flvdec: expose metadata through the generic metadata API
[libav.git] / libavformat / flvdec.c
1 /*
2 * FLV demuxer
3 * Copyright (c) 2003 The FFmpeg 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 FFmpeg.
11 *
12 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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 "libavcodec/bytestream.h"
29 #include "libavcodec/mpeg4audio.h"
30 #include "avformat.h"
31 #include "flv.h"
32
33 typedef struct {
34 int wrong_dts; ///< wrong dts due to negative cts
35 } FLVContext;
36
37 static int flv_probe(AVProbeData *p)
38 {
39 const uint8_t *d;
40
41 d = p->buf;
42 if (d[0] == 'F' && d[1] == 'L' && d[2] == 'V' && d[3] < 5 && d[5]==0) {
43 return AVPROBE_SCORE_MAX;
44 }
45 return 0;
46 }
47
48 /**
49 * Builds a Speex header.
50 * This is not needed for the libavcodec libspeex decoder, but is needed for
51 * stream copy and for decoders which require a header.
52 */
53 static void flv_build_speex_header(uint8_t *extradata)
54 {
55 memset(extradata, 0, 80);
56 bytestream_put_buffer(&extradata, "Speex ", 8); // speex_string
57 bytestream_put_buffer(&extradata, "1.2rc1", 6); // speex_version
58 extradata += 14; // speex_version padding
59 bytestream_put_le32(&extradata, 1); // speex_version_id
60 bytestream_put_le32(&extradata, 80); // header_size
61 bytestream_put_le32(&extradata, 16000); // rate
62 bytestream_put_le32(&extradata, 1); // mode
63 bytestream_put_le32(&extradata, 4); // mode_bitstream_version
64 bytestream_put_le32(&extradata, 1); // nb_channels
65 bytestream_put_le32(&extradata, -1); // bitrate
66 bytestream_put_le32(&extradata, 320); // frame_size
67 // vbr = 0
68 // frames_per_packet = 0
69 // extra_headers = 0
70 // reserved1 = 0
71 // reserved2 = 0
72 }
73
74 static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream, int flv_codecid) {
75 AVCodecContext *acodec = astream->codec;
76 switch(flv_codecid) {
77 //no distinction between S16 and S8 PCM codec flags
78 case FLV_CODECID_PCM:
79 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_S8 :
80 #ifdef WORDS_BIGENDIAN
81 CODEC_ID_PCM_S16BE;
82 #else
83 CODEC_ID_PCM_S16LE;
84 #endif
85 break;
86 case FLV_CODECID_PCM_LE:
87 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_S8 : CODEC_ID_PCM_S16LE; break;
88 case FLV_CODECID_AAC : acodec->codec_id = CODEC_ID_AAC; break;
89 case FLV_CODECID_ADPCM: acodec->codec_id = CODEC_ID_ADPCM_SWF; break;
90 case FLV_CODECID_SPEEX:
91 acodec->codec_id = CODEC_ID_SPEEX;
92 acodec->sample_rate = 16000;
93 acodec->extradata = av_mallocz(80 + FF_INPUT_BUFFER_PADDING_SIZE);
94 if (acodec->extradata) {
95 acodec->extradata_size = 80;
96 flv_build_speex_header(acodec->extradata);
97 } else {
98 av_log(s, AV_LOG_WARNING, "Unable to create Speex extradata\n");
99 }
100 break;
101 case FLV_CODECID_MP3 : acodec->codec_id = CODEC_ID_MP3 ; astream->need_parsing = AVSTREAM_PARSE_FULL; break;
102 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
103 acodec->sample_rate = 8000; //in case metadata does not otherwise declare samplerate
104 case FLV_CODECID_NELLYMOSER:
105 acodec->codec_id = CODEC_ID_NELLYMOSER;
106 break;
107 default:
108 av_log(s, AV_LOG_INFO, "Unsupported audio codec (%x)\n", flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
109 acodec->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
110 }
111 }
112
113 static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream, int flv_codecid) {
114 AVCodecContext *vcodec = vstream->codec;
115 switch(flv_codecid) {
116 case FLV_CODECID_H263 : vcodec->codec_id = CODEC_ID_FLV1 ; break;
117 case FLV_CODECID_SCREEN: vcodec->codec_id = CODEC_ID_FLASHSV; break;
118 case FLV_CODECID_VP6 : vcodec->codec_id = CODEC_ID_VP6F ;
119 case FLV_CODECID_VP6A :
120 if(flv_codecid == FLV_CODECID_VP6A)
121 vcodec->codec_id = CODEC_ID_VP6A;
122 if(vcodec->extradata_size != 1) {
123 vcodec->extradata_size = 1;
