Set protocol-level metadata in AVFormatContext any time a packet is read.
[libav.git] / libavformat / flvdec.c
CommitLineData
d4f5d74a 1/*
7fbde343 2 * FLV demuxer
2912e87a 3 * Copyright (c) 2003 The Libav Project
d4f5d74a 4 *
7b94177e
DB
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 *
2912e87a 10 * This file is part of Libav.
b78e7197 11 *
2912e87a 12 * Libav is free software; you can redistribute it and/or
d4f5d74a
GM
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
b78e7197 15 * version 2.1 of the License, or (at your option) any later version.
d4f5d74a 16 *
2912e87a 17 * Libav is distributed in the hope that it will be useful,
d4f5d74a
GM
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
2912e87a 23 * License along with Libav; if not, write to the Free Software
5509bffa 24 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
d4f5d74a 25 */
d2718187 26
df2bd71a 27#include "libavutil/avstring.h"
644d8d2e 28#include "libavutil/channel_layout.h"
d2d67e42 29#include "libavutil/dict.h"
5b54a90c 30#include "libavutil/opt.h"
3383a53e 31#include "libavutil/intfloat.h"
d3f751e6 32#include "libavutil/mathematics.h"
c1c206b3 33#include "libavcodec/bytestream.h"
d2718187 34#include "libavcodec/mpeg4audio.h"
d4f5d74a 35#include "avformat.h"
c3f9ebf7 36#include "internal.h"
933e90a6 37#include "avio_internal.h"
6cac3a3b 38#include "flv.h"
d4f5d74a 39
cb7e2c1c
KA
40#define KEYFRAMES_TAG "keyframes"
41#define KEYFRAMES_TIMESTAMP_TAG "times"
42#define KEYFRAMES_BYTEOFFSET_TAG "filepositions"
43
7e297a46
MS
44#define VALIDATE_INDEX_TS_THRESH 2500
45
ebd61055 46typedef struct {
5b54a90c 47 const AVClass *class; ///< Class for private options.
e4529df9
DB
48 int trust_metadata; ///< configure streams according onMetaData
49 int wrong_dts; ///< wrong dts due to negative cts
251f320f 50 uint8_t *new_extradata[2];
e4529df9
DB
51 int new_extradata_size[2];
52 int last_sample_rate;
53 int last_channels;
7e297a46
MS
54 struct {
55 int64_t dts;
56 int64_t pos;
57 } validate_index[2];
58 int validate_next;
59 int validate_count;
ebd61055
BC
60} FLVContext;
61
d4f5d74a
GM
62static int flv_probe(AVProbeData *p)
63{
64 const uint8_t *d;
65
d4f5d74a 66 d = p->buf;
e4529df9
DB
67 if (d[0] == 'F' &&
68 d[1] == 'L' &&
69 d[2] == 'V' &&
70 d[3] < 5 && d[5] == 0 &&
71 AV_RB32(d + 5) > 8) {
74248229 72 return AVPROBE_SCORE_MAX;
d4f5d74a
GM
73 }
74 return 0;
75}
76
41f43202 77static AVStream *create_stream(AVFormatContext *s, int codec_type)
21e2dc9f
LB
78{
79 AVStream *st = avformat_new_stream(s, NULL);
80 if (!st)
81 return NULL;
21e2dc9f
LB
82 st->codec->codec_type = codec_type;
83 avpriv_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
84 return st;
85}
e4529df9 86
09a445ce
LB
87static int flv_same_audio_codec(AVCodecContext *acodec, int flags)
88{
89 int bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
e4529df9 90 int flv_codecid = flags & FLV_AUDIO_CODECID_MASK;
09a445ce
LB
91 int codec_id;
92
93 if (!acodec->codec_id && !acodec->codec_tag)
94 return 1;
95
96 if (acodec->bits_per_coded_sample != bits_per_coded_sample)
97 return 0;
98
e4529df9
DB
99 switch (flv_codecid) {
100 // no distinction between S16 and S8 PCM codec flags
09a445ce 101 case FLV_CODECID_PCM:
e4529df9
DB
102 codec_id = bits_per_coded_sample == 8
103 ? AV_CODEC_ID_PCM_U8
09a445ce 104#if HAVE_BIGENDIAN
e4529df9 105 : AV_CODEC_ID_PCM_S16BE;
09a445ce 106#else
e4529df9 107 : AV_CODEC_ID_PCM_S16LE;
09a445ce
LB
108#endif
109 return codec_id == acodec->codec_id;
110 case FLV_CODECID_PCM_LE:
e4529df9
DB
111 codec_id = bits_per_coded_sample == 8
112 ? AV_CODEC_ID_PCM_U8
113 : AV_CODEC_ID_PCM_S16LE;
09a445ce
LB
114 return codec_id == acodec->codec_id;
115 case FLV_CODECID_AAC:
36ef5369 116 return acodec->codec_id == AV_CODEC_ID_AAC;
09a445ce 117 case FLV_CODECID_ADPCM:
36ef5369 118 return acodec->codec_id == AV_CODEC_ID_ADPCM_SWF;
09a445ce 119 case FLV_CODECID_SPEEX:
36ef5369 120 return acodec->codec_id == AV_CODEC_ID_SPEEX;
09a445ce 121 case FLV_CODECID_MP3:
36ef5369 122 return acodec->codec_id == AV_CODEC_ID_MP3;
09a445ce 123 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
09a445ce 124 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
09a445ce 125 case FLV_CODECID_NELLYMOSER:
36ef5369 126 return acodec->codec_id == AV_CODEC_ID_NELLYMOSER;
09a445ce
LB
127 case FLV_CODECID_PCM_MULAW:
128 return acodec->sample_rate == 8000 &&
e4529df9 129 acodec->codec_id == AV_CODEC_ID_PCM_MULAW;
09a445ce 130 case FLV_CODECID_PCM_ALAW:
