3 * Copyright (c) 2003 The Libav Project
5 * This demuxer will generate a 1 byte extradata for VP6F content.
7 * - upper 4bits: difference between encoded width and visible width
8 * - lower 4bits: difference between encoded height and visible height
10 * This file is part of Libav.
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.
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.
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
27 #include "libavutil/avstring.h"
28 #include "libavcodec/bytestream.h"
29 #include "libavcodec/mpeg4audio.h"
31 #include "avio_internal.h"
34 #define KEYFRAMES_TAG "keyframes"
35 #define KEYFRAMES_TIMESTAMP_TAG "times"
36 #define KEYFRAMES_BYTEOFFSET_TAG "filepositions"
39 int wrong_dts; ///< wrong dts due to negative cts
42 static int flv_probe(AVProbeData *p)
47 if (d[0] == 'F' && d[1] == 'L' && d[2] == 'V' && d[3] < 5 && d[5]==0 && AV_RB32(d+5)>8) {
48 return AVPROBE_SCORE_MAX;
53 static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream, int flv_codecid) {
54 AVCodecContext *acodec = astream->codec;
56 //no distinction between S16 and S8 PCM codec flags
58 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 :
65 case FLV_CODECID_PCM_LE:
66 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 : CODEC_ID_PCM_S16LE; break;
67 case FLV_CODECID_AAC : acodec->codec_id = CODEC_ID_AAC; break;
68 case FLV_CODECID_ADPCM: acodec->codec_id = CODEC_ID_ADPCM_SWF; break;
69 case FLV_CODECID_SPEEX:
70 acodec->codec_id = CODEC_ID_SPEEX;
71 acodec->sample_rate = 16000;
73 case FLV_CODECID_MP3 : acodec->codec_id = CODEC_ID_MP3 ; astream->need_parsing = AVSTREAM_PARSE_FULL; break;
74 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
75 acodec->sample_rate = 8000; //in case metadata does not otherwise declare samplerate
76 acodec->codec_id = CODEC_ID_NELLYMOSER;
78 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
79 acodec->sample_rate = 16000;
80 acodec->codec_id = CODEC_ID_NELLYMOSER;
82 case FLV_CODECID_NELLYMOSER:
83 acodec->codec_id = CODEC_ID_NELLYMOSER;
86 av_log(s, AV_LOG_INFO, "Unsupported audio codec (%x)\n", flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
87 acodec->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
91 static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream, int flv_codecid) {
92 AVCodecContext *vcodec = vstream->codec;
94 case FLV_CODECID_H263 : vcodec->codec_id = CODEC_ID_FLV1 ; break;
95 case FLV_CODECID_SCREEN: vcodec->codec_id = CODEC_ID_FLASHSV; break;
96 case FLV_CODECID_SCREEN2: vcodec->codec_id = CODEC_ID_FLASHSV2; break;
97 case FLV_CODECID_VP6 : vcodec->codec_id = CODEC_ID_VP6F ;
98 case FLV_CODECID_VP6A :
99 if(flv_codecid == FLV_CODECID_VP6A)
100 vcodec->codec_id = CODEC_ID_VP6A;
101 if(vcodec->extradata_size != 1) {
102 vcodec->extradata_size = 1;
103 vcodec->extradata = av_malloc(1);
105 vcodec->extradata[0] = avio_r8(s->pb);
106 return 1; // 1 byte body size adjustment for flv_read_packet()
107 case FLV_CODECID_H264:
108 vcodec->codec_id = CODEC_ID_H264;
109 return 3; // not 4, reading packet type will consume one byte
111 av_log(s, AV_LOG_INFO, "Unsupported video codec (%x)\n", flv_codecid);
112 vcodec->codec_tag = flv_codecid;
118 static int amf_get_string(AVIOContext *ioc, char *buffer, int buffsize) {
119 int length = avio_rb16(ioc);
120 if(length >= buffsize) {
121 avio_skip(ioc, length);
125 avio_read(ioc, buffer, length);
127 buffer[length] = '\0';
132 static int parse_keyframes_index(AVFormatContext *s, AVIOContext *ioc, AVStream *vstream, int64_t max_pos) {
133 unsigned int arraylen = 0, timeslen = 0, fileposlen = 0, i;
136 int64_t *times = NULL;
137 int64_t *filepositions = NULL;
138 int ret = AVERROR(ENOSYS);
139 int64_t initial_pos = avio_tell(ioc);
141 while (avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
142 int64_t* current_array;
144 // Expect array object in context
145 if (avio_r8(ioc) != AMF_DATA_TYPE_ARRAY)
148 arraylen = avio_rb32(ioc);
150 * Expect only 'times' or 'filepositions' sub-arrays in other case refuse to use such metadata
153 if (!strcmp(KEYFRAMES_TIMESTAMP_TAG, str_val) && !times) {
154 if (!(times = av_mallocz(sizeof(*times) * arraylen))) {
