81a71d39f4cb9ec56f34d24233a73d6f2f17dd41
[libav.git] / libavformat / flvdec.c
1 /*
2 * FLV demuxer
3 * Copyright (c) 2003 The Libav Project
4 *
5 * This demuxer will generate a 1 byte extradata for VP6F content.
6 * It is composed of:
7 * - upper 4 bits: difference between encoded width and visible width
8 * - lower 4 bits: difference between encoded height and visible height
9 *
10 * This file is part of Libav.
11 *
12 * Libav is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
15 * version 2.1 of the License, or (at your option) any later version.
16 *
17 * Libav is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
21 *
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with Libav; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 */
26
27 #include "libavutil/avstring.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/dict.h"
30 #include "libavutil/opt.h"
31 #include "libavutil/intfloat.h"
32 #include "libavutil/mathematics.h"
33 #include "libavcodec/bytestream.h"
34 #include "libavcodec/mpeg4audio.h"
35 #include "avformat.h"
36 #include "internal.h"
37 #include "avio_internal.h"
38 #include "flv.h"
39
40 #define KEYFRAMES_TAG "keyframes"
41 #define KEYFRAMES_TIMESTAMP_TAG "times"
42 #define KEYFRAMES_BYTEOFFSET_TAG "filepositions"
43
44 #define VALIDATE_INDEX_TS_THRESH 2500
45
46 typedef struct FLVContext {
47 const AVClass *class; ///< Class for private options.
48 int trust_metadata; ///< configure streams according onMetaData
49 int wrong_dts; ///< wrong dts due to negative cts
50 uint8_t *new_extradata[2];
51 int new_extradata_size[2];
52 int last_sample_rate;
53 int last_channels;
54 struct {
55 int64_t dts;
56 int64_t pos;
57 } validate_index[2];
58 int validate_next;
59 int validate_count;
60 int searched_for_end;
61 AVRational framerate;
62 } FLVContext;
63
64 static int flv_probe(AVProbeData *p)
65 {
66 const uint8_t *d;
67
68 d = p->buf;
69 if (d[0] == 'F' &&
70 d[1] == 'L' &&
71 d[2] == 'V' &&
72 d[3] < 5 && d[5] == 0 &&
73 AV_RB32(d + 5) > 8) {
74 return AVPROBE_SCORE_MAX;
75 }
76 return 0;
77 }
78
79 static AVStream *create_stream(AVFormatContext *s, int codec_type)
80 {
81 FLVContext *flv = s->priv_data;
82 AVStream *st = avformat_new_stream(s, NULL);
83 if (!st)
84 return NULL;
85 st->codecpar->codec_type = codec_type;
86 if (codec_type == AVMEDIA_TYPE_VIDEO)
87 st->avg_frame_rate = flv->framerate;
88
89 avpriv_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
90 return st;
91 }
92
93 static int flv_same_audio_codec(AVCodecParameters *apar, int flags)
94 {
95 int bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
96 int flv_codecid = flags & FLV_AUDIO_CODECID_MASK;
97 int codec_id;
98
99 if (!apar->codec_id && !apar->codec_tag)
100 return 1;
101
102 if (apar->bits_per_coded_sample != bits_per_coded_sample)
103 return 0;
104
105 switch (flv_codecid) {
106 // no distinction between S16 and S8 PCM codec flags
107 case FLV_CODECID_PCM:
108 codec_id = bits_per_coded_sample == 8
109 ? AV_CODEC_ID_PCM_U8
110 #if HAVE_BIGENDIAN
111 : AV_CODEC_ID_PCM_S16BE;
112 #else
113 : AV_CODEC_ID_PCM_S16LE;
114 #endif
115 return codec_id == apar->codec_id;
116 case FLV_CODECID_PCM_LE:
117 codec_id = bits_per_coded_sample == 8
118 ? AV_CODEC_ID_PCM_U8
119 : AV_CODEC_ID_PCM_S16LE;
120 return codec_id == apar->codec_id;
121 case FLV_CODECID_AAC:
122 return apar->codec_id == AV_CODEC_ID_AAC;
123 case FLV_CODECID_ADPCM:
124 return apar->codec_id == AV_CODEC_ID_ADPCM_SWF;
125 case FLV_CODECID_SPEEX:
126 return apar->codec_id == AV_CODEC_ID_SPEEX;
127 case FLV_CODECID_MP3:
128 return apar->codec_id == AV_CODEC_ID_MP3;
129 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
130 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
131 case FLV_CODECID_NELLYMOSER:
132 return apar->codec_id == AV_CODEC_ID_NELLYMOSER;
133 case FLV_CODECID_PCM_MULAW:
134 return apar->sample_rate == 8000 &&
135 apar->codec_id == AV_CODEC_ID_PCM_MULAW;
136 case FLV_CODECID_PCM_ALAW:
137 return apar->sample_rate == 8000 &&
