flvdec: Treat all nellymoser versions as the same codec
[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 4bits: difference between encoded width and visible width
8 * - lower 4bits: 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/dict.h"
29 #include "libavutil/opt.h"
30 #include "libavutil/intfloat.h"
31 #include "libavutil/mathematics.h"
32 #include "libavcodec/bytestream.h"
33 #include "libavcodec/mpeg4audio.h"
34 #include "avformat.h"
35 #include "internal.h"
36 #include "avio_internal.h"
37 #include "flv.h"
38
39 #define KEYFRAMES_TAG "keyframes"
40 #define KEYFRAMES_TIMESTAMP_TAG "times"
41 #define KEYFRAMES_BYTEOFFSET_TAG "filepositions"
42
43 #define VALIDATE_INDEX_TS_THRESH 2500
44
45 typedef struct {
46 const AVClass *class; ///< Class for private options.
47 int trust_metadata; ///< configure streams according onMetaData
48 int wrong_dts; ///< wrong dts due to negative cts
49 uint8_t *new_extradata[2];
50 int new_extradata_size[2];
51 int last_sample_rate;
52 int last_channels;
53 struct {
54 int64_t dts;
55 int64_t pos;
56 } validate_index[2];
57 int validate_next;
58 int validate_count;
59 } FLVContext;
60
61 static int flv_probe(AVProbeData *p)
62 {
63 const uint8_t *d;
64
65 d = p->buf;
66 if (d[0] == 'F' && d[1] == 'L' && d[2] == 'V' && d[3] < 5 && d[5]==0 && AV_RB32(d+5)>8) {
67 return AVPROBE_SCORE_MAX;
68 }
69 return 0;
70 }
71
72 static AVStream *create_stream(AVFormatContext *s, int tag, int codec_type)
73 {
74 AVStream *st = avformat_new_stream(s, NULL);
75 if (!st)
76 return NULL;
77 st->id = tag;
78 st->codec->codec_type = codec_type;
79 avpriv_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
80 return st;
81 }
82 static int flv_same_audio_codec(AVCodecContext *acodec, int flags)
83 {
84 int bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
85 int flv_codecid = flags & FLV_AUDIO_CODECID_MASK;
86 int codec_id;
87
88 if (!acodec->codec_id && !acodec->codec_tag)
89 return 1;
90
91 if (acodec->bits_per_coded_sample != bits_per_coded_sample)
92 return 0;
93
94 switch(flv_codecid) {
95 //no distinction between S16 and S8 PCM codec flags
96 case FLV_CODECID_PCM:
97 codec_id = bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 :
98 #if HAVE_BIGENDIAN
99 CODEC_ID_PCM_S16BE;
100 #else
101 CODEC_ID_PCM_S16LE;
102 #endif
103 return codec_id == acodec->codec_id;
104 case FLV_CODECID_PCM_LE:
105 codec_id = bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 : CODEC_ID_PCM_S16LE;
106 return codec_id == acodec->codec_id;
107 case FLV_CODECID_AAC:
108 return acodec->codec_id == CODEC_ID_AAC;
109 case FLV_CODECID_ADPCM:
110 return acodec->codec_id == CODEC_ID_ADPCM_SWF;
111 case FLV_CODECID_SPEEX:
112 return acodec->codec_id == CODEC_ID_SPEEX;
113 case FLV_CODECID_MP3:
114 return acodec->codec_id == CODEC_ID_MP3;
115 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
116 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
117 case FLV_CODECID_NELLYMOSER:
118 return acodec->codec_id == CODEC_ID_NELLYMOSER;
119 case FLV_CODECID_PCM_MULAW:
120 return acodec->sample_rate == 8000 &&
121 acodec->codec_id == CODEC_ID_PCM_MULAW;
122 case FLV_CODECID_PCM_ALAW:
123 return acodec->sample_rate = 8000 &&
124 acodec->codec_id == CODEC_ID_PCM_ALAW;
125 default:
126 return acodec->codec_tag == (flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
127 }
128
129 return 0;
130 }
131
132 static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream, AVCodecContext *acodec, int flv_codecid) {
133 switch(flv_codecid) {
134 //no distinction between S16 and S8 PCM codec flags
135 case FLV_CODECID_PCM:
136 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 :
137 #if HAVE_BIGENDIAN
138 CODEC_ID_PCM_S16BE;
139 #else
140 CODEC_ID_PCM_S16LE;
141 #endif
142 break;
143 case FLV_CODECID_PCM_LE:
144 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 : CODEC_ID_PCM_S16LE; break;
145 case FLV_CODECID_AAC : acodec->codec_id = CODEC_ID_AAC; break;
146 case FLV_CODECID_ADPCM: acodec->codec_id = CODEC_ID_ADPCM_SWF; break;
147 case FLV_CODECID_SPEEX:
148 acodec->codec_id = CODEC_ID_SPEEX;
149 acodec->sample_rate = 16000;
150 break;
151 case FLV_CODECID_MP3 : acodec->codec_id = CODEC_ID_MP3 ; astream->need_parsing = AVSTREAM_PARSE_FULL; break;
152 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
153 acodec->sample_rate = 8000; //in case metadata does not otherwise declare samplerate
154 acodec->codec_id = CODEC_ID_NELLYMOSER;
155 break;
156 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
157 acodec->sample_rate = 16000;
158 acodec->codec_id = CODEC_ID_NELLYMOSER;
159 break;
160 case FLV_CODECID_NELLYMOSER:
161 acodec->codec_id = CODEC_ID_NELLYMOSER;
162 break;
163 case FLV_CODECID_PCM_MULAW:
164 acodec->sample_rate = 8000;
165 acodec->codec_id = CODEC_ID_PCM_MULAW;
166 break;
167 case FLV_CODECID_PCM_ALAW:
168 acodec->sample_rate = 8000;
169 acodec->codec_id = CODEC_ID_PCM_ALAW;
170 break;
171 default:
172 av_log(s, AV_LOG_INFO, "Unsupported audio codec (%x)\n", flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
173 acodec->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
174 }
175 }
176
177 static int flv_same_video_codec(AVCodecContext *vcodec, int flags)
178 {
179 int flv_codecid = flags & FLV_VIDEO_CODECID_MASK;
180
181 if (!vcodec->codec_id && !vcodec->codec_tag)
182 return 1;
183
184 switch (flv_codecid) {
185 case FLV_CODECID_H263:
186 return vcodec->codec_id == CODEC_ID_FLV1;
187 case FLV_CODECID_SCREEN:
188 return vcodec->codec_id == CODEC_ID_FLASHSV;
189 case FLV_CODECID_SCREEN2:
190 return vcodec->codec_id == CODEC_ID_FLASHSV2;
191 case FLV_CODECID_VP6:
192 return vcodec->codec_id == CODEC_ID_VP6F;
193 case FLV_CODECID_VP6A:
194 return vcodec->codec_id == CODEC_ID_VP6A;
195 case FLV_CODECID_H264:
196 return vcodec->codec_id == CODEC_ID_H264;
197 default:
198 return vcodec->codec_tag == flv_codecid;
199 }
200
201 return 0;
202 }
203
204 static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream, int flv_codecid) {
205 AVCodecContext *vcodec = vstream->codec;
206 switch(flv_codecid) {
207 case FLV_CODECID_H263 : vcodec->codec_id = CODEC_ID_FLV1 ; break;
208 case FLV_CODECID_SCREEN: vcodec->codec_id = CODEC_ID_FLASHSV; break;
209 case FLV_CODECID_SCREEN2: vcodec->codec_id = CODEC_ID_FLASHSV2; break;
210 case FLV_CODECID_VP6 : vcodec->codec_id = CODEC_ID_VP6F ;
211 case FLV_CODECID_VP6A :
212 if(flv_codecid == FLV_CODECID_VP6A)
213 vcodec->codec_id = CODEC_ID_VP6A;
214 if(vcodec->extradata_size != 1) {
215 vcodec->extradata_size = 1;
216 vcodec->extradata = av_malloc(1);
217 }
218 vcodec->extradata[0] = avio_r8(s->pb);
219 return 1; // 1 byte body size adjustment for flv_read_packet()
220 case FLV_CODECID_H264:
221 vcodec->codec_id = CODEC_ID_H264;
222 return 3; // not 4, reading packet type will consume one byte
223 default:
224 av_log(s, AV_LOG_INFO, "Unsupported video codec (%x)\n", flv_codecid);
225 vcodec->codec_tag = flv_codecid;
226 }
227
228 return 0;
229 }
230
231 static int amf_get_string(AVIOContext *ioc, char *buffer, int buffsize) {
232 int length = avio_rb16(ioc);
233 if(length >= buffsize) {
234 avio_skip(ioc, length);
235 return -1;
236 }
237
238 avio_read(ioc, buffer, length);
239
240 buffer[length] = '\0';
241
242 return length;
243 }