124 vcodec->extradata = av_malloc(1);
125 }
126 vcodec->extradata[0] = get_byte(s->pb);
127 return 1; // 1 byte body size adjustment for flv_read_packet()
128 case FLV_CODECID_H264:
129 vcodec->codec_id = CODEC_ID_H264;
130 return 3; // not 4, reading packet type will consume one byte
131 default:
132 av_log(s, AV_LOG_INFO, "Unsupported video codec (%x)\n", flv_codecid);
133 vcodec->codec_tag = flv_codecid;
134 }
135
136 return 0;
137 }
138
139 static int amf_get_string(ByteIOContext *ioc, char *buffer, int buffsize) {
140 int length = get_be16(ioc);
141 if(length >= buffsize) {
142 url_fskip(ioc, length);
143 return -1;
144 }
145
146 get_buffer(ioc, buffer, length);
147
148 buffer[length] = '\0';
149
150 return length;
151 }
152
153 static int amf_parse_object(AVFormatContext *s, AVStream *astream, AVStream *vstream, const char *key, int64_t max_pos, int depth) {
154 AVCodecContext *acodec, *vcodec;
155 ByteIOContext *ioc;
156 AMFDataType amf_type;
157 char str_val[256];
158 double num_val;
159
160 num_val = 0;
161 ioc = s->pb;
162
163 amf_type = get_byte(ioc);
164
165 switch(amf_type) {
166 case AMF_DATA_TYPE_NUMBER:
167 num_val = av_int2dbl(get_be64(ioc)); break;
168 case AMF_DATA_TYPE_BOOL:
169 num_val = get_byte(ioc); break;
170 case AMF_DATA_TYPE_STRING:
171 if(amf_get_string(ioc, str_val, sizeof(str_val)) < 0)
172 return -1;
173 break;
174 case AMF_DATA_TYPE_OBJECT: {
175 unsigned int keylen;
176
177 while(url_ftell(ioc) < max_pos - 2 && (keylen = get_be16(ioc))) {
178 url_fskip(ioc, keylen); //skip key string
179 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
180 return -1; //if we couldn't skip, bomb out.
181 }
182 if(get_byte(ioc) != AMF_END_OF_OBJECT)
183 return -1;
184 }
185 break;
186 case AMF_DATA_TYPE_NULL:
187 case AMF_DATA_TYPE_UNDEFINED:
188 case AMF_DATA_TYPE_UNSUPPORTED:
189 break; //these take up no additional space
190 case AMF_DATA_TYPE_MIXEDARRAY:
191 url_fskip(ioc, 4); //skip 32-bit max array index
192 while(url_ftell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
193 //this is the only case in which we would want a nested parse to not skip over the object
194 if(amf_parse_object(s, astream, vstream, str_val, max_pos, depth + 1) < 0)
195 return -1;
196 }
197 if(get_byte(ioc) != AMF_END_OF_OBJECT)
198 return -1;
199 break;
200 case AMF_DATA_TYPE_ARRAY: {
201 unsigned int arraylen, i;
202
203 arraylen = get_be32(ioc);
204 for(i = 0; i < arraylen && url_ftell(ioc) < max_pos - 1; i++) {
205 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
206 return -1; //if we couldn't skip, bomb out.
207 }
208 }
209 break;
210 case AMF_DATA_TYPE_DATE:
211 url_fskip(ioc, 8 + 2); //timestamp (double) and UTC offset (int16)
212 break;
213 default: //unsupported type, we couldn't skip
214 return -1;
215 }
216
217 if(depth == 1 && key) { //only look for metadata values when we are not nested and key != NULL
218 acodec = astream ? astream->codec : NULL;
219 vcodec = vstream ? vstream->codec : NULL;
220
221 if(amf_type == AMF_DATA_TYPE_BOOL) {
222 av_strlcpy(str_val, num_val > 0 ? "true" : "false", sizeof(str_val));
223 av_metadata_set(&s->metadata, key, str_val);
224 } else if(amf_type == AMF_DATA_TYPE_NUMBER) {
225 snprintf(str_val, sizeof(str_val), "%.f", num_val);
226 av_metadata_set(&s->metadata, key, str_val);
227 if(!strcmp(key, "duration")) s->duration = num_val * AV_TIME_BASE;
228 else if(!strcmp(key, "videodatarate") && vcodec && 0 <= (int)(num_val * 1024.0))
229 vcodec->bit_rate = num_val * 1024.0;
230 } else if (amf_type == AMF_DATA_TYPE_STRING)
231 av_metadata_set(&s->metadata, key, str_val);
232 }
233
234 return 0;
235 }
236
237 static int flv_read_metabody(AVFormatContext *s, int64_t next_pos) {
238 AMFDataType type;
239 AVStream *stream, *astream, *vstream;
240 ByteIOContext *ioc;
241 int i;
242 char buffer[11]; //only needs to hold the string "onMetaData". Anything longer is something we don't want.