138 apar->codec_id == AV_CODEC_ID_PCM_ALAW;
139 default:
140 return apar->codec_tag == (flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
141 }
142 }
143
144 static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream,
145 AVCodecParameters *apar, int flv_codecid)
146 {
147 switch (flv_codecid) {
148 // no distinction between S16 and S8 PCM codec flags
149 case FLV_CODECID_PCM:
150 apar->codec_id = apar->bits_per_coded_sample == 8
151 ? AV_CODEC_ID_PCM_U8
152 #if HAVE_BIGENDIAN
153 : AV_CODEC_ID_PCM_S16BE;
154 #else
155 : AV_CODEC_ID_PCM_S16LE;
156 #endif
157 break;
158 case FLV_CODECID_PCM_LE:
159 apar->codec_id = apar->bits_per_coded_sample == 8
160 ? AV_CODEC_ID_PCM_U8
161 : AV_CODEC_ID_PCM_S16LE;
162 break;
163 case FLV_CODECID_AAC:
164 apar->codec_id = AV_CODEC_ID_AAC;
165 break;
166 case FLV_CODECID_ADPCM:
167 apar->codec_id = AV_CODEC_ID_ADPCM_SWF;
168 break;
169 case FLV_CODECID_SPEEX:
170 apar->codec_id = AV_CODEC_ID_SPEEX;
171 apar->sample_rate = 16000;
172 break;
173 case FLV_CODECID_MP3:
174 apar->codec_id = AV_CODEC_ID_MP3;
175 astream->need_parsing = AVSTREAM_PARSE_FULL;
176 break;
177 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
178 // in case metadata does not otherwise declare samplerate
179 apar->sample_rate = 8000;
180 apar->codec_id = AV_CODEC_ID_NELLYMOSER;
181 break;
182 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
183 apar->sample_rate = 16000;
184 apar->codec_id = AV_CODEC_ID_NELLYMOSER;
185 break;
186 case FLV_CODECID_NELLYMOSER:
187 apar->codec_id = AV_CODEC_ID_NELLYMOSER;
188 break;
189 case FLV_CODECID_PCM_MULAW:
190 apar->sample_rate = 8000;
191 apar->codec_id = AV_CODEC_ID_PCM_MULAW;
192 break;
193 case FLV_CODECID_PCM_ALAW:
194 apar->sample_rate = 8000;
195 apar->codec_id = AV_CODEC_ID_PCM_ALAW;
196 break;
197 default:
198 av_log(s, AV_LOG_INFO, "Unsupported audio codec (%x)\n",
199 flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
200 apar->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
201 }
202 }
203
204 static int flv_same_video_codec(AVCodecParameters *vpar, int flags)
205 {
206 int flv_codecid = flags & FLV_VIDEO_CODECID_MASK;
207
208 if (!vpar->codec_id && !vpar->codec_tag)
209 return 1;
210
211 switch (flv_codecid) {
212 case FLV_CODECID_H263:
213 return vpar->codec_id == AV_CODEC_ID_FLV1;
214 case FLV_CODECID_SCREEN:
215 return vpar->codec_id == AV_CODEC_ID_FLASHSV;
216 case FLV_CODECID_SCREEN2:
217 return vpar->codec_id == AV_CODEC_ID_FLASHSV2;
218 case FLV_CODECID_VP6:
219 return vpar->codec_id == AV_CODEC_ID_VP6F;
220 case FLV_CODECID_VP6A:
221 return vpar->codec_id == AV_CODEC_ID_VP6A;
222 case FLV_CODECID_H264:
223 return vpar->codec_id == AV_CODEC_ID_H264;
224 default:
225 return vpar->codec_tag == flv_codecid;
226 }
227 }
228
229 static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream,
230 int flv_codecid, int read)
231 {
232 AVCodecParameters *par = vstream->codecpar;
233 switch (flv_codecid) {
234 case FLV_CODECID_H263:
235 par->codec_id = AV_CODEC_ID_FLV1;
236 break;
237 case FLV_CODECID_SCREEN:
238 par->codec_id = AV_CODEC_ID_FLASHSV;
239 break;
240 case FLV_CODECID_SCREEN2:
241 par->codec_id = AV_CODEC_ID_FLASHSV2;
242 break;
243 case FLV_CODECID_VP6:
244 par->codec_id = AV_CODEC_ID_VP6F;
245 case FLV_CODECID_VP6A:
246 if (flv_codecid == FLV_CODECID_VP6A)
247 par->codec_id = AV_CODEC_ID_VP6A;
248 if (read) {
249 if (par->extradata_size != 1) {
250 par->extradata = av_malloc(1);
251 if (par->extradata)
252 par->extradata_size = 1;
253 }
254 if (par->extradata)
255 par->extradata[0] = avio_r8(s->pb);
256 else
257 avio_skip(s->pb, 1);
258 }
259 return 1; // 1 byte body size adjustment for flv_read_packet()
260 case FLV_CODECID_H264:
261 par->codec_id = AV_CODEC_ID_H264;
262 return 3; // not 4, reading packet type will consume one byte
263 default:
264 av_log(s, AV_LOG_INFO, "Unsupported video codec (%x)\n", flv_codecid);
265 par->codec_tag = flv_codecid;
266 }
267
268 return 0;
269 }
270