155 ret = AVERROR(ENOMEM);
159 current_array = times;
160 } else if (!strcmp(KEYFRAMES_BYTEOFFSET_TAG, str_val) && !filepositions) {
161 if (!(filepositions = av_mallocz(sizeof(*filepositions) * arraylen))) {
162 ret = AVERROR(ENOMEM);
165 fileposlen = arraylen;
166 current_array = filepositions;
167 } else // unexpected metatag inside keyframes, will not use such metadata for indexing
170 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
171 if (avio_r8(ioc) != AMF_DATA_TYPE_NUMBER)
173 num_val = av_int2dbl(avio_rb64(ioc));
174 current_array[i] = num_val;
176 if (times && filepositions) {
177 // All done, exiting at a position allowing amf_parse_object
178 // to finish parsing the object
184 if (timeslen == fileposlen)
185 for(i = 0; i < arraylen; i++)
186 av_add_index_entry(vstream, filepositions[i], times[i]*1000, 0, 0, AVINDEX_KEYFRAME);
188 av_log(s, AV_LOG_WARNING, "Invalid keyframes object, skipping.\n");
192 av_freep(&filepositions);
193 // If we got unexpected data, but successfully reset back to
194 // the start pos, the caller can continue parsing
195 if (ret < 0 && avio_seek(ioc, initial_pos, SEEK_SET) > 0)
200 static int amf_parse_object(AVFormatContext *s, AVStream *astream, AVStream *vstream, const char *key, int64_t max_pos, int depth) {
201 AVCodecContext *acodec, *vcodec;
203 AMFDataType amf_type;
210 amf_type = avio_r8(ioc);
213 case AMF_DATA_TYPE_NUMBER:
214 num_val = av_int2dbl(avio_rb64(ioc)); break;
215 case AMF_DATA_TYPE_BOOL:
216 num_val = avio_r8(ioc); break;
217 case AMF_DATA_TYPE_STRING:
218 if(amf_get_string(ioc, str_val, sizeof(str_val)) < 0)
221 case AMF_DATA_TYPE_OBJECT: {
224 if (key && !strcmp(KEYFRAMES_TAG, key) && depth == 1)
225 if (parse_keyframes_index(s, ioc, vstream, max_pos) < 0)
228 while(avio_tell(ioc) < max_pos - 2 && (keylen = avio_rb16(ioc))) {
229 avio_skip(ioc, keylen); //skip key string
230 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
231 return -1; //if we couldn't skip, bomb out.
233 if(avio_r8(ioc) != AMF_END_OF_OBJECT)
237 case AMF_DATA_TYPE_NULL:
238 case AMF_DATA_TYPE_UNDEFINED:
239 case AMF_DATA_TYPE_UNSUPPORTED:
240 break; //these take up no additional space
241 case AMF_DATA_TYPE_MIXEDARRAY:
242 avio_skip(ioc, 4); //skip 32-bit max array index
243 while(avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
244 //this is the only case in which we would want a nested parse to not skip over the object
245 if(amf_parse_object(s, astream, vstream, str_val, max_pos, depth + 1) < 0)
248 if(avio_r8(ioc) != AMF_END_OF_OBJECT)
251 case AMF_DATA_TYPE_ARRAY: {
252 unsigned int arraylen, i;
254 arraylen = avio_rb32(ioc);
255 for(i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
256 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
257 return -1; //if we couldn't skip, bomb out.
261 case AMF_DATA_TYPE_DATE:
262 avio_skip(ioc, 8 + 2); //timestamp (double) and UTC offset (int16)
264 default: //unsupported type, we couldn't skip
268 if(depth == 1 && key) { //only look for metadata values when we are not nested and key != NULL
269 acodec = astream ? astream->codec : NULL;
270 vcodec = vstream ? vstream->codec : NULL;
272 if(amf_type == AMF_DATA_TYPE_BOOL) {
273 av_strlcpy(str_val, num_val > 0 ? "true" : "false", sizeof(str_val));
274 av_metadata_set2(&s->metadata, key, str_val, 0);
275 } else if(amf_type == AMF_DATA_TYPE_NUMBER) {
276 snprintf(str_val, sizeof(str_val), "%.f", num_val);
277 av_metadata_set2(&s->metadata, key, str_val, 0);
278 if(!strcmp(key, "duration")) s->duration = num_val * AV_TIME_BASE;
279 else if(!strcmp(key, "videodatarate") && vcodec && 0 <= (int)(num_val * 1024.0))
280 vcodec->bit_rate = num_val * 1024.0;
281 else if(!strcmp(key, "audiodatarate") && acodec && 0 <= (int)(num_val * 1024.0))
282 acodec->bit_rate = num_val * 1024.0;
283 } else if (amf_type == AMF_DATA_TYPE_STRING)
284 av_metadata_set2(&s->metadata, key, str_val, 0);
290 static int flv_read_metabody(AVFormatContext *s, int64_t next_pos) {
292 AVStream *stream, *astream, *vstream;
295 char buffer[11]; //only needs to hold the string "onMetaData". Anything longer is something we don't want.
