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