271 static int amf_get_string(AVIOContext *ioc, char *buffer, int buffsize)
272 {
273 int length = avio_rb16(ioc);
274 if (length >= buffsize) {
275 avio_skip(ioc, length);
276 return -1;
277 }
278
279 avio_read(ioc, buffer, length);
280
281 buffer[length] = '\0';
282
283 return length;
284 }
285
286 static int parse_keyframes_index(AVFormatContext *s, AVIOContext *ioc,
287 AVStream *vstream, int64_t max_pos)
288 {
289 FLVContext *flv = s->priv_data;
290 unsigned int arraylen = 0, timeslen = 0, fileposlen = 0, i;
291 double num_val;
292 char str_val[256];
293 int64_t *times = NULL;
294 int64_t *filepositions = NULL;
295 int ret = AVERROR(ENOSYS);
296 int64_t initial_pos = avio_tell(ioc);
297
298 if (s->flags & AVFMT_FLAG_IGNIDX)
299 return 0;
300
301 while (avio_tell(ioc) < max_pos - 2 &&
302 amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
303 int64_t *current_array;
304
305 // Expect array object in context
306 if (avio_r8(ioc) != AMF_DATA_TYPE_ARRAY)
307 break;
308
309 arraylen = avio_rb32(ioc);
310 if (arraylen >> 28)
311 break;
312
313 /* Expect only 'times' or 'filepositions' sub-arrays in other
314 * case refuse to use such metadata for indexing. */
315 if (!strcmp(KEYFRAMES_TIMESTAMP_TAG, str_val) && !times) {
316 if (!(times = av_mallocz(sizeof(*times) * arraylen))) {
317 ret = AVERROR(ENOMEM);
318 goto finish;
319 }
320 timeslen = arraylen;
321 current_array = times;
322 } else if (!strcmp(KEYFRAMES_BYTEOFFSET_TAG, str_val) &&
323 !filepositions) {
324 if (!(filepositions = av_mallocz(sizeof(*filepositions) * arraylen))) {
325 ret = AVERROR(ENOMEM);
326 goto finish;
327 }
328 fileposlen = arraylen;
329 current_array = filepositions;
330 } else
331 // unexpected metatag inside keyframes, will not use such
332 // metadata for indexing
333 break;
334
335 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
336 if (avio_r8(ioc) != AMF_DATA_TYPE_NUMBER)
337 goto finish;
338 num_val = av_int2double(avio_rb64(ioc));
339 current_array[i] = num_val;
340 }
341 if (times && filepositions) {
342 // All done, exiting at a position allowing amf_parse_object
343 // to finish parsing the object
344 ret = 0;
345 break;
346 }
347 }
348
349 if (!ret && timeslen == fileposlen) {
350 for (i = 0; i < fileposlen; i++) {
351 av_add_index_entry(vstream, filepositions[i], times[i] * 1000,
352 0, 0, AVINDEX_KEYFRAME);
353 if (i < 2) {
354 flv->validate_index[i].pos = filepositions[i];
355 flv->validate_index[i].dts = times[i] * 1000;
356 flv->validate_count = i + 1;
357 }
358 }
359 } else
360 av_log(s, AV_LOG_WARNING, "Invalid keyframes object, skipping.\n");
361
362 finish:
363 av_freep(&times);
364 av_freep(&filepositions);
365 // If we got unexpected data, but successfully reset back to
366 // the start pos, the caller can continue parsing
367 if (ret < 0 && avio_seek(ioc, initial_pos, SEEK_SET) > 0)
368 return 0;
369 return ret;
370 }
371
372 static int amf_parse_object(AVFormatContext *s, AVStream *astream,
373 AVStream *vstream, const char *key,
374 int64_t max_pos, int depth)
375 {
376 AVCodecParameters *apar, *vpar;
377 FLVContext *flv = s->priv_data;
378 AVIOContext *ioc;
379 AMFDataType amf_type;
380 char str_val[256];
381 double num_val;
382
383 num_val = 0;
384 ioc = s->pb;
385 amf_type = avio_r8(ioc);
386
387 switch (amf_type) {
388 case AMF_DATA_TYPE_NUMBER:
389 num_val = av_int2double(avio_rb64(ioc));
390 break;
391 case AMF_DATA_TYPE_BOOL:
392 num_val = avio_r8(ioc);
393 break;
394 case AMF_DATA_TYPE_STRING:
395 if (amf_get_string(ioc, str_val, sizeof(str_val)) < 0)
396 return -1;
397 break;
398 case AMF_DATA_TYPE_OBJECT:
399 if ((vstream || astream) && key &&
400 !strcmp(KEYFRAMES_TAG, key) && depth == 1)
401 if (parse_keyframes_index(s, ioc, vstream ? vstream : astream,
402 max_pos) < 0)
403 return -1;
404
405 while (avio_tell(ioc) < max_pos - 2 &&
406 amf_get_string(ioc, str_val, sizeof(str_val)) > 0)
407 if (amf_parse_object(s, astream, vstream, str_val, max_pos,
408 depth + 1) < 0)
409 return -1; // if we couldn't skip, bomb out.