390b4d70 131 return acodec->sample_rate == 8000 &&
e4529df9 132 acodec->codec_id == AV_CODEC_ID_PCM_ALAW;
09a445ce
LB
133 default:
134 return acodec->codec_tag == (flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
135 }
09a445ce 136}
21e2dc9f 137
e4529df9
DB
138static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream,
139 AVCodecContext *acodec, int flv_codecid)
140{
141 switch (flv_codecid) {
142 // no distinction between S16 and S8 PCM codec flags
143 case FLV_CODECID_PCM:
144 acodec->codec_id = acodec->bits_per_coded_sample == 8
145 ? AV_CODEC_ID_PCM_U8
63613fe6 146#if HAVE_BIGENDIAN
e4529df9 147 : AV_CODEC_ID_PCM_S16BE;
58293e57 148#else
e4529df9 149 : AV_CODEC_ID_PCM_S16LE;
58293e57 150#endif
e4529df9
DB
151 break;
152 case FLV_CODECID_PCM_LE:
153 acodec->codec_id = acodec->bits_per_coded_sample == 8
154 ? AV_CODEC_ID_PCM_U8
155 : AV_CODEC_ID_PCM_S16LE;
156 break;
157 case FLV_CODECID_AAC:
158 acodec->codec_id = AV_CODEC_ID_AAC;
159 break;
160 case FLV_CODECID_ADPCM:
161 acodec->codec_id = AV_CODEC_ID_ADPCM_SWF;
162 break;
163 case FLV_CODECID_SPEEX:
164 acodec->codec_id = AV_CODEC_ID_SPEEX;
165 acodec->sample_rate = 16000;
166 break;
167 case FLV_CODECID_MP3:
168 acodec->codec_id = AV_CODEC_ID_MP3;
169 astream->need_parsing = AVSTREAM_PARSE_FULL;
170 break;
171 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
172 // in case metadata does not otherwise declare samplerate
173 acodec->sample_rate = 8000;
174 acodec->codec_id = AV_CODEC_ID_NELLYMOSER;
175 break;
176 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
177 acodec->sample_rate = 16000;
178 acodec->codec_id = AV_CODEC_ID_NELLYMOSER;
179 break;
180 case FLV_CODECID_NELLYMOSER:
181 acodec->codec_id = AV_CODEC_ID_NELLYMOSER;
182 break;
183 case FLV_CODECID_PCM_MULAW:
184 acodec->sample_rate = 8000;
185 acodec->codec_id = AV_CODEC_ID_PCM_MULAW;
186 break;
187 case FLV_CODECID_PCM_ALAW:
188 acodec->sample_rate = 8000;
189 acodec->codec_id = AV_CODEC_ID_PCM_ALAW;
190 break;
191 default:
192 av_log(s, AV_LOG_INFO, "Unsupported audio codec (%x)\n",
193 flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
194 acodec->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
428cc588
AH
195 }
196}
197
09a445ce
LB
198static int flv_same_video_codec(AVCodecContext *vcodec, int flags)
199{
200 int flv_codecid = flags & FLV_VIDEO_CODECID_MASK;
201
202 if (!vcodec->codec_id && !vcodec->codec_tag)
203 return 1;
204
205 switch (flv_codecid) {
e4529df9
DB
206 case FLV_CODECID_H263:
207 return vcodec->codec_id == AV_CODEC_ID_FLV1;
208 case FLV_CODECID_SCREEN:
209 return vcodec->codec_id == AV_CODEC_ID_FLASHSV;
210 case FLV_CODECID_SCREEN2:
211 return vcodec->codec_id == AV_CODEC_ID_FLASHSV2;
212 case FLV_CODECID_VP6:
213 return vcodec->codec_id == AV_CODEC_ID_VP6F;
214 case FLV_CODECID_VP6A:
215 return vcodec->codec_id == AV_CODEC_ID_VP6A;
216 case FLV_CODECID_H264:
217 return vcodec->codec_id == AV_CODEC_ID_H264;
218 default:
219 return vcodec->codec_tag == flv_codecid;
09a445ce 220 }
09a445ce
LB
221}
222
e4529df9
DB
223static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream,
224 int flv_codecid, int read)
225{
428cc588 226 AVCodecContext *vcodec = vstream->codec;
e4529df9
DB
227 switch (flv_codecid) {
228 case FLV_CODECID_H263:
229 vcodec->codec_id = AV_CODEC_ID_FLV1;
230 break;
231 case FLV_CODECID_SCREEN:
232 vcodec->codec_id = AV_CODEC_ID_FLASHSV;
233 break;
234 case FLV_CODECID_SCREEN2:
235 vcodec->codec_id = AV_CODEC_ID_FLASHSV2;
236 break;
237 case FLV_CODECID_VP6:
238 vcodec->codec_id = AV_CODEC_ID_VP6F;
239 case FLV_CODECID_VP6A:
240 if (flv_codecid == FLV_CODECID_VP6A)
241 vcodec->codec_id = AV_CODEC_ID_VP6A;
242 if (read) {
243 if (vcodec->extradata_size != 1) {
244 vcodec->extradata = av_malloc(1);
c5a738ca 245 if (vcodec->extradata)
e4529df9 246 vcodec->extradata_size = 1;
428cc588 247 }
e4529df9
DB
248 if (vcodec->extradata)
249 vcodec->extradata[0] = avio_r8(s->pb);
250 else
251 avio_skip(s->pb, 1);
252 }
253 return 1; // 1 byte body size adjustment for flv_read_packet()
254 case FLV_CODECID_H264:
255 vcodec->codec_id = AV_CODEC_ID_H264;
256 return 3; // not 4, reading packet type will consume one byte
257 default:
258 av_log(s, AV_LOG_INFO, "Unsupported video codec (%x)\n", flv_codecid);
259 vcodec->codec_tag = flv_codecid;
428cc588
AH
260 }
261
262 return 0;
263}
264
e4529df9
DB
265static int amf_get_string(AVIOContext *ioc, char *buffer, int buffsize)
266{
b7effd4e 267 int length = avio_rb16(ioc);