243
244 astream = NULL;
245 vstream = NULL;
246 ioc = s->pb;
247
248 //first object needs to be "onMetaData" string
249 type = get_byte(ioc);
250 if(type != AMF_DATA_TYPE_STRING || amf_get_string(ioc, buffer, sizeof(buffer)) < 0 || strcmp(buffer, "onMetaData"))
251 return -1;
252
253 //find the streams now so that amf_parse_object doesn't need to do the lookup every time it is called.
254 for(i = 0; i < s->nb_streams; i++) {
255 stream = s->streams[i];
256 if (stream->codec->codec_type == CODEC_TYPE_AUDIO) astream = stream;
257 else if(stream->codec->codec_type == CODEC_TYPE_VIDEO) vstream = stream;
258 }
259
260 //parse the second object (we want a mixed array)
261 if(amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
262 return -1;
263
264 return 0;
265 }
266
267 static AVStream *create_stream(AVFormatContext *s, int is_audio){
268 AVStream *st = av_new_stream(s, is_audio);
269 if (!st)
270 return NULL;
271 st->codec->codec_type = is_audio ? CODEC_TYPE_AUDIO : CODEC_TYPE_VIDEO;
272 av_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
273 return st;
274 }
275
276 static int flv_read_header(AVFormatContext *s,
277 AVFormatParameters *ap)
278 {
279 int offset, flags;
280
281 url_fskip(s->pb, 4);
282 flags = get_byte(s->pb);
283 /* old flvtool cleared this field */
284 /* FIXME: better fix needed */
285 if (!flags) {
286 flags = FLV_HEADER_FLAG_HASVIDEO | FLV_HEADER_FLAG_HASAUDIO;
287 av_log(s, AV_LOG_WARNING, "Broken FLV file, which says no streams present, this might fail\n");
288 }
289
290 if((flags & (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
291 != (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
292 s->ctx_flags |= AVFMTCTX_NOHEADER;
293
294 if(flags & FLV_HEADER_FLAG_HASVIDEO){
295 if(!create_stream(s, 0))
296 return AVERROR(ENOMEM);
297 }
298 if(flags & FLV_HEADER_FLAG_HASAUDIO){
299 if(!create_stream(s, 1))
300 return AVERROR(ENOMEM);
301 }
302
303 offset = get_be32(s->pb);
304 url_fseek(s->pb, offset, SEEK_SET);
305
306 s->start_time = 0;
307
308 return 0;
309 }
310
311 static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
312 {
313 av_free(st->codec->extradata);
314 st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
315 if (!st->codec->extradata)
316 return AVERROR(ENOMEM);
317 st->codec->extradata_size = size;
318 get_buffer(s->pb, st->codec->extradata, st->codec->extradata_size);
319 return 0;
320 }
321
322 static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
323 {
324 FLVContext *flv = s->priv_data;
325 int ret, i, type, size, flags, is_audio;
326 int64_t next, pos;
327 int64_t dts, pts = AV_NOPTS_VALUE;
328 AVStream *st = NULL;
329
330 for(;;){
331 pos = url_ftell(s->pb);
332 url_fskip(s->pb, 4); /* size of previous packet */
333 type = get_byte(s->pb);
334 size = get_be24(s->pb);
335 dts = get_be24(s->pb);
336 dts |= get_byte(s->pb) << 24;
337 // av_log(s, AV_LOG_DEBUG, "type:%d, size:%d, dts:%d\n", type, size, dts);
338 if (url_feof(s->pb))
339 return AVERROR_EOF;
340 url_fskip(s->pb, 3); /* stream id, always 0 */
341 flags = 0;
342
343 if(size == 0)
344 continue;
345
346 next= size + url_ftell(s->pb);
347
348 if (type == FLV_TAG_TYPE_AUDIO) {
349 is_audio=1;
350 flags = get_byte(s->pb);
351 size--;
352 } else if (type == FLV_TAG_TYPE_VIDEO) {
353 is_audio=0;
354 flags = get_byte(s->pb);
355 size--;
356 if ((flags & 0xf0) == 0x50) /* video info / command frame */
357 goto skip;
358 } else {
359 if (type == FLV_TAG_TYPE_META && size > 13+1+4)
360 flv_read_metabody(s, next);
361 else /* skip packet */
362 av_log(s, AV_LOG_DEBUG, "skipping flv packet: type %d, size %d, flags %d\n", type, size, flags);
363 skip:
364 url_fseek(s->pb, next, SEEK_SET);
365 continue;
366 }
367
368 /* skip empty data packets */
369 if (!size)