244
245 static int parse_keyframes_index(AVFormatContext *s, AVIOContext *ioc, AVStream *vstream, int64_t max_pos) {
246 FLVContext *flv = s->priv_data;
247 unsigned int arraylen = 0, timeslen = 0, fileposlen = 0, i;
248 double num_val;
249 char str_val[256];
250 int64_t *times = NULL;
251 int64_t *filepositions = NULL;
252 int ret = AVERROR(ENOSYS);
253 int64_t initial_pos = avio_tell(ioc);
254
255 if (s->flags & AVFMT_FLAG_IGNIDX)
256 return 0;
257
258 while (avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
259 int64_t* current_array;
260
261 // Expect array object in context
262 if (avio_r8(ioc) != AMF_DATA_TYPE_ARRAY)
263 break;
264
265 arraylen = avio_rb32(ioc);
266 if (arraylen >> 28)
267 break;
268
269 /*
270 * Expect only 'times' or 'filepositions' sub-arrays in other case refuse to use such metadata
271 * for indexing
272 */
273 if (!strcmp(KEYFRAMES_TIMESTAMP_TAG, str_val) && !times) {
274 if (!(times = av_mallocz(sizeof(*times) * arraylen))) {
275 ret = AVERROR(ENOMEM);
276 goto finish;
277 }
278 timeslen = arraylen;
279 current_array = times;
280 } else if (!strcmp(KEYFRAMES_BYTEOFFSET_TAG, str_val) && !filepositions) {
281 if (!(filepositions = av_mallocz(sizeof(*filepositions) * arraylen))) {
282 ret = AVERROR(ENOMEM);
283 goto finish;
284 }
285 fileposlen = arraylen;
286 current_array = filepositions;
287 } else // unexpected metatag inside keyframes, will not use such metadata for indexing
288 break;
289
290 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
291 if (avio_r8(ioc) != AMF_DATA_TYPE_NUMBER)
292 goto finish;
293 num_val = av_int2double(avio_rb64(ioc));
294 current_array[i] = num_val;
295 }
296 if (times && filepositions) {
297 // All done, exiting at a position allowing amf_parse_object
298 // to finish parsing the object
299 ret = 0;
300 break;
301 }
302 }
303
304 if (!ret && timeslen == fileposlen) {
305 for (i = 0; i < fileposlen; i++) {
306 av_add_index_entry(vstream, filepositions[i], times[i]*1000,
307 0, 0, AVINDEX_KEYFRAME);
308 if (i < 2) {
309 flv->validate_index[i].pos = filepositions[i];
310 flv->validate_index[i].dts = times[i] * 1000;
311 flv->validate_count = i + 1;
312 }
313 }
314 } else
315 av_log(s, AV_LOG_WARNING, "Invalid keyframes object, skipping.\n");
316
317 finish:
318 av_freep(&times);
319 av_freep(&filepositions);
320 // If we got unexpected data, but successfully reset back to
321 // the start pos, the caller can continue parsing
322 if (ret < 0 && avio_seek(ioc, initial_pos, SEEK_SET) > 0)
323 return 0;
324 return ret;
325 }
326
327 static int amf_parse_object(AVFormatContext *s, AVStream *astream, AVStream *vstream, const char *key, int64_t max_pos, int depth) {
328 AVCodecContext *acodec, *vcodec;
329 FLVContext *flv = s->priv_data;
330 AVIOContext *ioc;
331 AMFDataType amf_type;
332 char str_val[256];
333 double num_val;
334
335 num_val = 0;
336 ioc = s->pb;
337
338 amf_type = avio_r8(ioc);
339
340 switch(amf_type) {
341 case AMF_DATA_TYPE_NUMBER:
342 num_val = av_int2double(avio_rb64(ioc)); break;
343 case AMF_DATA_TYPE_BOOL:
344 num_val = avio_r8(ioc); break;
345 case AMF_DATA_TYPE_STRING:
346 if(amf_get_string(ioc, str_val, sizeof(str_val)) < 0)
347 return -1;
348 break;
349 case AMF_DATA_TYPE_OBJECT:
350 if ((vstream || astream) && key && !strcmp(KEYFRAMES_TAG, key) && depth == 1)
351 if (parse_keyframes_index(s, ioc, vstream ? vstream : astream,
352 max_pos) < 0)
353 return -1;
354
355 while (avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
356 if (amf_parse_object(s, astream, vstream, str_val, max_pos, depth + 1) < 0)
357 return -1; //if we couldn't skip, bomb out.