e4529df9 268 if (length >= buffsize) {
45a8a02a 269 avio_skip(ioc, length);
cc38e063
MN
270 return -1;
271 }
896bcd2e 272
b7effd4e 273 avio_read(ioc, buffer, length);
896bcd2e 274
cc38e063 275 buffer[length] = '\0';
896bcd2e 276
cc38e063 277 return length;
896bcd2e
MN
278}
279
e4529df9
DB
280static int parse_keyframes_index(AVFormatContext *s, AVIOContext *ioc,
281 AVStream *vstream, int64_t max_pos)
282{
283 FLVContext *flv = s->priv_data;
cb7e2c1c
KA
284 unsigned int arraylen = 0, timeslen = 0, fileposlen = 0, i;
285 double num_val;
286 char str_val[256];
e4529df9 287 int64_t *times = NULL;
cb7e2c1c 288 int64_t *filepositions = NULL;
e4529df9
DB
289 int ret = AVERROR(ENOSYS);
290 int64_t initial_pos = avio_tell(ioc);
cb7e2c1c 291
b70f04c2
MS
292 if (s->flags & AVFMT_FLAG_IGNIDX)
293 return 0;
294
e4529df9
DB
295 while (avio_tell(ioc) < max_pos - 2 &&
296 amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
297 int64_t *current_array;
cb7e2c1c
KA
298
299 // Expect array object in context
300 if (avio_r8(ioc) != AMF_DATA_TYPE_ARRAY)
301 break;
302
303 arraylen = avio_rb32(ioc);
a246cefa
MN
304 if (arraylen >> 28)
305 break;
306
e4529df9
DB
307 /* Expect only 'times' or 'filepositions' sub-arrays in other
308 * case refuse to use such metadata for indexing. */
cb7e2c1c
KA
309 if (!strcmp(KEYFRAMES_TIMESTAMP_TAG, str_val) && !times) {
310 if (!(times = av_mallocz(sizeof(*times) * arraylen))) {
311 ret = AVERROR(ENOMEM);
312 goto finish;
313 }
e4529df9 314 timeslen = arraylen;
cb7e2c1c 315 current_array = times;
e4529df9
DB
316 } else if (!strcmp(KEYFRAMES_BYTEOFFSET_TAG, str_val) &&
317 !filepositions) {
cb7e2c1c
KA
318 if (!(filepositions = av_mallocz(sizeof(*filepositions) * arraylen))) {
319 ret = AVERROR(ENOMEM);
320 goto finish;
321 }
e4529df9 322 fileposlen = arraylen;
cb7e2c1c 323 current_array = filepositions;
e4529df9
DB
324 } else
325 // unexpected metatag inside keyframes, will not use such
326 // metadata for indexing
cb7e2c1c
KA
327 break;
328
329 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
330 if (avio_r8(ioc) != AMF_DATA_TYPE_NUMBER)
331 goto finish;
e4529df9 332 num_val = av_int2double(avio_rb64(ioc));
cb7e2c1c
KA
333 current_array[i] = num_val;
334 }
578d6861
MS
335 if (times && filepositions) {
336 // All done, exiting at a position allowing amf_parse_object
337 // to finish parsing the object
338 ret = 0;
339 break;
340 }
cb7e2c1c
KA
341 }
342
7e297a46
MS
343 if (!ret && timeslen == fileposlen) {
344 for (i = 0; i < fileposlen; i++) {
e4529df9 345 av_add_index_entry(vstream, filepositions[i], times[i] * 1000,
0a7ce3ca 346 0, 0, AVINDEX_KEYFRAME);
7e297a46
MS
347 if (i < 2) {
348 flv->validate_index[i].pos = filepositions[i];
349 flv->validate_index[i].dts = times[i] * 1000;
e4529df9 350 flv->validate_count = i + 1;
7e297a46
MS
351 }
352 }
353 } else
cb7e2c1c
KA
354 av_log(s, AV_LOG_WARNING, "Invalid keyframes object, skipping.\n");
355
356finish:
357 av_freep(&times);
358 av_freep(&filepositions);
578d6861
MS
359 // If we got unexpected data, but successfully reset back to
360 // the start pos, the caller can continue parsing
361 if (ret < 0 && avio_seek(ioc, initial_pos, SEEK_SET) > 0)
362 return 0;
cb7e2c1c
KA
363 return ret;
364}
365
e4529df9
DB
366static int amf_parse_object(AVFormatContext *s, AVStream *astream,
367 AVStream *vstream, const char *key,
368 int64_t max_pos, int depth)
369{
428cc588 370 AVCodecContext *acodec, *vcodec;
5b54a90c 371 FLVContext *flv = s->priv_data;
ae628ec1 372 AVIOContext *ioc;
428cc588 373 AMFDataType amf_type;
cc38e063 374 char str_val[256];
428cc588
AH
375 double num_val;
376
e4529df9
DB
377 num_val = 0;
378 ioc = s->pb;
b7effd4e 379 amf_type = avio_r8(ioc);
428cc588 380
e4529df9
DB
381 switch (amf_type) {
382 case AMF_DATA_TYPE_NUMBER:
383 num_val = av_int2double(avio_rb64(ioc));
384 break;
385 case AMF_DATA_TYPE_BOOL:
386 num_val = avio_r8(ioc);
387 break;
388 case AMF_DATA_TYPE_STRING:
389 if (amf_get_string(ioc, str_val, sizeof(str_val)) < 0)
390 return -1;
391 break;
392 case AMF_DATA_TYPE_OBJECT:
393 if ((vstream || astream) && key &&
394 !strcmp(KEYFRAMES_TAG, key) && depth == 1)
395 if (parse_keyframes_index(s, ioc, vstream ? vstream : astream,
396 max_pos) < 0)
428cc588 397 return -1;
428cc588 398
e4529df9
DB
399 while (avio_tell(ioc) < max_pos - 2 &&
400 amf_get_string(ioc, str_val, sizeof(str_val)) > 0)
401 if (amf_parse_object(s, astream, vstream, str_val, max_pos,
402 depth + 1) < 0)
403 return -1; // if we couldn't skip, bomb out.