301 //first object needs to be "onMetaData" string
303 if(type != AMF_DATA_TYPE_STRING || amf_get_string(ioc, buffer, sizeof(buffer)) < 0 || strcmp(buffer, "onMetaData"))
306 //find the streams now so that amf_parse_object doesn't need to do the lookup every time it is called.
307 for(i = 0; i < s->nb_streams; i++) {
308 stream = s->streams[i];
309 if (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO) astream = stream;
310 else if(stream->codec->codec_type == AVMEDIA_TYPE_VIDEO) vstream = stream;
313 //parse the second object (we want a mixed array)
314 if(amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
320 static AVStream *create_stream(AVFormatContext *s, int is_audio){
321 AVStream *st = av_new_stream(s, is_audio);
324 st->codec->codec_type = is_audio ? AVMEDIA_TYPE_AUDIO : AVMEDIA_TYPE_VIDEO;
325 av_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
329 static int flv_read_header(AVFormatContext *s,
330 AVFormatParameters *ap)
335 flags = avio_r8(s->pb);
336 /* old flvtool cleared this field */
337 /* FIXME: better fix needed */
339 flags = FLV_HEADER_FLAG_HASVIDEO | FLV_HEADER_FLAG_HASAUDIO;
340 av_log(s, AV_LOG_WARNING, "Broken FLV file, which says no streams present, this might fail\n");
343 if((flags & (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
344 != (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
345 s->ctx_flags |= AVFMTCTX_NOHEADER;
347 if(flags & FLV_HEADER_FLAG_HASVIDEO){
348 if(!create_stream(s, 0))
349 return AVERROR(ENOMEM);
351 if(flags & FLV_HEADER_FLAG_HASAUDIO){
352 if(!create_stream(s, 1))
353 return AVERROR(ENOMEM);
356 offset = avio_rb32(s->pb);
357 avio_seek(s->pb, offset, SEEK_SET);
365 static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
367 av_free(st->codec->extradata);
368 st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
369 if (!st->codec->extradata)
370 return AVERROR(ENOMEM);
371 st->codec->extradata_size = size;
372 avio_read(s->pb, st->codec->extradata, st->codec->extradata_size);
376 static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
378 FLVContext *flv = s->priv_data;
379 int ret, i, type, size, flags, is_audio;
381 int64_t dts, pts = AV_NOPTS_VALUE;
384 for(;;avio_skip(s->pb, 4)){ /* pkt size is repeated at end. skip it */
385 pos = avio_tell(s->pb);
386 type = avio_r8(s->pb);
387 size = avio_rb24(s->pb);
388 dts = avio_rb24(s->pb);
389 dts |= avio_r8(s->pb) << 24;
390 // av_log(s, AV_LOG_DEBUG, "type:%d, size:%d, dts:%d\n", type, size, dts);
391 if (s->pb->eof_reached)
393 avio_skip(s->pb, 3); /* stream id, always 0 */
399 next= size + avio_tell(s->pb);
401 if (type == FLV_TAG_TYPE_AUDIO) {
403 flags = avio_r8(s->pb);
405 } else if (type == FLV_TAG_TYPE_VIDEO) {
407 flags = avio_r8(s->pb);
409 if ((flags & 0xf0) == 0x50) /* video info / command frame */
412 if (type == FLV_TAG_TYPE_META && size > 13+1+4)
413 flv_read_metabody(s, next);
414 else /* skip packet */
415 av_log(s, AV_LOG_DEBUG, "skipping flv packet: type %d, size %d, flags %d\n", type, size, flags);
417 avio_seek(s->pb, next, SEEK_SET);
421 /* skip empty data packets */
425 /* now find stream */
426 for(i=0;i<s->nb_streams;i++) {
428 if (st->id == is_audio)
431 if(i == s->nb_streams){
432 av_log(s, AV_LOG_ERROR, "invalid stream\n");
433 st= create_stream(s, is_audio);
434 s->ctx_flags &= ~AVFMTCTX_NOHEADER;
436 // av_log(s, AV_LOG_DEBUG, "%d %X %d \n", is_audio, flags, st->discard);
437 if( (st->discard >= AVDISCARD_NONKEY && !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY || is_audio))
438 ||(st->discard >= AVDISCARD_BIDIR && ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && !is_audio))