358 }
359 if(avio_r8(ioc) != AMF_END_OF_OBJECT)
360 return -1;
361 break;
362 case AMF_DATA_TYPE_NULL:
363 case AMF_DATA_TYPE_UNDEFINED:
364 case AMF_DATA_TYPE_UNSUPPORTED:
365 break; //these take up no additional space
366 case AMF_DATA_TYPE_MIXEDARRAY:
367 avio_skip(ioc, 4); //skip 32-bit max array index
368 while(avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
369 //this is the only case in which we would want a nested parse to not skip over the object
370 if(amf_parse_object(s, astream, vstream, str_val, max_pos, depth + 1) < 0)
371 return -1;
372 }
373 if(avio_r8(ioc) != AMF_END_OF_OBJECT)
374 return -1;
375 break;
376 case AMF_DATA_TYPE_ARRAY: {
377 unsigned int arraylen, i;
378
379 arraylen = avio_rb32(ioc);
380 for(i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
381 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
382 return -1; //if we couldn't skip, bomb out.
383 }
384 }
385 break;
386 case AMF_DATA_TYPE_DATE:
387 avio_skip(ioc, 8 + 2); //timestamp (double) and UTC offset (int16)
388 break;
389 default: //unsupported type, we couldn't skip
390 return -1;
391 }
392
393 if(depth == 1 && key) { //only look for metadata values when we are not nested and key != NULL
394 acodec = astream ? astream->codec : NULL;
395 vcodec = vstream ? vstream->codec : NULL;
396
397 if (amf_type == AMF_DATA_TYPE_NUMBER) {
398 if (!strcmp(key, "duration"))
399 s->duration = num_val * AV_TIME_BASE;
400 else if (!strcmp(key, "videodatarate") && vcodec && 0 <= (int)(num_val * 1024.0))
401 vcodec->bit_rate = num_val * 1024.0;
402 else if (!strcmp(key, "audiodatarate") && acodec && 0 <= (int)(num_val * 1024.0))
403 acodec->bit_rate = num_val * 1024.0;
404 else if (!strcmp(key, "datastream")) {
405 AVStream *st = create_stream(s, 2, AVMEDIA_TYPE_DATA);
406 if (!st)
407 return AVERROR(ENOMEM);
408 st->codec->codec_id = CODEC_ID_TEXT;
409 } else if (flv->trust_metadata) {
410 if (!strcmp(key, "videocodecid") && vcodec) {
411 flv_set_video_codec(s, vstream, num_val);
412 } else
413 if (!strcmp(key, "audiocodecid") && acodec) {
414 flv_set_audio_codec(s, astream, acodec, num_val);
415 } else
416 if (!strcmp(key, "audiosamplerate") && acodec) {
417 acodec->sample_rate = num_val;
418 } else
419 if (!strcmp(key, "width") && vcodec) {
420 vcodec->width = num_val;
421 } else
422 if (!strcmp(key, "height") && vcodec) {
423 vcodec->height = num_val;
424 }
425 }
426 }
427
428 if (!strcmp(key, "duration") ||
429 !strcmp(key, "filesize") ||
430 !strcmp(key, "width") ||
431 !strcmp(key, "height") ||
432 !strcmp(key, "videodatarate") ||
433 !strcmp(key, "framerate") ||
434 !strcmp(key, "videocodecid") ||
435 !strcmp(key, "audiodatarate") ||
436 !strcmp(key, "audiosamplerate") ||
437 !strcmp(key, "audiosamplesize") ||
438 !strcmp(key, "stereo") ||
439 !strcmp(key, "audiocodecid"))
440 return 0;
441
442 if(amf_type == AMF_DATA_TYPE_BOOL) {
443 av_strlcpy(str_val, num_val > 0 ? "true" : "false", sizeof(str_val));
444 av_dict_set(&s->metadata, key, str_val, 0);
445 } else if(amf_type == AMF_DATA_TYPE_NUMBER) {
446 snprintf(str_val, sizeof(str_val), "%.f", num_val);
447 av_dict_set(&s->metadata, key, str_val, 0);
448 } else if (amf_type == AMF_DATA_TYPE_STRING)
449 av_dict_set(&s->metadata, key, str_val, 0);
450 }
451
452 return 0;
453 }
454
455 static int flv_read_metabody(AVFormatContext *s, int64_t next_pos) {
456 AMFDataType type;
457 AVStream *stream, *astream, *vstream;
458 AVIOContext *ioc;
459 int i;
460 char buffer[11]; //only needs to hold the string "onMetaData". Anything longer is something we don't want.