404 if (avio_r8(ioc) != AMF_END_OF_OBJECT)
428cc588 405 return -1;
e4529df9
DB
406 break;
407 case AMF_DATA_TYPE_NULL:
408 case AMF_DATA_TYPE_UNDEFINED:
409 case AMF_DATA_TYPE_UNSUPPORTED:
410 break; // these take up no additional space
411 case AMF_DATA_TYPE_MIXEDARRAY:
412 avio_skip(ioc, 4); // skip 32-bit max array index
413 while (avio_tell(ioc) < max_pos - 2 &&
414 amf_get_string(ioc, str_val, sizeof(str_val)) > 0)
415 // this is the only case in which we would want a nested
416 // parse to not skip over the object
417 if (amf_parse_object(s, astream, vstream, str_val, max_pos,
418 depth + 1) < 0)
419 return -1;
420 if (avio_r8(ioc) != AMF_END_OF_OBJECT)
421 return -1;
422 break;
423 case AMF_DATA_TYPE_ARRAY:
424 {
425 unsigned int arraylen, i;
426
427 arraylen = avio_rb32(ioc);
428 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++)
429 if (amf_parse_object(s, NULL, NULL, NULL, max_pos,
430 depth + 1) < 0)
431 return -1; // if we couldn't skip, bomb out.
432 }
433 break;
434 case AMF_DATA_TYPE_DATE:
435 avio_skip(ioc, 8 + 2); // timestamp (double) and UTC offset (int16)
436 break;
437 default: // unsupported type, we couldn't skip
438 return -1;
428cc588
AH
439 }
440
e4529df9
DB
441 // only look for metadata values when we are not nested and key != NULL
442 if (depth == 1 && key) {
428cc588
AH
443 acodec = astream ? astream->codec : NULL;
444 vcodec = vstream ? vstream->codec : NULL;
445
e4529df9
DB
446 if (amf_type == AMF_DATA_TYPE_NUMBER ||
447 amf_type == AMF_DATA_TYPE_BOOL) {
b204c46d
MS
448 if (!strcmp(key, "duration"))
449 s->duration = num_val * AV_TIME_BASE;
e4529df9
DB
450 else if (!strcmp(key, "videodatarate") && vcodec &&
451 0 <= (int)(num_val * 1024.0))
b204c46d 452 vcodec->bit_rate = num_val * 1024.0;
e4529df9
DB
453 else if (!strcmp(key, "audiodatarate") && acodec &&
454 0 <= (int)(num_val * 1024.0))
b204c46d 455 acodec->bit_rate = num_val * 1024.0;
21e2dc9f 456 else if (!strcmp(key, "datastream")) {
41f43202 457 AVStream *st = create_stream(s, AVMEDIA_TYPE_DATA);
21e2dc9f
LB
458 if (!st)
459 return AVERROR(ENOMEM);
36ef5369 460 st->codec->codec_id = AV_CODEC_ID_TEXT;
5b54a90c
LB
461 } else if (flv->trust_metadata) {
462 if (!strcmp(key, "videocodecid") && vcodec) {
c91c63b5 463 flv_set_video_codec(s, vstream, num_val, 0);
e4529df9 464 } else if (!strcmp(key, "audiocodecid") && acodec) {
c3d01577
JR
465 int id = ((int)num_val) << FLV_AUDIO_CODECID_OFFSET;
466 flv_set_audio_codec(s, astream, acodec, id);
e4529df9 467 } else if (!strcmp(key, "audiosamplerate") && acodec) {
5b54a90c 468 acodec->sample_rate = num_val;
e46a2a73
JR
469 } else if (!strcmp(key, "audiosamplesize") && acodec) {
470 acodec->bits_per_coded_sample = num_val;
471 } else if (!strcmp(key, "stereo") && acodec) {
e4529df9 472 acodec->channels = num_val + 1;
e46a2a73
JR
473 acodec->channel_layout = acodec->channels == 2 ?
