libavformat/isom.h

   1 /*
   2  * ISO Media common code
   3  * copyright (c) 2001 Fabrice Bellard
   4  * copyright (c) 2002 Francois Revol <revol@free.fr>
   5  * copyright (c) 2006 Baptiste Coudurier <baptiste.coudurier@free.fr>
   6  *
   7  * This file is part of FFmpeg.
   8  *
   9  * FFmpeg is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public
  11  * License as published by the Free Software Foundation; either
  12  * version 2.1 of the License, or (at your option) any later version.
  13  *
  14  * FFmpeg is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with FFmpeg; if not, write to the Free Software
  21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  22  */
  23
  24 #ifndef AVFORMAT_ISOM_H
  25 #define AVFORMAT_ISOM_H
  26
  27 #include <stddef.h>
  28 #include <stdint.h>
  29
  30 #include "libavutil/encryption_info.h"
  31 #include "libavutil/mastering_display_metadata.h"
  32 #include "libavutil/ambient_viewing_environment.h"
  33 #include "libavutil/spherical.h"
  34 #include "libavutil/stereo3d.h"
  35
  36 #include "avio.h"
  37 #include "internal.h"
  38 #include "dv.h"
  39
  40 /* isom.c */
  41 extern const AVCodecTag ff_mp4_obj_type[];
  42 extern const AVCodecTag ff_codec_movvideo_tags[];
  43 extern const AVCodecTag ff_codec_movaudio_tags[];
  44 extern const AVCodecTag ff_codec_movsubtitle_tags[];
  45 extern const AVCodecTag ff_codec_movdata_tags[];
  46
  47 int ff_mov_iso639_to_lang(const char lang[4], int mp4);
  48 int ff_mov_lang_to_iso639(unsigned code, char to[4]);
  49
  50 struct AVAESCTR;
  51
  52 /* the QuickTime file format is quite convoluted...
  53  * it has lots of index tables, each indexing something in another one...
  54  * Here we just use what is needed to read the chunks
  55  */
  56
  57 typedef struct MOVTimeToSample {
  58     unsigned int count;
  59     unsigned int duration;
  60     int offset;
  61 } MOVTimeToSample;
  62
  63 typedef struct MOVStts {
  64     unsigned int count;
  65     unsigned int duration;
  66 } MOVStts;
  67
  68 typedef struct MOVCtts {
  69     unsigned int count;
  70     int offset;
  71 } MOVCtts;
  72
  73 typedef struct MOVStsc {
  74     int first;
  75     int count;
  76     int id;
  77 } MOVStsc;
  78
  79 typedef struct MOVElst {
  80     int64_t duration;
  81     int64_t time;
  82     float rate;
  83 } MOVElst;
  84
  85 typedef struct MOVDref {
  86     uint32_t type;
  87     char *path;
  88     char *dir;
  89     char volume[28];
  90     char filename[64];
  91     int16_t nlvl_to, nlvl_from;
  92 } MOVDref;
  93
  94 typedef struct MOVAtom {
  95     uint32_t type;
  96     int64_t size; /* total size (excluding the size and type fields) */
  97 } MOVAtom;
  98
  99 struct MOVParseTableEntry;
 100
 101 typedef struct MOVFragment {
 102     int found_tfhd;
 103     unsigned track_id;
 104     uint64_t base_data_offset;
 105     uint64_t moof_offset;
 106     uint64_t implicit_offset;
 107     unsigned stsd_id;
 108     unsigned duration;
 109     unsigned size;
 110     unsigned flags;
 111 } MOVFragment;
 112
 113 typedef struct MOVTrackExt {
 114     unsigned track_id;
 115     unsigned stsd_id;
 116     unsigned duration;
 117     unsigned size;
 118     unsigned flags;
 119 } MOVTrackExt;
 120
 121 typedef struct MOVSbgp {
 122     unsigned int count;
 123     unsigned int index;
 124 } MOVSbgp;
 125
 126 typedef struct MOVEncryptionIndex {
 127     // Individual encrypted samples.  If there are no elements, then the default
 128     // settings will be used.