461
462 astream = NULL;
463 vstream = NULL;
464 ioc = s->pb;
465
466 //first object needs to be "onMetaData" string
467 type = avio_r8(ioc);
468 if (type != AMF_DATA_TYPE_STRING ||
469 amf_get_string(ioc, buffer, sizeof(buffer)) < 0)
470 return -1;
471
472 if (!strcmp(buffer, "onTextData"))
473 return 1;
474
475 if (strcmp(buffer, "onMetaData"))
476 return -1;
477
478 //find the streams now so that amf_parse_object doesn't need to do the lookup every time it is called.
479 for(i = 0; i < s->nb_streams; i++) {
480 stream = s->streams[i];
481 if (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO) astream = stream;
482 else if(stream->codec->codec_type == AVMEDIA_TYPE_VIDEO) vstream = stream;
483 }
484
485 //parse the second object (we want a mixed array)
486 if(amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
487 return -1;
488
489 return 0;
490 }
491
492 static int flv_read_header(AVFormatContext *s)
493 {
494 int offset, flags;
495
496 avio_skip(s->pb, 4);
497 flags = avio_r8(s->pb);
498 /* old flvtool cleared this field */
499 /* FIXME: better fix needed */
500 if (!flags) {
501 flags = FLV_HEADER_FLAG_HASVIDEO | FLV_HEADER_FLAG_HASAUDIO;
502 av_log(s, AV_LOG_WARNING, "Broken FLV file, which says no streams present, this might fail\n");
503 }
504
505 if((flags & (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
506 != (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
507 s->ctx_flags |= AVFMTCTX_NOHEADER;
508
509 if(flags & FLV_HEADER_FLAG_HASVIDEO){
510 if(!create_stream(s, 0, AVMEDIA_TYPE_VIDEO))
511 return AVERROR(ENOMEM);
512 }
513 if(flags & FLV_HEADER_FLAG_HASAUDIO){
514 if(!create_stream(s, 1, AVMEDIA_TYPE_AUDIO))
515 return AVERROR(ENOMEM);
516 }
517
518 offset = avio_rb32(s->pb);
519 avio_seek(s->pb, offset, SEEK_SET);
520 avio_skip(s->pb, 4);
521
522 s->start_time = 0;
523
524 return 0;
525 }
526
527 static int flv_read_close(AVFormatContext *s)
528 {
529 FLVContext *flv = s->priv_data;
530 av_freep(&flv->new_extradata[0]);
531 av_freep(&flv->new_extradata[1]);
532 return 0;
533 }
534
535 static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
536 {
537 av_free(st->codec->extradata);
538 st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
539 if (!st->codec->extradata)
540 return AVERROR(ENOMEM);
541 st->codec->extradata_size = size;
542 avio_read(s->pb, st->codec->extradata, st->codec->extradata_size);
543 return 0;
544 }
545
546 static int flv_queue_extradata(FLVContext *flv, AVIOContext *pb, int stream,
547 int size)
548 {
549 av_free(flv->new_extradata[stream]);
550 flv->new_extradata[stream] = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
551 if (!flv->new_extradata[stream])
552 return AVERROR(ENOMEM);
553 flv->new_extradata_size[stream] = size;
554 avio_read(pb, flv->new_extradata[stream], size);
555 return 0;
556 }
557
558 static void clear_index_entries(AVFormatContext *s, int64_t pos)
559 {
560 int i, j, out;
561 av_log(s, AV_LOG_WARNING, "Found invalid index entries, clearing the index.\n");
562 for (i = 0; i < s->nb_streams; i++) {
563 AVStream *st = s->streams[i];
564 /* Remove all index entries that point to >= pos */
565 out = 0;
566 for (j = 0; j < st->nb_index_entries; j++) {
567 if (st->index_entries[j].pos < pos)