474 AV_CH_LAYOUT_STEREO :
475 AV_CH_LAYOUT_MONO;
e4529df9 476 } else if (!strcmp(key, "width") && vcodec) {
5b54a90c 477 vcodec->width = num_val;
e4529df9 478 } else if (!strcmp(key, "height") && vcodec) {
5b54a90c
LB
479 vcodec->height = num_val;
480 }
21e2dc9f 481 }
b204c46d
MS
482 }
483
5e87222f
MS
484 if (!strcmp(key, "duration") ||
485 !strcmp(key, "filesize") ||
486 !strcmp(key, "width") ||
487 !strcmp(key, "height") ||
488 !strcmp(key, "videodatarate") ||
489 !strcmp(key, "framerate") ||
490 !strcmp(key, "videocodecid") ||
491 !strcmp(key, "audiodatarate") ||
492 !strcmp(key, "audiosamplerate") ||
493 !strcmp(key, "audiosamplesize") ||
494 !strcmp(key, "stereo") ||
0a9425d7
LB
495 !strcmp(key, "audiocodecid") ||
496 !strcmp(key, "datastream"))
5e87222f
MS
497 return 0;
498
e4529df9
DB
499 if (amf_type == AMF_DATA_TYPE_BOOL) {
500 av_strlcpy(str_val, num_val > 0 ? "true" : "false",
501 sizeof(str_val));
d2d67e42 502 av_dict_set(&s->metadata, key, str_val, 0);
e4529df9 503 } else if (amf_type == AMF_DATA_TYPE_NUMBER) {
cc38e063 504 snprintf(str_val, sizeof(str_val), "%.f", num_val);
d2d67e42 505 av_dict_set(&s->metadata, key, str_val, 0);
df2bd71a 506 } else if (amf_type == AMF_DATA_TYPE_STRING)
d2d67e42 507 av_dict_set(&s->metadata, key, str_val, 0);
428cc588
AH
508 }
509
510 return 0;
511}
512
e4529df9
DB
513static int flv_read_metabody(AVFormatContext *s, int64_t next_pos)
514{
428cc588
AH
515 AMFDataType type;
516 AVStream *stream, *astream, *vstream;
ae628ec1 517 AVIOContext *ioc;
4eec2606 518 int i;
e4529df9
DB
519 // only needs to hold the string "onMetaData".
520 // Anything longer is something we don't want.
521 char buffer[11];
428cc588
AH
522
523 astream = NULL;
524 vstream = NULL;
e4529df9 525 ioc = s->pb;
428cc588 526
e4529df9 527 // first object needs to be "onMetaData" string
b7effd4e 528 type = avio_r8(ioc);
21e2dc9f
LB
529 if (type != AMF_DATA_TYPE_STRING ||
530 amf_get_string(ioc, buffer, sizeof(buffer)) < 0)
531 return -1;
532
533 if (!strcmp(buffer, "onTextData"))
534 return 1;
535
536 if (strcmp(buffer, "onMetaData"))
428cc588
AH
537 return -1;
538
e4529df9
DB
539 // find the streams now so that amf_parse_object doesn't need to do
540 // the lookup every time it is called.
541 for (i = 0; i < s->nb_streams; i++) {
428cc588 542 stream = s->streams[i];
e4529df9
DB
543 if (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO)
544 astream = stream;
545 else if (stream->codec->codec_type == AVMEDIA_TYPE_VIDEO)
546 vstream = stream;
428cc588
AH
547 }
548
e4529df9
DB
549 // parse the second object (we want a mixed array)
550 if (amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
428cc588
AH
551 return -1;
552
553 return 0;
554}
555
6e9651d1 556static int flv_read_header(AVFormatContext *s)
d4f5d74a 557{
15f14fc7 558 int offset, flags;
d4f5d74a 559
45a8a02a 560 avio_skip(s->pb, 4);
b7effd4e 561 flags = avio_r8(s->pb);
d4f5d74a 562
ee0dadc1 563 s->ctx_flags |= AVFMTCTX_NOHEADER;
b41497e9 564
e4529df9
DB
565 if (flags & FLV_HEADER_FLAG_HASVIDEO)
566 if (!create_stream(s, AVMEDIA_TYPE_VIDEO))
769e10f0 567 return AVERROR(ENOMEM);
e4529df9
DB
568 if (flags & FLV_HEADER_FLAG_HASAUDIO)
569 if (!create_stream(s, AVMEDIA_TYPE_AUDIO))
769e10f0 570 return AVERROR(ENOMEM);
4eb0c665 571
b7effd4e 572 offset = avio_rb32(s->pb);
6b4aa5da 573 avio_seek(s->pb, offset, SEEK_SET);
45a8a02a 574 avio_skip(s->pb, 4);
d4f5d74a 575
aeb20f7f
N
576 s->start_time = 0;
577
d4f5d74a
GM
578 return 0;
579}
580
251f320f
MS
581static int flv_read_close(AVFormatContext *s)
582{
583 FLVContext *flv = s->priv_data;
584 av_freep(&flv->new_extradata[0]);
585 av_freep(&flv->new_extradata[1]);
586 return 0;
587}
588
04fd3e81
BC
589static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
590{
591 av_free(st->codec->extradata);
592 st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
593 if (!st->codec->extradata)
594 return AVERROR(ENOMEM);
595 st->codec->extradata_size = size;
b7effd4e 596 avio_read(s->pb, st->codec->extradata, st->codec->extradata_size);
04fd3e81
BC
597 return 0;
598}
599
251f320f
MS
600static int flv_queue_extradata(FLVContext *flv, AVIOContext *pb, int stream,
601 int size)
602{
603 av_free(flv->new_extradata[stream]);
e4529df9
DB
604 flv->new_extradata[stream] = av_mallocz(size +
605 FF_INPUT_BUFFER_PADDING_SIZE);
251f320f
MS
606 if (!flv->new_extradata[stream])
