG('I','D','3',0):
1588 if (CheckMp3(buffer, buffer_size, true))
1589 return CONTAINER_MP3;
1590 break;
1593 // Maybe the first 2 characters are something we can use.
1594 uint32 first2 = Read16(buffer);
1595 switch (first2) {
1596 case kAc3SyncWord:
1597 if (CheckAc3(buffer, buffer_size))
1598 return CONTAINER_AC3;
1599 if (CheckEac3(buffer, buffer_size))
1600 return CONTAINER_EAC3;
1601 break;
1603 case 0xfff0:
1604 case 0xfff1:
1605 case 0xfff8:
1606 case 0xfff9:
1607 if (CheckAac(buffer, buffer_size))
1608 return CONTAINER_AAC;
1609 break;
1612 // Check if the file is in MP3 format without the header.
1613 if (CheckMp3(buffer, buffer_size, false))
1614 return CONTAINER_MP3;
1616 return CONTAINER_UNKNOWN;
1619 // Attempt to determine the container name from the buffer provided.
1620 MediaContainerName DetermineContainer(const uint8* buffer, int buffer_size) {
1621 DCHECK(buffer);
1623 // Since MOV/QuickTime/MPEG4 streams are common, check for them first.
1624 if (CheckMov(buffer, buffer_size))
1625 return CONTAINER_MOV;
1627 // Next attempt the simple checks, that typically look at just the
1628 // first few bytes of the file.
1629 MediaContainerName result = LookupContainerByFirst4(buffer, buffer_size);
1630 if (result != CONTAINER_UNKNOWN)
1631 return result;
1633 // Additional checks that may scan a portion of the buffer.
1634 if (CheckMpeg2ProgramStream(buffer, buffer_size))
1635 return CONTAINER_MPEG2PS;
1636 if (CheckMpeg2TransportStream(buffer, buffer_size))
1637 return CONTAINER_MPEG2TS;
1638 if (CheckMJpeg(buffer, buffer_size))
1639 return CONTAINER_MJPEG;
1640 if (CheckDV(buffer, buffer_size))
1641 return CONTAINER_DV;
1642 if (CheckH261(buffer, buffer_size))
1643 return CONTAINER_H261;
1644 if (CheckH263(buffer, buffer_size))
1645 return CONTAINER_H263;
1646 if (CheckH264(buffer, buffer_size))
1647 return CONTAINER_H264;
1648 if (CheckMpeg4BitStream(buffer, buffer_size))
1649 return CONTAINER_MPEG4BS;
1650 if (CheckVC1(buffer, buffer_size))
1651 return CONTAINER_VC1;
1652 if (CheckSrt(buffer, buffer_size))
1653 return CONTAINER_SRT;
1654 if (CheckGsm(buffer, buffer_size))
1655 return CONTAINER_GSM;
1657 // AC3/EAC3 might not start at the beginning of the stream,
1658 // so scan for a start code.
1659 int offset = 1; // No need to start at byte 0 due to First4 check.
1660 if (AdvanceToStartCode(buffer, buffer_size, &offset, 4, 16, kAc3SyncWord)) {
1661 if (CheckAc3(buffer + offset, buffer_size - offset))
1662 return CONTAINER_AC3;
1663 if (CheckEac3(buffer + offset, buffer_size - offset))
1664 return CONTAINER_EAC3;
1667 return CONTAINER_UNKNOWN;
1670 } // namespace container_names
1672 } // namespace media