568 st->index_entries[out++] = st->index_entries[j];
569 }
570 st->nb_index_entries = out;
571 }
572 }
573
574
575 static int flv_data_packet(AVFormatContext *s, AVPacket *pkt,
576 int64_t dts, int64_t next)
577 {
578 int ret = AVERROR_INVALIDDATA, i;
579 AVIOContext *pb = s->pb;
580 AVStream *st = NULL;
581 AMFDataType type;
582 char buf[20];
583 int length;
584
585 type = avio_r8(pb);
586 if (type == AMF_DATA_TYPE_MIXEDARRAY)
587 avio_seek(pb, 4, SEEK_CUR);
588 else if (type != AMF_DATA_TYPE_OBJECT)
589 goto out;
590
591 amf_get_string(pb, buf, sizeof(buf));
592 if (strcmp(buf, "type") || avio_r8(pb) != AMF_DATA_TYPE_STRING)
593 goto out;
594
595 amf_get_string(pb, buf, sizeof(buf));
596 //FIXME parse it as codec_id
597 amf_get_string(pb, buf, sizeof(buf));
598 if (strcmp(buf, "text") || avio_r8(pb) != AMF_DATA_TYPE_STRING)
599 goto out;
600
601 length = avio_rb16(pb);
602 ret = av_get_packet(s->pb, pkt, length);
603 if (ret < 0) {
604 ret = AVERROR(EIO);
605 goto out;
606 }
607
608 for (i = 0; i < s->nb_streams; i++) {
609 st = s->streams[i];
610 if (st->codec->codec_type == AVMEDIA_TYPE_DATA)
611 break;
612 }
613
614 if (i == s->nb_streams) {
615 st = create_stream(s, 2, AVMEDIA_TYPE_DATA);
616 if (!st)
617 goto out;
618 st->codec->codec_id = CODEC_ID_TEXT;
619 }
620
621 pkt->dts = dts;
622 pkt->pts = dts;
623 pkt->size = ret;
624
625 pkt->stream_index = st->index;
626 pkt->flags |= AV_PKT_FLAG_KEY;
627
628 avio_seek(s->pb, next + 4, SEEK_SET);
629 out:
630 return ret;
631 }
632
633 static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
634 {
635 FLVContext *flv = s->priv_data;
636 int ret, i, type, size, flags, is_audio;
637 int64_t next, pos;
638 int64_t dts, pts = AV_NOPTS_VALUE;
639 int sample_rate = 0, channels = 0;
640 AVStream *st = NULL;
641
642 for(;;avio_skip(s->pb, 4)){ /* pkt size is repeated at end. skip it */
643 pos = avio_tell(s->pb);
644 type = avio_r8(s->pb);
645 size = avio_rb24(s->pb);
646 dts = avio_rb24(s->pb);
647 dts |= avio_r8(s->pb) << 24;
648 av_dlog(s, "type:%d, size:%d, dts:%"PRId64"\n", type, size, dts);
649 if (s->pb->eof_reached)
650 return AVERROR_EOF;
651 avio_skip(s->pb, 3); /* stream id, always 0 */
652 flags = 0;
653
654 if (flv->validate_next < flv->validate_count) {
655 int64_t validate_pos = flv->validate_index[flv->validate_next].pos;
656 if (pos == validate_pos) {
657 if (FFABS(dts - flv->validate_index[flv->validate_next].dts) <=
658 VALIDATE_INDEX_TS_THRESH) {
659 flv->validate_next++;
660 } else {
661 clear_index_entries(s, validate_pos);
662 flv->validate_count = 0;
663 }
664 } else if (pos > validate_pos) {
665 clear_index_entries(s, validate_pos);
666 flv->validate_count = 0;
667 }
668 }
669
670 if(size == 0)
671 continue;
672
673 next= size + avio_tell(s->pb);
674
675 if (type == FLV_TAG_TYPE_AUDIO) {
676 is_audio=1;
677 flags = avio_r8(s->pb);
678 size--;
679 } else if (type == FLV_TAG_TYPE_VIDEO) {
680 is_audio=0;
681 flags = avio_r8(s->pb);
682 size--;
683 if ((flags & 0xf0) == 0x50) /* video info / command frame */
684 goto skip;
685 } else {
686 if (type == FLV_TAG_TYPE_META && size > 13+1+4)
687 if (flv_read_metabody(s, next) > 0) {
688 return flv_data_packet(s, pkt, dts, next);
689 }
690 else /* skip packet */