607 return AVERROR(ENOMEM);
608 flv->new_extradata_size[stream] = size;
609 avio_read(pb, flv->new_extradata[stream], size);
610 return 0;
611}
612
7e297a46
MS
613static void clear_index_entries(AVFormatContext *s, int64_t pos)
614{
615 int i, j, out;
e4529df9
DB
616 av_log(s, AV_LOG_WARNING,
617 "Found invalid index entries, clearing the index.\n");
7e297a46
MS
618 for (i = 0; i < s->nb_streams; i++) {
619 AVStream *st = s->streams[i];
620 /* Remove all index entries that point to >= pos */
621 out = 0;
e4529df9 622 for (j = 0; j < st->nb_index_entries; j++)
7e297a46
MS
623 if (st->index_entries[j].pos < pos)
624 st->index_entries[out++] = st->index_entries[j];
7e297a46
MS
625 st->nb_index_entries = out;
626 }
627}
628
c951e4b4
LB
629static int amf_skip_tag(AVIOContext *pb, AMFDataType type)
630{
631 int nb = -1, ret, parse_name = 1;
632
633 switch (type) {
634 case AMF_DATA_TYPE_NUMBER:
635 avio_skip(pb, 8);
636 break;
637 case AMF_DATA_TYPE_BOOL:
638 avio_skip(pb, 1);
639 break;
640 case AMF_DATA_TYPE_STRING:
641 avio_skip(pb, avio_rb16(pb));
642 break;
643 case AMF_DATA_TYPE_ARRAY:
644 parse_name = 0;
645 case AMF_DATA_TYPE_MIXEDARRAY:
646 nb = avio_rb32(pb);
647 case AMF_DATA_TYPE_OBJECT:
648 while(!pb->eof_reached && (nb-- > 0 || type != AMF_DATA_TYPE_ARRAY)) {
649 if (parse_name) {
650 int size = avio_rb16(pb);
651 if (!size) {
652 avio_skip(pb, 1);
653 break;
654 }
655 avio_skip(pb, size);
656 }
657 if ((ret = amf_skip_tag(pb, avio_r8(pb))) < 0)
658 return ret;
659 }
660 break;
661 case AMF_DATA_TYPE_NULL:
662 case AMF_DATA_TYPE_OBJECT_END:
663 break;
664 default:
665 return AVERROR_INVALIDDATA;
666 }
667 return 0;
668}
669
21e2dc9f
LB
670static int flv_data_packet(AVFormatContext *s, AVPacket *pkt,
671 int64_t dts, int64_t next)
672{
21e2dc9f 673 AVIOContext *pb = s->pb;
e4529df9 674 AVStream *st = NULL;
21e2dc9f 675 char buf[20];
c951e4b4
LB
676 int ret = AVERROR_INVALIDDATA;
677 int i, length = -1;
21e2dc9f 678
c951e4b4
LB
679 switch (avio_r8(pb)) {
680 case AMF_DATA_TYPE_MIXEDARRAY:
21e2dc9f 681 avio_seek(pb, 4, SEEK_CUR);
c951e4b4
LB
682 case AMF_DATA_TYPE_OBJECT:
683 break;
684 default:
685 goto skip;
686 }
21e2dc9f 687
c951e4b4
LB
688 while ((ret = amf_get_string(pb, buf, sizeof(buf))) > 0) {
689 AMFDataType type = avio_r8(pb);
690 if (type == AMF_DATA_TYPE_STRING && !strcmp(buf, "text")) {
691 length = avio_rb16(pb);
692 ret = av_get_packet(pb, pkt, length);
693 if (ret < 0)
694 goto skip;
695 else
696 break;
697 } else {
698 if ((ret = amf_skip_tag(pb, type)) < 0)
699 goto skip;
700 }
701 }
21e2dc9f 702
c951e4b4
LB
703 if (length < 0) {
704 ret = AVERROR_INVALIDDATA;
705 goto skip;
706 }
21e2dc9f
LB
707
708 for (i = 0; i < s->nb_streams; i++) {
709 st = s->streams[i];
09a445ce 710 if (st->codec->codec_type == AVMEDIA_TYPE_DATA)
21e2dc9f
LB
711 break;
712 }
713
714 if (i == s->nb_streams) {
41f43202 715 st = create_stream(s, AVMEDIA_TYPE_DATA);
21e2dc9f 716 if (!st)
f900f35a 717 return AVERROR_INVALIDDATA;
36ef5369 718 st->codec->codec_id = AV_CODEC_ID_TEXT;
21e2dc9f
LB
719 }
720
721 pkt->dts = dts;
722 pkt->pts = dts;
723 pkt->size = ret;
724
725 pkt->stream_index = st->index;
e4529df9 726 pkt->flags |= AV_PKT_FLAG_KEY;
21e2dc9f 727
c951e4b4 728skip:
21e2dc9f 729 avio_seek(s->pb, next + 4, SEEK_SET);
e4529df9 730
21e2dc9f
LB
731 return ret;
732}
733
d4f5d74a
GM
734static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
735{
ebd61055 736 FLVContext *flv = s->priv_data;
4fe8a452
PM
737 int ret, i, type, size, flags, is_audio;
738 int64_t next, pos;
ebd61055 739 int64_t dts, pts = AV_NOPTS_VALUE;
2df73eef 740 int sample_rate = 0, channels = 0;
e4529df9
DB
741 AVStream *st = NULL;
742
743 /* pkt size is repeated at end. skip it */
744 for (;; avio_skip(s->pb, 4)) {
745 pos = avio_tell(s->pb);
746 type = avio_r8(s->pb);
747 size = avio_rb24(s->pb);
748 dts = avio_rb24(s->pb);
749 dts |= avio_r8(s->pb) << 24;
750 av_dlog(s, "type:%d, size:%d, dts:%"PRId64"\n", type, size, dts);
751 if (s->pb->eof_reached)
752 return AVERROR_EOF;
753 avio_skip(s->pb, 3); /* stream id, always 0 */
754 flags = 0;
755
756 if (flv->validate_next < flv->validate_count) {
757 int64_t validate_pos = flv->validate_index[flv->validate_next].pos;
758 if (pos == validate_pos) {
759 if (FFABS(dts - flv->validate_index[flv->validate_next].dts) <=
760 VALIDATE_INDEX_TS_THRESH) {
761 flv->validate_next++;