410 if (avio_r8(ioc) != AMF_END_OF_OBJECT)
411 return -1;
412 break;
413 case AMF_DATA_TYPE_NULL:
414 case AMF_DATA_TYPE_UNDEFINED:
415 case AMF_DATA_TYPE_UNSUPPORTED:
416 break; // these take up no additional space
417 case AMF_DATA_TYPE_MIXEDARRAY:
418 avio_skip(ioc, 4); // skip 32-bit max array index
419 while (avio_tell(ioc) < max_pos - 2 &&
420 amf_get_string(ioc, str_val, sizeof(str_val)) > 0)
421 // this is the only case in which we would want a nested
422 // parse to not skip over the object
423 if (amf_parse_object(s, astream, vstream, str_val, max_pos,
424 depth + 1) < 0)
425 return -1;
426 if (avio_r8(ioc) != AMF_END_OF_OBJECT)
427 return -1;
428 break;
429 case AMF_DATA_TYPE_ARRAY:
430 {
431 unsigned int arraylen, i;
432
433 arraylen = avio_rb32(ioc);
434 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++)
435 if (amf_parse_object(s, NULL, NULL, NULL, max_pos,
436 depth + 1) < 0)
437 return -1; // if we couldn't skip, bomb out.
438 }
439 break;
440 case AMF_DATA_TYPE_DATE:
441 avio_skip(ioc, 8 + 2); // timestamp (double) and UTC offset (int16)
442 break;
443 default: // unsupported type, we couldn't skip
444 return -1;
445 }
446
447 if (key) {
448 // stream info doesn't live any deeper than the first object
449 if (depth == 1) {
450 apar = astream ? astream->codecpar : NULL;
451 vpar = vstream ? vstream->codecpar : NULL;
452
453 if (amf_type == AMF_DATA_TYPE_NUMBER ||
454 amf_type == AMF_DATA_TYPE_BOOL) {
455 if (!strcmp(key, "duration"))
456 s->duration = num_val * AV_TIME_BASE;
457 else if (!strcmp(key, "videodatarate") && vpar &&
458 0 <= (int)(num_val * 1024.0))
459 vpar->bit_rate = num_val * 1024.0;
460 else if (!strcmp(key, "audiodatarate") && apar &&
461 0 <= (int)(num_val * 1024.0))
462 apar->bit_rate = num_val * 1024.0;
463 else if (!strcmp(key, "datastream")) {
464 AVStream *st = create_stream(s, AVMEDIA_TYPE_DATA);
465 if (!st)
466 return AVERROR(ENOMEM);
467 st->codecpar->codec_id = AV_CODEC_ID_TEXT;
468 } else if (!strcmp(key, "framerate")) {
469 flv->framerate = av_d2q(num_val, 1000);
470 if (vstream)
471 vstream->avg_frame_rate = flv->framerate;
472 } else if (flv->trust_metadata) {
473 if (!strcmp(key, "videocodecid") && vpar) {
474 flv_set_video_codec(s, vstream, num_val, 0);
475 } else if (!strcmp(key, "audiocodecid") && apar) {
476 int id = ((int)num_val) << FLV_AUDIO_CODECID_OFFSET;
477 flv_set_audio_codec(s, astream, apar, id);
478 } else if (!strcmp(key, "audiosamplerate") && apar) {
479 apar->sample_rate = num_val;
480 } else if (!strcmp(key, "audiosamplesize") && apar) {
481 apar->bits_per_coded_sample = num_val;
482 } else if (!strcmp(key, "stereo") && apar) {
483 apar->channels = num_val + 1;
484 apar->channel_layout = apar->channels == 2 ?
485 AV_CH_LAYOUT_STEREO :
486 AV_CH_LAYOUT_MONO;
487 } else if (!strcmp(key, "width") && vpar) {
488 vpar->width = num_val;
489 } else if (!strcmp(key, "height") && vpar) {
490 vpar->height = num_val;
491 }
492 }
493 }
494 }
495
496 if (!strcmp(key, "duration") ||
497 !strcmp(key, "filesize") ||
498 !strcmp(key, "width") ||
499 !strcmp(key, "height") ||
500 !strcmp(key, "videodatarate") ||
501 !strcmp(key, "framerate") ||
502 !strcmp(key, "videocodecid") ||
503 !strcmp(key, "audiodatarate") ||
504 !strcmp(key, "audiosamplerate") ||
505 !strcmp(key, "audiosamplesize") ||
506 !strcmp(key, "stereo") ||
507 !strcmp(key, "audiocodecid") ||
508 !strcmp(key, "datastream"))
509 return 0;
510
511 s->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
512 if (amf_type == AMF_DATA_TYPE_BOOL) {
513 av_strlcpy(str_val, num_val > 0 ? "true" : "false",
514 sizeof(str_val));
515 av_dict_set(&s->metadata, key, str_val, 0);
516 } else if (amf_type == AMF_DATA_TYPE_NUMBER) {
517 snprintf(str_val, sizeof(str_val), "%.f", num_val);
518 av_dict_set(&s->metadata, key, str_val, 0);
519 } else if (amf_type == AMF_DATA_TYPE_STRING)
520 av_dict_set(&s->metadata, key, str_val, 0);
521 }
522
523 return 0;
524 }
525
526 static int flv_read_metabody(AVFormatContext *s, int64_t next_pos)
527 {
528 AMFDataType type;
529 AVStream *stream, *astream, *vstream;
530 AVIOContext *ioc;
531 int i;
532 // only needs to hold the string "onMetaData".