439 || st->discard >= AVDISCARD_ALL
441 avio_seek(s->pb, next, SEEK_SET);
444 if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY)
445 av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
449 // if not streamed and no duration from metadata then seek to end to find the duration from the timestamps
450 if(s->pb->seekable && (!s->duration || s->duration==AV_NOPTS_VALUE)){
452 const int64_t pos= avio_tell(s->pb);
453 const int64_t fsize= avio_size(s->pb);
454 avio_seek(s->pb, fsize-4, SEEK_SET);
455 size= avio_rb32(s->pb);
456 avio_seek(s->pb, fsize-3-size, SEEK_SET);
457 if(size == avio_rb24(s->pb) + 11){
458 uint32_t ts = avio_rb24(s->pb);
459 ts |= avio_r8(s->pb) << 24;
460 s->duration = ts * (int64_t)AV_TIME_BASE / 1000;
462 avio_seek(s->pb, pos, SEEK_SET);
466 if(!st->codec->channels || !st->codec->sample_rate || !st->codec->bits_per_coded_sample) {
467 st->codec->channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
468 st->codec->sample_rate = (44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3);
469 st->codec->bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
471 if(!st->codec->codec_id){
472 flv_set_audio_codec(s, st, flags & FLV_AUDIO_CODECID_MASK);
475 size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK);
478 if (st->codec->codec_id == CODEC_ID_AAC ||
479 st->codec->codec_id == CODEC_ID_H264) {
480 int type = avio_r8(s->pb);
482 if (st->codec->codec_id == CODEC_ID_H264) {
483 int32_t cts = (avio_rb24(s->pb)+0xff800000)^0xff800000; // sign extension
485 if (cts < 0) { // dts are wrong
487 av_log(s, AV_LOG_WARNING, "negative cts, previous timestamps might be wrong\n");
490 dts = AV_NOPTS_VALUE;
493 if ((ret = flv_get_extradata(s, st, size)) < 0)
495 if (st->codec->codec_id == CODEC_ID_AAC) {
496 MPEG4AudioConfig cfg;
497 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
498 st->codec->extradata_size);
499 st->codec->channels = cfg.channels;
500 if (cfg.ext_sample_rate)
501 st->codec->sample_rate = cfg.ext_sample_rate;
503 st->codec->sample_rate = cfg.sample_rate;
504 av_dlog(s, "mp4a config channels %d sample rate %d\n",
505 st->codec->channels, st->codec->sample_rate);
508 ret = AVERROR(EAGAIN);
513 /* skip empty data packets */
515 ret = AVERROR(EAGAIN);
519 ret= av_get_packet(s->pb, pkt, size);
523 /* note: we need to modify the packet size here to handle the last
527 pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
528 pkt->stream_index = st->index;
530 if (is_audio || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY))
531 pkt->flags |= AV_PKT_FLAG_KEY;
538 static int flv_read_seek(AVFormatContext *s, int stream_index,
539 int64_t ts, int flags)
541 return avio_seek_time(s->pb, stream_index, ts, flags);
544 #if 0 /* don't know enough to implement this */
545 static int flv_read_seek2(AVFormatContext *s, int stream_index,
546 int64_t min_ts, int64_t ts, int64_t max_ts, int flags)
548 int ret = AVERROR(ENOSYS);
550 if (ts - min_ts > (uint64_t)(max_ts - ts)) flags |= AVSEEK_FLAG_BACKWARD;
552 if (!s->pb->seekable) {
553 if (stream_index < 0) {
554 stream_index = av_find_default_stream_index(s);
555 if (stream_index < 0)
558 /* timestamp for default must be expressed in AV_TIME_BASE units */
559 ts = av_rescale_rnd(ts, 1000, AV_TIME_BASE,
560 flags & AVSEEK_FLAG_BACKWARD ? AV_ROUND_DOWN : AV_ROUND_UP);
562 ret = avio_seek_time(s->pb, stream_index, ts, flags);
565 if (ret == AVERROR(ENOSYS))
566 ret = av_seek_frame(s, stream_index, ts, flags);
571 AVInputFormat ff_flv_demuxer = {
573 NULL_IF_CONFIG_SMALL("FLV format"),
578 .read_seek = flv_read_seek,
580 .read_seek2 = flv_read_seek2,
583 .value = CODEC_ID_FLV1,