762 } else {
763 clear_index_entries(s, validate_pos);
764 flv->validate_count = 0;
765 }
766 } else if (pos > validate_pos) {
7e297a46
MS
767 clear_index_entries(s, validate_pos);
768 flv->validate_count = 0;
769 }
7e297a46 770 }
115329f1 771
e4529df9
DB
772 if (size == 0)
773 continue;
774
775 next = size + avio_tell(s->pb);
776
777 if (type == FLV_TAG_TYPE_AUDIO) {
778 is_audio = 1;
779 flags = avio_r8(s->pb);
780 size--;
781 } else if (type == FLV_TAG_TYPE_VIDEO) {
782 is_audio = 0;
783 flags = avio_r8(s->pb);
784 size--;
785 if ((flags & 0xf0) == 0x50) /* video info / command frame */
786 goto skip;
787 } else {
788 if (type == FLV_TAG_TYPE_META && size > 13 + 1 + 4)
789 if (flv_read_metabody(s, next) > 0) {
790 return flv_data_packet(s, pkt, dts, next);
791 } else /* skip packet */
792 av_log(s, AV_LOG_DEBUG,
374fdc8c 793 "Skipping flv packet: type %d, size %d, flags %d.\n",
e4529df9
DB
794 type, size, flags);
795
796skip:
797 avio_seek(s->pb, next, SEEK_SET);
798 continue;
799 }
ae58b54b 800
e4529df9
DB
801 /* skip empty data packets */
802 if (!size)
803 continue;
804
805 /* now find stream */
806 for (i = 0; i < s->nb_streams; i++) {
807 st = s->streams[i];
808 if (is_audio && st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
809 if (flv_same_audio_codec(st->codec, flags))
810 break;
811 } else if (!is_audio &&
812 st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
813 if (flv_same_video_codec(st->codec, flags))
814 break;
09a445ce
LB
815 }
816 }
e4529df9
DB
817 if (i == s->nb_streams)
818 st = create_stream(s, is_audio ? AVMEDIA_TYPE_AUDIO
819 : AVMEDIA_TYPE_VIDEO);
820 av_dlog(s, "%d %X %d \n", is_audio, flags, st->discard);
fa14804c
LB
821
822 if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY ||
823 is_audio)
824 av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
825
e4529df9
DB
826 if ((st->discard >= AVDISCARD_NONKEY &&
827 !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY || is_audio)) ||
828 (st->discard >= AVDISCARD_BIDIR &&
829 ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && !is_audio)) ||
830 st->discard >= AVDISCARD_ALL) {
831 avio_seek(s->pb, next, SEEK_SET);
832 continue;
833 }
e4529df9 834 break;
d4f5d74a 835 }
068f2a22 836
e4529df9
DB
837 // if not streamed and no duration from metadata then seek to end to find
838 // the duration from the timestamps
839 if (s->pb->seekable && (!s->duration || s->duration == AV_NOPTS_VALUE)) {
15f14fc7 840 int size;
e4529df9
DB
841 const int64_t pos = avio_tell(s->pb);
842 const int64_t fsize = avio_size(s->pb);
843 avio_seek(s->pb, fsize - 4, SEEK_SET);
844 size = avio_rb32(s->pb);
845 avio_seek(s->pb, fsize - 3 - size, SEEK_SET);
846 if (size == avio_rb24(s->pb) + 11) {
b7effd4e 847 uint32_t ts = avio_rb24(s->pb);
e4529df9 848 ts |= avio_r8(s->pb) << 24;
b126dee9 849 s->duration = ts * (int64_t)AV_TIME_BASE / 1000;
15f14fc7 850 }
6b4aa5da 851 avio_seek(s->pb, pos, SEEK_SET);
15f14fc7
MN
852 }
853
e4529df9 854 if (is_audio) {
2215c39e 855 int bits_per_coded_sample;
e4529df9
DB
856 channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
857 sample_rate = 44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >>
858 FLV_AUDIO_SAMPLERATE_OFFSET) >> 3;
2215c39e 859 bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
e4529df9
DB
860 if (!st->codec->channels || !st->codec->sample_rate ||
861 !st->codec->bits_per_coded_sample) {
2215c39e 862 st->codec->channels = channels;
e4529df9
DB
863 st->codec->channel_layout = channels == 1
864 ? AV_CH_LAYOUT_MONO
865 : AV_CH_LAYOUT_STEREO;
2215c39e
MS
866 st->codec->sample_rate = sample_rate;
867 st->codec->bits_per_coded_sample = bits_per_coded_sample;
2f3d7ea9 868 }
e4529df9
DB
869 if (!st->codec->codec_id) {
870 flv_set_audio_codec(s, st, st->codec,
871 flags & FLV_AUDIO_CODECID_MASK);
872 flv->last_sample_rate =
873 sample_rate = st->codec->sample_rate;
874 flv->last_channels =
875 channels = st->codec->channels;
2215c39e
MS
876 } else {
877 AVCodecContext ctx;
878 ctx.sample_rate = sample_rate;
879 flv_set_audio_codec(s, st, &ctx, flags & FLV_AUDIO_CODECID_MASK);
880 sample_rate = ctx.sample_rate;
068f2a22 881 }
e4529df9 882 } else {
c91c63b5 883 size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK, 1);
bb01a3f0
MN
884 }
885
36ef5369
AK
886 if (st->codec->codec_id == AV_CODEC_ID_AAC ||
887 st->codec->codec_id == AV_CODEC_ID_H264) {