533 // Anything longer is something we don't want.
534 char buffer[11];
535
536 astream = NULL;
537 vstream = NULL;
538 ioc = s->pb;
539
540 // first object needs to be "onMetaData" string
541 type = avio_r8(ioc);
542 if (type != AMF_DATA_TYPE_STRING ||
543 amf_get_string(ioc, buffer, sizeof(buffer)) < 0)
544 return -1;
545
546 if (!strcmp(buffer, "onTextData"))
547 return 1;
548
549 if (strcmp(buffer, "onMetaData") && strcmp(buffer, "onCuePoint"))
550 return -1;
551
552 // find the streams now so that amf_parse_object doesn't need to do
553 // the lookup every time it is called.
554 for (i = 0; i < s->nb_streams; i++) {
555 stream = s->streams[i];
556 if (stream->codecpar->codec_type == AVMEDIA_TYPE_AUDIO)
557 astream = stream;
558 else if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO)
559 vstream = stream;
560 }
561
562 // parse the second object (we want a mixed array)
563 if (amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
564 return -1;
565
566 return 0;
567 }
568
569 static int flv_read_header(AVFormatContext *s)
570 {
571 int offset;
572
573 avio_skip(s->pb, 4);
574 avio_r8(s->pb); // flags
575
576 s->ctx_flags |= AVFMTCTX_NOHEADER;
577
578 offset = avio_rb32(s->pb);
579 avio_seek(s->pb, offset, SEEK_SET);
580 avio_skip(s->pb, 4);
581
582 s->start_time = 0;
583
584 return 0;
585 }
586
587 static int flv_read_close(AVFormatContext *s)
588 {
589 FLVContext *flv = s->priv_data;
590 av_freep(&flv->new_extradata[0]);
591 av_freep(&flv->new_extradata[1]);
592 return 0;
593 }
594
595 static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
596 {
597 av_free(st->codecpar->extradata);
598 st->codecpar->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
599 if (!st->codecpar->extradata)
600 return AVERROR(ENOMEM);
601 st->codecpar->extradata_size = size;
602 avio_read(s->pb, st->codecpar->extradata, st->codecpar->extradata_size);
603 return 0;
604 }
605
606 static int flv_queue_extradata(FLVContext *flv, AVIOContext *pb, int stream,
607 int size)
608 {
609 av_free(flv->new_extradata[stream]);
610 flv->new_extradata[stream] = av_mallocz(size +
611 AV_INPUT_BUFFER_PADDING_SIZE);
612 if (!flv->new_extradata[stream])
613 return AVERROR(ENOMEM);
614 flv->new_extradata_size[stream] = size;
615 avio_read(pb, flv->new_extradata[stream], size);
616 return 0;
617 }
618
619 static void clear_index_entries(AVFormatContext *s, int64_t pos)
620 {
621 int i, j, out;
622 av_log(s, AV_LOG_WARNING,
623 "Found invalid index entries, clearing the index.\n");
624 for (i = 0; i < s->nb_streams; i++) {
625 AVStream *st = s->streams[i];
626 /* Remove all index entries that point to >= pos */
627 out = 0;
628 for (j = 0; j < st->nb_index_entries; j++)
629 if (st->index_entries[j].pos < pos)
630 st->index_entries[out++] = st->index_entries[j];
631 st->nb_index_entries = out;
632 }
633 }
634
635 static int amf_skip_tag(AVIOContext *pb, AMFDataType type)
636 {
637 int nb = -1, ret, parse_name = 1;
638
639 switch (type) {
640 case AMF_DATA_TYPE_NUMBER:
641 avio_skip(pb, 8);
642 break;
643 case AMF_DATA_TYPE_BOOL:
644 avio_skip(pb, 1);
645 break;
646 case AMF_DATA_TYPE_STRING:
647 avio_skip(pb, avio_rb16(pb));
648 break;
649 case AMF_DATA_TYPE_ARRAY:
650 parse_name = 0;
651 case AMF_DATA_TYPE_MIXEDARRAY:
652 nb = avio_rb32(pb);
653 case AMF_DATA_TYPE_OBJECT:
654 while(!pb->eof_reached && (nb-- > 0 || type != AMF_DATA_TYPE_ARRAY)) {
655 if (parse_name) {
656 int size = avio_rb16(pb);
657 if (!size) {
658 avio_skip(pb, 1);
659 break;
660 }
661 avio_skip(pb, size);
662 }
663 if ((ret = amf_skip_tag(pb, avio_r8(pb))) < 0)
664 return ret;
665 }
666 break;
667 case AMF_DATA_TYPE_NULL:
668 case AMF_DATA_TYPE_OBJECT_END:
669 break;
670 default:
671 return AVERROR_INVALIDDATA;
672 }
673 return 0;
674 }
675
676 static int flv_data_packet(AVFormatContext *s, AVPacket *pkt,
677 int64_t dts, int64_t next)
678 {
679 AVIOContext *pb = s->pb;
680 AVStream *st = NULL;
681 char buf[20];
682 int ret = AVERROR_INVALIDDATA;
683 int i, length = -1;
684
685 switch (avio_r8(pb)) {
686 case AMF_DATA_TYPE_MIXEDARRAY:
687 avio_seek(pb, 4, SEEK_CUR);
688 case AMF_DATA_TYPE_OBJECT:
689 break;
690 default:
691 goto skip;
692 }
693
694 while ((ret = amf_get_string(pb, buf, sizeof(buf))) > 0) {
695 AMFDataType type = avio_r8(pb);
696 if (type == AMF_DATA_TYPE_STRING && !strcmp(buf, "text")) {
697 length = avio_rb16(pb);
698 ret = av_get_packet(pb, pkt, length);
699 if (ret < 0)
700 goto skip;
701 else
702 break;
703 } else {
704 if ((ret = amf_skip_tag(pb, type)) < 0)
705 goto skip;
706 }
707 }
708
709 if (length < 0) {
710 ret = AVERROR_INVALIDDATA;
711 goto skip;
712 }
713
714 for (i = 0; i < s->nb_streams; i++) {
715 st = s->streams[i];
716 if (st->codecpar->codec_type == AVMEDIA_TYPE_DATA)
717 break;
718 }
719
720 if (i == s->nb_streams) {
721 st = create_stream(s, AVMEDIA_TYPE_DATA);
722 if (!st)
723 return AVERROR(ENOMEM);
724 st->codecpar->codec_id = AV_CODEC_ID_TEXT;
725 }
726
727 pkt->dts = dts;
728 pkt->pts = dts;
729 pkt->size = ret;
730
731 pkt->stream_index = st->index;
732 pkt->flags |= AV_PKT_FLAG_KEY;
733
734 skip:
735 avio_seek(s->pb, next + 4, SEEK_SET);
736
737 return ret;
738 }
739
740 static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
741 {
742 FLVContext *flv = s->priv_data;
743 int ret, i, size, flags, is_audio;
744 enum FlvTagType type;
745 int64_t next, pos;
746 int64_t dts, pts = AV_NOPTS_VALUE;
747 int sample_rate = 0, channels = 0;
748 AVStream *st = NULL;
749
750 /* pkt size is repeated at end. skip it */
751 for (;; avio_skip(s->pb, 4)) {
752 pos = avio_tell(s->pb);
753 type = avio_r8(s->pb);
754 size = avio_rb24(s->pb);
755 dts = avio_rb24(s->pb);
756 dts |= avio_r8(s->pb) << 24;
757 av_log(s, AV_LOG_TRACE, "type:%d, size:%d, dts:%"PRId64"\n", type, size, dts);
758 if (s->pb->eof_reached)
759 return AVERROR_EOF;
760 avio_skip(s->pb, 3); /* stream id, always 0 */
761 flags = 0;
762
763 if (flv->validate_next < flv->validate_count) {
764 int64_t validate_pos = flv->validate_index[flv->validate_next].pos;
765 if (pos == validate_pos) {
766 if (FFABS(dts - flv->validate_index[flv->validate_next].dts) <=
767 VALIDATE_INDEX_TS_THRESH) {
768 flv->validate_next++;
769 } else {
770 clear_index_entries(s, validate_pos);
771 flv->validate_count = 0;
772 }
773 } else if (pos > validate_pos) {
774 clear_index_entries(s, validate_pos);
775 flv->validate_count = 0;
776 }
777 }
778
779 if (size == 0)
780 continue;
781
782 next = size + avio_tell(s->pb);
783
784 if (type == FLV_TAG_TYPE_AUDIO) {
785 is_audio = 1;
786 flags = avio_r8(s->pb);
787 size--;
788 } else if (type == FLV_TAG_TYPE_VIDEO) {
789 is_audio = 0;
790 flags = avio_r8(s->pb);
791 size--;
792 if ((flags & 0xf0) == 0x50) /* video info / command frame */
793 goto skip;
794 } else {
795 if (type == FLV_TAG_TYPE_META && size > 13 + 1 + 4)
796 if (flv_read_metabody(s, next) > 0) {
797 return flv_data_packet(s, pkt, dts, next);
798 } else /* skip packet */
799 av_log(s, AV_LOG_DEBUG,
800 "Skipping flv packet: type %d, size %d, flags %d.\n",
801 type, size, flags);
802
803 skip:
804 if (avio_seek(s->pb, next, SEEK_SET) != next) {
805 // This can happen if flv_read_metabody above read past
806 // next, on a non-seekable input, and the preceding data has
807 // been flushed out from the IO buffer.