691 av_log(s, AV_LOG_DEBUG, "skipping flv packet: type %d, size %d, flags %d\n", type, size, flags);
692 skip:
693 avio_seek(s->pb, next, SEEK_SET);
694 continue;
695 }
696
697 /* skip empty data packets */
698 if (!size)
699 continue;
700
701 /* now find stream */
702 for(i=0;i<s->nb_streams;i++) {
703 st = s->streams[i];
704 if (is_audio && st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
705 if (flv_same_audio_codec(st->codec, flags)) {
706 break;
707 }
708 } else
709 if (!is_audio && st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
710 if (flv_same_video_codec(st->codec, flags)) {
711 break;
712 }
713 }
714 }
715 if(i == s->nb_streams){
716 st = create_stream(s, is_audio,
717 is_audio ? AVMEDIA_TYPE_AUDIO : AVMEDIA_TYPE_VIDEO);
718 s->ctx_flags &= ~AVFMTCTX_NOHEADER;
719 }
720 av_dlog(s, "%d %X %d \n", is_audio, flags, st->discard);
721 if( (st->discard >= AVDISCARD_NONKEY && !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY || is_audio))
722 ||(st->discard >= AVDISCARD_BIDIR && ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && !is_audio))
723 || st->discard >= AVDISCARD_ALL
724 ){
725 avio_seek(s->pb, next, SEEK_SET);
726 continue;
727 }
728 if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY)
729 av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
730 break;
731 }
732
733 // if not streamed and no duration from metadata then seek to end to find the duration from the timestamps
734 if(s->pb->seekable && (!s->duration || s->duration==AV_NOPTS_VALUE)){
735 int size;
736 const int64_t pos= avio_tell(s->pb);
737 const int64_t fsize= avio_size(s->pb);
738 avio_seek(s->pb, fsize-4, SEEK_SET);
739 size= avio_rb32(s->pb);
740 avio_seek(s->pb, fsize-3-size, SEEK_SET);
741 if(size == avio_rb24(s->pb) + 11){
742 uint32_t ts = avio_rb24(s->pb);
743 ts |= avio_r8(s->pb) << 24;
744 s->duration = ts * (int64_t)AV_TIME_BASE / 1000;
745 }
746 avio_seek(s->pb, pos, SEEK_SET);
747 }
748
749 if(is_audio){
750 int bits_per_coded_sample;
751 channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
752 sample_rate = (44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3);
753 bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
754 if(!st->codec->channels || !st->codec->sample_rate || !st->codec->bits_per_coded_sample) {
755 st->codec->channels = channels;
756 st->codec->sample_rate = sample_rate;
757 st->codec->bits_per_coded_sample = bits_per_coded_sample;
758 }
759 if(!st->codec->codec_id){
760 flv_set_audio_codec(s, st, st->codec, flags & FLV_AUDIO_CODECID_MASK);
761 flv->last_sample_rate = sample_rate = st->codec->sample_rate;
762 flv->last_channels = channels = st->codec->channels;
763 } else {
764 AVCodecContext ctx;
765 ctx.sample_rate = sample_rate;
766 flv_set_audio_codec(s, st, &ctx, flags & FLV_AUDIO_CODECID_MASK);
767 sample_rate = ctx.sample_rate;
768 }
769 }else{
770 size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK);
771 }
772
773 if (st->codec->codec_id == CODEC_ID_AAC ||
774 st->codec->codec_id == CODEC_ID_H264) {
775 int type = avio_r8(s->pb);
776 size--;
777 if (st->codec->codec_id == CODEC_ID_H264) {
778 int32_t cts = (avio_rb24(s->pb)+0xff800000)^0xff800000; // sign extension
779 pts = dts + cts;
780 if (cts < 0) { // dts are wrong
781 flv->wrong_dts = 1;