 129     unsigned int nb_encrypted_samples;
 130     AVEncryptionInfo **encrypted_samples;
 131
 132     uint8_t* auxiliary_info_sizes;
 133     size_t auxiliary_info_sample_count;
 134     uint8_t auxiliary_info_default_size;
 135     uint64_t* auxiliary_offsets;  ///< Absolute seek position
 136     size_t auxiliary_offsets_count;
 137 } MOVEncryptionIndex;
 138
 139 typedef struct MOVFragmentStreamInfo {
 140     int id;
 141     int64_t sidx_pts;
 142     int64_t first_tfra_pts;
 143     int64_t tfdt_dts;
 144     int64_t next_trun_dts;
 145     // Index of the first sample/trun in the fragment.
 146     int index_base;
 147     int index_entry;
 148     MOVEncryptionIndex *encryption_index;
 149     int stsd_id; // current fragment stsd_id
 150 } MOVFragmentStreamInfo;
 151
 152 typedef struct MOVFragmentIndexItem {
 153     int64_t moof_offset;
 154     int headers_read;
 155     int current;
 156     int nb_stream_info;
 157     MOVFragmentStreamInfo * stream_info;
 158 } MOVFragmentIndexItem;
 159
 160 typedef struct MOVFragmentIndex {
 161     int allocated_size;
 162     int complete;
 163     int current;
 164     int nb_items;
 165     MOVFragmentIndexItem * item;
 166 } MOVFragmentIndex;
 167
 168 typedef struct MOVIndexRange {
 169     int64_t start;
 170     int64_t end;
 171 } MOVIndexRange;
 172
 173 typedef struct MOVStreamContext {
 174     AVIOContext *pb;
 175     int refcount;
 176     int pb_is_copied;
 177     int id;               ///< AVStream id
 178     int ffindex;          ///< AVStream index
 179     int next_chunk;
 180     unsigned int chunk_count;
 181     int64_t *chunk_offsets;
 182     unsigned int tts_count;
 183     unsigned int tts_allocated_size;
 184     MOVTimeToSample *tts_data;
 185     unsigned int stts_count;
 186     unsigned int stts_allocated_size;
 187     MOVStts *stts_data;
 188     unsigned int sdtp_count;
 189     uint8_t *sdtp_data;
 190     unsigned int ctts_count;
 191     unsigned int ctts_allocated_size;
 192     MOVCtts *ctts_data;
 193     unsigned int stsc_count;
 194     MOVStsc *stsc_data;
 195     unsigned int stsc_index;
 196     int stsc_sample;
 197     unsigned int stps_count;
 198     unsigned *stps_data;  ///< partial sync sample for mpeg-2 open gop
 199     MOVElst *elst_data;
 200     unsigned int elst_count;
 201     int tts_index;
 202     int tts_sample;
 203     unsigned int sample_size; ///< may contain value calculated from stsd or value from stsz atom
 204     unsigned int stsz_sample_size; ///< always contains sample size from stsz atom
 205     unsigned int sample_count;
 206     unsigned int *sample_sizes;
 207     int keyframe_absent;
 208     unsigned int keyframe_count;
 209     int *keyframes;
 210     int time_scale;
 211     int64_t time_offset;  ///< time offset of the edit list entries
 212     int64_t min_corrected_pts;  ///< minimum Composition time shown by the edits excluding empty edits.