808 av_log(s, AV_LOG_ERROR, "Unable to seek to the next packet\n");
809 return AVERROR_INVALIDDATA;
810 }
811 continue;
812 }
813
814 /* skip empty data packets */
815 if (!size)
816 continue;
817
818 /* now find stream */
819 for (i = 0; i < s->nb_streams; i++) {
820 st = s->streams[i];
821 if (is_audio && st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
822 if (flv_same_audio_codec(st->codecpar, flags))
823 break;
824 } else if (!is_audio &&
825 st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
826 if (flv_same_video_codec(st->codecpar, flags))
827 break;
828 }
829 }
830 if (i == s->nb_streams) {
831 st = create_stream(s, is_audio ? AVMEDIA_TYPE_AUDIO
832 : AVMEDIA_TYPE_VIDEO);
833 if (!st)
834 return AVERROR(ENOMEM);
835 }
836 av_log(s, AV_LOG_TRACE, "%d %X %d \n", is_audio, flags, st->discard);
837
838 if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY ||
839 is_audio)
840 av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
841
842 if ((st->discard >= AVDISCARD_NONKEY &&
843 !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY || is_audio)) ||
844 (st->discard >= AVDISCARD_BIDIR &&
845 ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && !is_audio)) ||
846 st->discard >= AVDISCARD_ALL) {
847 avio_seek(s->pb, next, SEEK_SET);
848 continue;
849 }
850 break;
851 }
852
853 // if not streamed and no duration from metadata then seek to end to find
854 // the duration from the timestamps
855 if ((s->pb->seekable & AVIO_SEEKABLE_NORMAL) &&
856 (!s->duration || s->duration == AV_NOPTS_VALUE) &&
857 !flv->searched_for_end) {
858 int size;
859 const int64_t pos = avio_tell(s->pb);
860 // Read the last 4 bytes of the file, this should be the size of the
861 // previous FLV tag. Use the timestamp of its payload as duration.
862 const int64_t fsize = avio_size(s->pb);
863 avio_seek(s->pb, fsize - 4, SEEK_SET);
864 size = avio_rb32(s->pb);
865 if (size > 0 && size < fsize) {
866 // Seek to the start of the last FLV tag at position (fsize - 4 - size)
867 // but skip the byte indicating the type.
868 avio_seek(s->pb, fsize - 3 - size, SEEK_SET);
869 if (size == avio_rb24(s->pb) + 11) {
870 uint32_t ts = avio_rb24(s->pb);
871 ts |= avio_r8(s->pb) << 24;
872 s->duration = ts * (int64_t)AV_TIME_BASE / 1000;
873 }
874 }
875 avio_seek(s->pb, pos, SEEK_SET);
876 flv->searched_for_end = 1;
877 }
878
879 if (is_audio) {
880 int bits_per_coded_sample;
881 channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
882 sample_rate = 44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >>
883 FLV_AUDIO_SAMPLERATE_OFFSET) >> 3;
884 bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
885 if (!st->codecpar->channels || !st->codecpar->sample_rate ||
886 !st->codecpar->bits_per_coded_sample) {
887 st->codecpar->channels = channels;
888 st->codecpar->channel_layout = channels == 1
889 ? AV_CH_LAYOUT_MONO
890 : AV_CH_LAYOUT_STEREO;
891 st->codecpar->sample_rate = sample_rate;
892 st->codecpar->bits_per_coded_sample = bits_per_coded_sample;
893 }
894 if (!st->codecpar->codec_id) {
895 flv_set_audio_codec(s, st, st->codecpar,
896 flags & FLV_AUDIO_CODECID_MASK);
897 flv->last_sample_rate =
898 sample_rate = st->codecpar->sample_rate;
899 flv->last_channels =
900 channels = st->codecpar->channels;
901 } else {
902 AVCodecParameters *par = avcodec_parameters_alloc();
903 if (!par) {
904 ret = AVERROR(ENOMEM);
905 goto leave;
906 }
907 par->sample_rate = sample_rate;
908 par->bits_per_coded_sample = bits_per_coded_sample;
909 flv_set_audio_codec(s, st, par, flags & FLV_AUDIO_CODECID_MASK);
910 sample_rate = par->sample_rate;
911 avcodec_parameters_free(&par);
912 }
913 } else {
914 size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK, 1);
915 }
916
917 if (st->codecpar->codec_id == AV_CODEC_ID_AAC ||
918 st->codecpar->codec_id == AV_CODEC_ID_H264) {
919 int type = avio_r8(s->pb);
920 size--;
921