782 av_log(s, AV_LOG_WARNING, "negative cts, previous timestamps might be wrong\n");
783 }
784 if (flv->wrong_dts)
785 dts = AV_NOPTS_VALUE;
786 }
787 if (type == 0) {
788 if (st->codec->extradata) {
789 if ((ret = flv_queue_extradata(flv, s->pb, is_audio, size)) < 0)
790 return ret;
791 ret = AVERROR(EAGAIN);
792 goto leave;
793 }
794 if ((ret = flv_get_extradata(s, st, size)) < 0)
795 return ret;
796 if (st->codec->codec_id == CODEC_ID_AAC) {
797 MPEG4AudioConfig cfg;
798 avpriv_mpeg4audio_get_config(&cfg, st->codec->extradata,
799 st->codec->extradata_size * 8, 1);
800 st->codec->channels = cfg.channels;
801 if (cfg.ext_sample_rate)
802 st->codec->sample_rate = cfg.ext_sample_rate;
803 else
804 st->codec->sample_rate = cfg.sample_rate;
805 av_dlog(s, "mp4a config channels %d sample rate %d\n",
806 st->codec->channels, st->codec->sample_rate);
807 }
808
809 ret = AVERROR(EAGAIN);
810 goto leave;
811 }
812 }
813
814 /* skip empty data packets */
815 if (!size) {
816 ret = AVERROR(EAGAIN);
817 goto leave;
818 }
819
820 ret= av_get_packet(s->pb, pkt, size);
821 if (ret < 0) {
822 return AVERROR(EIO);
823 }
824 /* note: we need to modify the packet size here to handle the last
825 packet */
826 pkt->size = ret;
827 pkt->dts = dts;
828 pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
829 pkt->stream_index = st->index;
830 if (flv->new_extradata[is_audio]) {
831 uint8_t *side = av_packet_new_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA,
832 flv->new_extradata_size[is_audio]);
833 if (side) {
834 memcpy(side, flv->new_extradata[is_audio],
835 flv->new_extradata_size[is_audio]);
836 av_freep(&flv->new_extradata[is_audio]);
837 flv->new_extradata_size[is_audio] = 0;
838 }
839 }
840 if (is_audio && (sample_rate != flv->last_sample_rate ||
841 channels != flv->last_channels)) {
842 flv->last_sample_rate = sample_rate;
843 flv->last_channels = channels;
844 ff_add_param_change(pkt, channels, 0, sample_rate, 0, 0);
845 }
846
847 if (is_audio || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY))
848 pkt->flags |= AV_PKT_FLAG_KEY;
849
850 leave:
851 avio_skip(s->pb, 4);
852 return ret;
853 }
854
855 static int flv_read_seek(AVFormatContext *s, int stream_index,
856 int64_t ts, int flags)
857 {
858 FLVContext *flv = s->priv_data;
859 flv->validate_count = 0;
860 return avio_seek_time(s->pb, stream_index, ts, flags);
861 }
862
863 #define OFFSET(x) offsetof(FLVContext, x)
864 #define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM
865 static const AVOption options[] = {
866 { "flv_metadata", "Allocate streams according the onMetaData array", OFFSET(trust_metadata), AV_OPT_TYPE_INT, { 0 }, 0, 1, VD},
867 { NULL }
868 };
869
870 static const AVClass class = {
871 .class_name = "flvdec",
872 .item_name = av_default_item_name,
873 .option = options,
874 .version = LIBAVUTIL_VERSION_INT,
875 };
876
877 AVInputFormat ff_flv_demuxer = {
878 .name = "flv",
879 .long_name = NULL_IF_CONFIG_SMALL("FLV format"),
880 .priv_data_size = sizeof(FLVContext),
881 .read_probe = flv_probe,
882 .read_header = flv_read_header,
883 .read_packet = flv_read_packet,
884 .read_seek = flv_read_seek,
885 .read_close = flv_read_close,
886 .extensions = "flv",
887 .priv_class = &class,
888 };