 213     int current_sample;
 214     int64_t current_index;
 215     MOVIndexRange* index_ranges;
 216     MOVIndexRange* current_index_range;
 217     unsigned int bytes_per_frame;
 218     unsigned int samples_per_frame;
 219     int dv_audio_container;
 220     int pseudo_stream_id; ///< -1 means demux all ids
 221     int16_t audio_cid;    ///< stsd audio compression id
 222     unsigned drefs_count;
 223     MOVDref *drefs;
 224     int dref_id;
 225     unsigned tref_flags;
 226     int tref_id;
 227     int timecode_track;
 228     int width;            ///< tkhd width
 229     int height;           ///< tkhd height
 230     int h_spacing;        ///< pasp hSpacing
 231     int v_spacing;        ///< pasp vSpacing
 232     int dts_shift;        ///< dts shift when ctts is negative
 233     uint32_t palette[256];
 234     int has_palette;
 235     int64_t data_size;
 236     uint32_t tmcd_flags;  ///< tmcd track flags
 237     uint8_t tmcd_nb_frames;  ///< tmcd number of frames per tick / second
 238     int64_t track_end;    ///< used for dts generation in fragmented movie files
 239     int start_pad;        ///< amount of samples to skip due to enc-dec delay
 240     unsigned int rap_group_count;
 241     MOVSbgp *rap_group;
 242     unsigned int sync_group_count;
 243     MOVSbgp *sync_group;
 244     uint8_t *sgpd_sync;
 245     uint32_t sgpd_sync_count;
 246     int32_t *sample_offsets;
 247     int sample_offsets_count;
 248     int *open_key_samples;
 249     int open_key_samples_count;
 250     uint32_t min_sample_duration;
 251
 252     int nb_frames_for_fps;
 253     int64_t duration_for_fps;
 254
 255     /** extradata array (and size) for multiple stsd */
 256     uint8_t **extradata;
 257     int *extradata_size;
 258     int last_stsd_index;
 259     int stsd_count;
 260     int stsd_version;
 261
 262     int32_t *display_matrix;
 263     AVStereo3D *stereo3d;
 264     size_t stereo3d_size;
 265     AVSphericalMapping *spherical;
 266     size_t spherical_size;
 267     AVMasteringDisplayMetadata *mastering;
 268     size_t mastering_size;
 269     AVContentLightMetadata *coll;
 270     size_t coll_size;
 271     AVAmbientViewingEnvironment *ambient;
 272     size_t ambient_size;
 273
 274     uint32_t format;
 275
 276     int has_sidx;  // If there is an sidx entry for this stream.
 277     struct {
 278         struct AVAESCTR* aes_ctr;
 279         struct AVAES *aes_ctx;
 280         unsigned int per_sample_iv_size;  // Either 0, 8, or 16.
 281         AVEncryptionInfo *default_encrypted_sample;
 282         MOVEncryptionIndex *encryption_index;
 283     } cenc;
 284
 285     struct IAMFDemuxContext *iamf;
 286     int iamf_stream_offset;
 287 } MOVStreamContext;
 288
 289 typedef struct HEIFItemRef {
 290     uint32_t type;
 291     int item_id;
 292 } HEIFItemRef;
 293
 294 typedef struct HEIFItem {
 295     AVStream *st;
 296     HEIFItemRef *iref_list;
 297     int nb_iref_list;
 298     char *name;
 299     int item_id;
 300     int64_t extent_length;
 301     int64_t extent_offset;
 302     int width;
 303     int height;
 304     int rotation;
 305     int hflip;
 306     int vflip;
 307     int type;
 308     int is_idat_relative;
 309     uint8_t *icc_profile;
 310     size_t icc_profile_size;
 311 } HEIFItem;
 312
 313 typedef struct HEIFGrid {
 314     HEIFItem *item;
 315     HEIFItem **tile_item_list;
 316     int16_t *tile_id_list;
 317     unsigned *tile_idx_list;
 318     int nb_tiles;
 319 } HEIFGrid;
 320