922 if (size < 0) {
923 ret = AVERROR_INVALIDDATA;
924 goto leave;
925 }
926
927 if (st->codecpar->codec_id == AV_CODEC_ID_H264) {
928 // sign extension
929 int32_t cts = (avio_rb24(s->pb) + 0xff800000) ^ 0xff800000;
930 pts = dts + cts;
931 if (cts < 0 && !flv->wrong_dts) { // dts might be wrong
932 flv->wrong_dts = 1;
933 av_log(s, AV_LOG_WARNING,
934 "Negative cts, previous timestamps might be wrong.\n");
935 }
936 }
937 if (type == 0) {
938 if (st->codecpar->extradata) {
939 if ((ret = flv_queue_extradata(flv, s->pb, is_audio, size)) < 0)
940 return ret;
941 ret = AVERROR(EAGAIN);
942 goto leave;
943 }
944 if ((ret = flv_get_extradata(s, st, size)) < 0)
945 return ret;
946 if (st->codecpar->codec_id == AV_CODEC_ID_AAC) {
947 MPEG4AudioConfig cfg;
948
949 /* Workaround for buggy Omnia A/XE encoder */
950 AVDictionaryEntry *t = av_dict_get(s->metadata, "Encoder", NULL, 0);
951 if (t && !strcmp(t->value, "Omnia A/XE"))
952 st->codecpar->extradata_size = 2;
953
954 avpriv_mpeg4audio_get_config(&cfg, st->codecpar->extradata,
955 st->codecpar->extradata_size * 8, 1);
956 st->codecpar->channels = cfg.channels;
957 st->codecpar->channel_layout = 0;
958 if (cfg.ext_sample_rate)
959 st->codecpar->sample_rate = cfg.ext_sample_rate;
960 else
961 st->codecpar->sample_rate = cfg.sample_rate;
962 av_log(s, AV_LOG_TRACE, "mp4a config channels %d sample rate %d\n",
963 st->codecpar->channels, st->codecpar->sample_rate);
964 }
965
966 ret = AVERROR(EAGAIN);
967 goto leave;
968 }
969 }
970
971 /* skip empty data packets */
972 if (!size) {
973 ret = AVERROR(EAGAIN);
974 goto leave;
975 }
976
977 ret = av_get_packet(s->pb, pkt, size);
978 if (ret < 0)
979 return AVERROR(EIO);
980 /* note: we need to modify the packet size here to handle the last
981 * packet */
982 pkt->size = ret;
983 pkt->dts = dts;
984 pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
985 pkt->stream_index = st->index;
986 if (flv->new_extradata[is_audio]) {
987 uint8_t *side = av_packet_new_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA,
988 flv->new_extradata_size[is_audio]);
989 if (side) {
990 memcpy(side, flv->new_extradata[is_audio],
991 flv->new_extradata_size[is_audio]);
992 av_freep(&flv->new_extradata[is_audio]);
993 flv->new_extradata_size[is_audio] = 0;
994 }
995 }
996 if (is_audio && (sample_rate != flv->last_sample_rate ||
997 channels != flv->last_channels)) {
998 flv->last_sample_rate = sample_rate;
999 flv->last_channels = channels;
1000 ff_add_param_change(pkt, channels, 0, sample_rate, 0, 0);
1001 }
1002
1003 if (is_audio || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY))
1004 pkt->flags |= AV_PKT_FLAG_KEY;
1005
1006 leave:
1007 avio_skip(s->pb, 4);
1008 return ret;
1009 }
1010
1011 static int flv_read_seek(AVFormatContext *s, int stream_index,
1012 int64_t ts, int flags)
1013 {
1014 FLVContext *flv = s->priv_data;
1015 flv->validate_count = 0;
1016 return avio_seek_time(s->pb, stream_index, ts, flags);
1017 }
1018
1019 #define OFFSET(x) offsetof(FLVContext, x)
1020 #define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM
1021 static const AVOption options[] = {
1022 { "flv_metadata", "Allocate streams according to the onMetaData array", OFFSET(trust_metadata), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VD },
1023 { NULL }
1024 };
1025
1026 static const AVClass class = {
1027 .class_name = "flvdec",
1028 .item_name = av_default_item_name,
1029 .option = options,
1030 .version = LIBAVUTIL_VERSION_INT,
1031 };
1032
1033 AVInputFormat ff_flv_demuxer = {
1034 .name = "flv",
1035 .long_name = NULL_IF_CONFIG_SMALL("FLV (Flash Video)"),
1036 .priv_data_size = sizeof(FLVContext),
1037 .read_probe = flv_probe,
1038 .read_header = flv_read_header,
1039 .read_packet = flv_read_packet,
1040 .read_seek = flv_read_seek,
1041 .read_close = flv_read_close,
1042 .extensions = "flv",
1043 .priv_class = &class,
1044 };