370 continue;
371
372 /* now find stream */
373 for(i=0;i<s->nb_streams;i++) {
374 st = s->streams[i];
375 if (st->id == is_audio)
376 break;
377 }
378 if(i == s->nb_streams){
379 av_log(s, AV_LOG_ERROR, "invalid stream\n");
380 st= create_stream(s, is_audio);
381 s->ctx_flags &= ~AVFMTCTX_NOHEADER;
382 }
383 // av_log(s, AV_LOG_DEBUG, "%d %X %d \n", is_audio, flags, st->discard);
384 if( (st->discard >= AVDISCARD_NONKEY && !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY || is_audio))
385 ||(st->discard >= AVDISCARD_BIDIR && ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && !is_audio))
386 || st->discard >= AVDISCARD_ALL
387 ){
388 url_fseek(s->pb, next, SEEK_SET);
389 continue;
390 }
391 if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY)
392 av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
393 break;
394 }
395
396 // if not streamed and no duration from metadata then seek to end to find the duration from the timestamps
397 if(!url_is_streamed(s->pb) && s->duration==AV_NOPTS_VALUE){
398 int size;
399 const int64_t pos= url_ftell(s->pb);
400 const int64_t fsize= url_fsize(s->pb);
401 url_fseek(s->pb, fsize-4, SEEK_SET);
402 size= get_be32(s->pb);
403 url_fseek(s->pb, fsize-3-size, SEEK_SET);
404 if(size == get_be24(s->pb) + 11){
405 s->duration= get_be24(s->pb) * (int64_t)AV_TIME_BASE / 1000;
406 }
407 url_fseek(s->pb, pos, SEEK_SET);
408 }
409
410 if(is_audio){
411 if(!st->codec->channels || !st->codec->sample_rate || !st->codec->bits_per_coded_sample) {
412 st->codec->channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
413 st->codec->sample_rate = (44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3);
414 st->codec->bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
415 }
416 if(!st->codec->codec_id){
417 flv_set_audio_codec(s, st, flags & FLV_AUDIO_CODECID_MASK);
418 }
419 }else{
420 size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK);
421 }
422
423 if (st->codec->codec_id == CODEC_ID_AAC ||
424 st->codec->codec_id == CODEC_ID_H264) {
425 int type = get_byte(s->pb);
426 size--;
427 if (st->codec->codec_id == CODEC_ID_H264) {
428 int32_t cts = (get_be24(s->pb)+0xff800000)^0xff800000; // sign extension
429 pts = dts + cts;
430 if (cts < 0) { // dts are wrong
431 flv->wrong_dts = 1;
432 av_log(s, AV_LOG_WARNING, "negative cts, previous timestamps might be wrong\n");
433 }
434 if (flv->wrong_dts)
435 dts = AV_NOPTS_VALUE;
436 }
437 if (type == 0) {
438 if ((ret = flv_get_extradata(s, st, size)) < 0)
439 return ret;
440 if (st->codec->codec_id == CODEC_ID_AAC) {
441 MPEG4AudioConfig cfg;
442 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
443 st->codec->extradata_size);
444 if (cfg.chan_config > 7)
445 return -1;
446 st->codec->channels = ff_mpeg4audio_channels[cfg.chan_config];
447 st->codec->sample_rate = cfg.sample_rate;
448 dprintf(s, "mp4a config channels %d sample rate %d\n",
449 st->codec->channels, st->codec->sample_rate);
450 }
451
452 return AVERROR(EAGAIN);
453 }
454 }
455
456 /* skip empty data packets */
457 if (!size)
458 return AVERROR(EAGAIN);
459
460 ret= av_get_packet(s->pb, pkt, size);
461 if (ret < 0) {
462 return AVERROR(EIO);
463 }
464 /* note: we need to modify the packet size here to handle the last
465 packet */
466 pkt->size = ret;
467 pkt->dts = dts;
468 pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
469 pkt->stream_index = st->index;
470
471 if (is_audio || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY))
472 pkt->flags |= PKT_FLAG_KEY;
473
474 return ret;
475 }
476
477 AVInputFormat flv_demuxer = {
478 "flv",
479 NULL_IF_CONFIG_SMALL("FLV format"),
480 sizeof(FLVContext),
481 flv_probe,
482 flv_read_header,
483 flv_read_packet,
484 .extensions = "flv",
485 .value = CODEC_ID_FLV1,
486 };