 321 typedef struct MOVContext {
 322     const AVClass *class; ///< class for private options
 323     AVFormatContext *fc;
 324     int time_scale;
 325     int64_t duration;     ///< duration of the longest track
 326     int found_moov;       ///< 'moov' atom has been found
 327     int found_iloc;       ///< 'iloc' atom has been found
 328     int found_iinf;       ///< 'iinf' atom has been found
 329     int found_mdat;       ///< 'mdat' atom has been found
 330     int found_hdlr_mdta;  ///< 'hdlr' atom with type 'mdta' has been found
 331     int trak_index;       ///< Index of the current 'trak'
 332     char **meta_keys;
 333     unsigned meta_keys_count;
 334     DVDemuxContext *dv_demux;
 335     AVFormatContext *dv_fctx;
 336     int isom;             ///< 1 if file is ISO Media (mp4/3gp)
 337     MOVFragment fragment; ///< current fragment in moof atom
 338     MOVTrackExt *trex_data;
 339     unsigned trex_count;
 340     int itunes_metadata;  ///< metadata are itunes style
 341     int handbrake_version;
 342     int *chapter_tracks;
 343     unsigned int nb_chapter_tracks;
 344     int use_absolute_path;
 345     int ignore_editlist;
 346     int advanced_editlist;
 347     int advanced_editlist_autodisabled;
 348     int ignore_chapters;
 349     int seek_individually;
 350     int64_t next_root_atom; ///< offset of the next root atom
 351     int export_all;
 352     int export_xmp;
 353     int *bitrates;          ///< bitrates read before streams creation
 354     int bitrates_count;
 355     int moov_retry;
 356     int use_mfra_for;
 357     int has_looked_for_mfra;
 358     int use_tfdt;
 359     MOVFragmentIndex frag_index;
 360     int atom_depth;
 361     unsigned int aax_mode;  ///< 'aax' file has been detected
 362     uint8_t file_key[20];
 363     uint8_t file_iv[20];
 364     void *activation_bytes;
 365     int activation_bytes_size;
 366     void *audible_fixed_key;
 367     int audible_fixed_key_size;
 368     void *audible_key;
 369     int audible_key_size;
 370     void *audible_iv;
 371     int audible_iv_size;
 372     struct AVAES *aes_decrypt;
 373     uint8_t *decryption_key;
 374     int decryption_key_len;
 375     int enable_drefs;
 376     int32_t movie_display_matrix[3][3]; ///< display matrix from mvhd
 377     int have_read_mfra_size;
 378     uint32_t mfra_size;
 379     uint32_t max_stts_delta;
 380     int primary_item_id;
 381     int cur_item_id;
 382     HEIFItem **heif_item;
 383     int nb_heif_item;
 384     HEIFGrid *heif_grid;
 385     int nb_heif_grid;
 386     int64_t idat_offset;
 387     int interleaved_read;
 388 } MOVContext;
 389
 390 int ff_mp4_read_descr_len(AVIOContext *pb);
 391 int ff_mp4_read_descr(void *logctx, AVIOContext *pb, int *tag);
 392 int ff_mp4_read_dec_config_descr(void *logctx, AVStream *st, AVIOContext *pb);
 393 void ff_mp4_parse_es_descr(AVIOContext *pb, int *es_id);
 394
 395 #define MP4ODescrTag                    0x01
 396 #define MP4IODescrTag                   0x02
 397 #define MP4ESDescrTag                   0x03
 398 #define MP4DecConfigDescrTag            0x04
 399 #define MP4DecSpecificDescrTag          0x05
 400 #define MP4SLDescrTag                   0x06
 401
 402 #define MOV_TFHD_BASE_DATA_OFFSET       0x01
 403 #define MOV_TFHD_STSD_ID                0x02
 404 #define MOV_TFHD_DEFAULT_DURATION       0x08
 405 #define MOV_TFHD_DEFAULT_SIZE           0x10
 406 #define MOV_TFHD_DEFAULT_FLAGS          0x20
 407 #define MOV_TFHD_DURATION_IS_EMPTY  0x010000