b7effd4e 888 int type = avio_r8(s->pb);
04fd3e81 889 size--;
36ef5369 890 if (st->codec->codec_id == AV_CODEC_ID_H264) {
e4529df9
DB
891 // sign extension
892 int32_t cts = (avio_rb24(s->pb) + 0xff800000) ^ 0xff800000;
ebd61055 893 pts = dts + cts;
5d983fdb 894 if (cts < 0 && !flv->wrong_dts) { // dts might be wrong
ebd61055 895 flv->wrong_dts = 1;
e4529df9 896 av_log(s, AV_LOG_WARNING,
374fdc8c 897 "Negative cts, previous timestamps might be wrong.\n");
ebd61055 898 }
04fd3e81
BC
899 }
900 if (type == 0) {
251f320f
MS
901 if (st->codec->extradata) {
902 if ((ret = flv_queue_extradata(flv, s->pb, is_audio, size)) < 0)
903 return ret;
904 ret = AVERROR(EAGAIN);
905 goto leave;
906 }
6298eb81 907 if ((ret = flv_get_extradata(s, st, size)) < 0)
04fd3e81 908 return ret;
36ef5369 909 if (st->codec->codec_id == AV_CODEC_ID_AAC) {
d2718187 910 MPEG4AudioConfig cfg;
547f8345
BL
911
912 /* Workaround for buggy Omnia A/XE encoder */
913 AVDictionaryEntry *t = av_dict_get(s->metadata, "Encoder", NULL, 0);
914 if (t && !strcmp(t->value, "Omnia A/XE"))
915 st->codec->extradata_size = 2;
916
59a9a235 917 avpriv_mpeg4audio_get_config(&cfg, st->codec->extradata,
fd095539 918 st->codec->extradata_size * 8, 1);
e4529df9 919 st->codec->channels = cfg.channels;
644d8d2e 920 st->codec->channel_layout = 0;
7d6096e4
BC
921 if (cfg.ext_sample_rate)
922 st->codec->sample_rate = cfg.ext_sample_rate;
923 else
924 st->codec->sample_rate = cfg.sample_rate;
dfd2a005 925 av_dlog(s, "mp4a config channels %d sample rate %d\n",
d2718187
BC
926 st->codec->channels, st->codec->sample_rate);
927 }
928
527c2e64
HC
929 ret = AVERROR(EAGAIN);
930 goto leave;
04fd3e81
BC
931 }
932 }
933
fcb4228c 934 /* skip empty data packets */
527c2e64
HC
935 if (!size) {
936 ret = AVERROR(EAGAIN);
937 goto leave;
938 }
fcb4228c 939
e4529df9
DB
940 ret = av_get_packet(s->pb, pkt, size);
941 if (ret < 0)
6f3e0b21 942 return AVERROR(EIO);
d4f5d74a 943 /* note: we need to modify the packet size here to handle the last
e4529df9
DB
944 * packet */
945 pkt->size = ret;
946 pkt->dts = dts;
947 pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
d4f5d74a 948 pkt->stream_index = st->index;
251f320f
MS
949 if (flv->new_extradata[is_audio]) {
950 uint8_t *side = av_packet_new_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA,
951 flv->new_extradata_size[is_audio]);
952 if (side) {
953 memcpy(side, flv->new_extradata[is_audio],
954 flv->new_extradata_size[is_audio]);
955 av_freep(&flv->new_extradata[is_audio]);
956 flv->new_extradata_size[is_audio] = 0;
957 }
958 }
2215c39e 959 if (is_audio && (sample_rate != flv->last_sample_rate ||
e4529df9 960 channels != flv->last_channels)) {
2215c39e
MS
961 flv->last_sample_rate = sample_rate;
962 flv->last_channels = channels;
963 ff_add_param_change(pkt, channels, 0, sample_rate, 0, 0);
964 }
115329f1 965
6cac3a3b 966 if (is_audio || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY))
cc947f04 967 pkt->flags |= AV_PKT_FLAG_KEY;
115329f1 968
527c2e64 969leave:
45a8a02a 970 avio_skip(s->pb, 4);
d4f5d74a
GM
971 return ret;
972}
973
fc8fa007 974static int flv_read_seek(AVFormatContext *s, int stream_index,
e4529df9 975 int64_t ts, int flags)
fc8fa007 976{
7e297a46
MS
977 FLVContext *flv = s->priv_data;
978 flv->validate_count = 0;
ff1ec0c3 979 return avio_seek_time(s->pb, stream_index, ts, flags);
fc8fa007
HC
980}
981
5b54a90c
LB
982#define OFFSET(x) offsetof(FLVContext, x)
983#define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM
984static const AVOption options[] = {
f4634ae8 985 { "flv_metadata", "Allocate streams according to the onMetaData array", OFFSET(trust_metadata), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VD },
5b54a90c
LB
986 { NULL }
987};
988
989static const AVClass class = {
990 .class_name = "flvdec",
991 .item_name = av_default_item_name,
992 .option = options,
993 .version = LIBAVUTIL_VERSION_INT,
994};
995
c6610a21 996AVInputFormat ff_flv_demuxer = {
dfc2c4d9 997 .name = "flv",
0177b7d2 998 .long_name = NULL_IF_CONFIG_SMALL("FLV (Flash Video)"),
dfc2c4d9
AK
999 .priv_data_size = sizeof(FLVContext),
1000 .read_probe = flv_probe,
1001 .read_header = flv_read_header,
1002 .read_packet = flv_read_packet,
20234a4b 1003 .read_seek = flv_read_seek,
20234a4b
MS
1004 .read_close = flv_read_close,
1005 .extensions = "flv",
5b54a90c 1006 .priv_class = &class,
d4f5d74a 1007};