 408 #define MOV_TFHD_DEFAULT_BASE_IS_MOOF 0x020000
 409
 410 #define MOV_TRUN_DATA_OFFSET            0x01
 411 #define MOV_TRUN_FIRST_SAMPLE_FLAGS     0x04
 412 #define MOV_TRUN_SAMPLE_DURATION       0x100
 413 #define MOV_TRUN_SAMPLE_SIZE           0x200
 414 #define MOV_TRUN_SAMPLE_FLAGS          0x400
 415 #define MOV_TRUN_SAMPLE_CTS            0x800
 416
 417 #define MOV_FRAG_SAMPLE_FLAG_DEGRADATION_PRIORITY_MASK 0x0000ffff
 418 #define MOV_FRAG_SAMPLE_FLAG_IS_NON_SYNC               0x00010000
 419 #define MOV_FRAG_SAMPLE_FLAG_PADDING_MASK              0x000e0000
 420 #define MOV_FRAG_SAMPLE_FLAG_REDUNDANCY_MASK           0x00300000
 421 #define MOV_FRAG_SAMPLE_FLAG_DEPENDED_MASK             0x00c00000
 422 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_MASK              0x03000000
 423
 424 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_NO                0x02000000
 425 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_YES               0x01000000
 426
 427 #define MOV_TKHD_FLAG_ENABLED       0x0001
 428 #define MOV_TKHD_FLAG_IN_MOVIE      0x0002
 429 #define MOV_TKHD_FLAG_IN_PREVIEW    0x0004
 430 #define MOV_TKHD_FLAG_IN_POSTER     0x0008
 431
 432 #define MOV_SAMPLE_DEPENDENCY_UNKNOWN 0x0
 433 #define MOV_SAMPLE_DEPENDENCY_YES     0x1
 434 #define MOV_SAMPLE_DEPENDENCY_NO      0x2
 435
 436 #define MOV_TREF_FLAG_ENHANCEMENT     0x1
 437
 438 #define TAG_IS_AVCI(tag)                    \
 439     ((tag) == MKTAG('a', 'i', '5', 'p') ||  \
 440      (tag) == MKTAG('a', 'i', '5', 'q') ||  \
 441      (tag) == MKTAG('a', 'i', '5', '2') ||  \
 442      (tag) == MKTAG('a', 'i', '5', '3') ||  \
 443      (tag) == MKTAG('a', 'i', '5', '5') ||  \
 444      (tag) == MKTAG('a', 'i', '5', '6') ||  \
 445      (tag) == MKTAG('a', 'i', '1', 'p') ||  \
 446      (tag) == MKTAG('a', 'i', '1', 'q') ||  \
 447      (tag) == MKTAG('a', 'i', '1', '2') ||  \
 448      (tag) == MKTAG('a', 'i', '1', '3') ||  \
 449      (tag) == MKTAG('a', 'i', '1', '5') ||  \
 450      (tag) == MKTAG('a', 'i', '1', '6') ||  \
 451      (tag) == MKTAG('a', 'i', 'v', 'x') ||  \
 452      (tag) == MKTAG('A', 'V', 'i', 'n'))
 453
 454
 455 int ff_mov_read_esds(AVFormatContext *fc, AVIOContext *pb);
 456
 457 int ff_mov_read_stsd_entries(MOVContext *c, AVIOContext *pb, int entries);
 458 void ff_mov_write_chan(AVIOContext *pb, int64_t channel_layout);
 459
 460 #define FF_MOV_FLAG_MFRA_AUTO -1
 461 #define FF_MOV_FLAG_MFRA_DTS 1
 462 #define FF_MOV_FLAG_MFRA_PTS 2
 463
 464 /**
 465  * Compute codec id for 'lpcm' tag.
 466  * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
 467  */
 468 static inline enum AVCodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
 469 {
 470     /* lpcm flags:
 471      * 0x1 = float
 472      * 0x2 = big-endian
 473      * 0x4 = signed
 474      */
 475     return ff_get_pcm_codec_id(bps, flags & 1, flags & 2, flags & 4 ? -1 : 0);
 476 }
 477
 478 #define MOV_ISMV_TTML_TAG MKTAG('d', 'f', 'x', 'p')
 479 #define MOV_MP4_TTML_TAG  MKTAG('s', 't', 'p', 'p')
 480 #define MOV_MP4_FPCM_TAG  MKTAG('f', 'p', 'c', 'm')
 481 #define MOV_MP4_IPCM_TAG  MKTAG('i', 'p', 'c', 'm')
 482
 483 struct MP4TrackKindValueMapping {
 484     int         disposition;
 485     const char *value;
 486 };
 487
 488 struct MP4TrackKindMapping {
 489     const char   *scheme_uri;
 490     const struct  MP4TrackKindValueMapping *value_maps;
 491 };
 492
 493 extern const struct MP4TrackKindMapping ff_mov_track_kind_table[];
 494
 495 #endif /* AVFORMAT_ISOM_H */