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 "libavutil/dict.h"
29 #include "libavutil/intfloat_readwrite.h"
30 #include "libavutil/mathematics.h"
31 #include "libavcodec/bytestream.h"
32 #include "libavcodec/mpeg4audio.h"
35 #include "avio_internal.h"
38 #define KEYFRAMES_TAG "keyframes"
39 #define KEYFRAMES_TIMESTAMP_TAG "times"
40 #define KEYFRAMES_BYTEOFFSET_TAG "filepositions"
43 int wrong_dts; ///< wrong dts due to negative cts
46 static int flv_probe(AVProbeData *p)
51 if (d[0] == 'F' && d[1] == 'L' && d[2] == 'V' && d[3] < 5 && d[5]==0 && AV_RB32(d+5)>8) {
52 return AVPROBE_SCORE_MAX;
57 static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream, int flv_codecid) {
58 AVCodecContext *acodec = astream->codec;
60 //no distinction between S16 and S8 PCM codec flags
62 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 :
69 case FLV_CODECID_PCM_LE:
70 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 : CODEC_ID_PCM_S16LE; break;
71 case FLV_CODECID_AAC : acodec->codec_id = CODEC_ID_AAC; break;
72 case FLV_CODECID_ADPCM: acodec->codec_id = CODEC_ID_ADPCM_SWF; break;
73 case FLV_CODECID_SPEEX:
74 acodec->codec_id = CODEC_ID_SPEEX;
75 acodec->sample_rate = 16000;
77 case FLV_CODECID_MP3 : acodec->codec_id = CODEC_ID_MP3 ; astream->need_parsing = AVSTREAM_PARSE_FULL; break;
78 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
79 acodec->sample_rate = 8000; //in case metadata does not otherwise declare samplerate
80 acodec->codec_id = CODEC_ID_NELLYMOSER;
82 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
83 acodec->sample_rate = 16000;
84 acodec->codec_id = CODEC_ID_NELLYMOSER;
86 case FLV_CODECID_NELLYMOSER:
87 acodec->codec_id = CODEC_ID_NELLYMOSER;
90 av_log(s, AV_LOG_INFO, "Unsupported audio codec (%x)\n", flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
91 acodec->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
95 static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream, int flv_codecid) {
96 AVCodecContext *vcodec = vstream->codec;
98 case FLV_CODECID_H263 : vcodec->codec_id = CODEC_ID_FLV1 ; break;
99 case FLV_CODECID_SCREEN: vcodec->codec_id = CODEC_ID_FLASHSV; break;
100 case FLV_CODECID_SCREEN2: vcodec->codec_id = CODEC_ID_FLASHSV2; break;
101 case FLV_CODECID_VP6 : vcodec->codec_id = CODEC_ID_VP6F ;
102 case FLV_CODECID_VP6A :
103 if(flv_codecid == FLV_CODECID_VP6A)
104 vcodec->codec_id = CODEC_ID_VP6A;
105 if(vcodec->extradata_size != 1) {
106 vcodec->extradata_size = 1;
107 vcodec->extradata = av_malloc(1);
109 vcodec->extradata[0] = avio_r8(s->pb);
110 return 1; // 1 byte body size adjustment for flv_read_packet()
111 case FLV_CODECID_H264:
112 vcodec->codec_id = CODEC_ID_H264;
113 return 3; // not 4, reading packet type will consume one byte
115 av_log(s, AV_LOG_INFO, "Unsupported video codec (%x)\n", flv_codecid);
116 vcodec->codec_tag = flv_codecid;
122 static int amf_get_string(AVIOContext *ioc, char *buffer, int buffsize) {
123 int length = avio_rb16(ioc);
124 if(length >= buffsize) {
125 avio_skip(ioc, length);
129 avio_read(ioc, buffer, length);
131 buffer[length] = '\0';
136 static int parse_keyframes_index(AVFormatContext *s, AVIOContext *ioc, AVStream *vstream, int64_t max_pos) {
137 unsigned int arraylen = 0, timeslen = 0, fileposlen = 0, i;
140 int64_t *times = NULL;
141 int64_t *filepositions = NULL;
142 int ret = AVERROR(ENOSYS);
143 int64_t initial_pos = avio_tell(ioc);
144 AVDictionaryEntry *creator = av_dict_get(s->metadata, "metadatacreator",
147 if (creator && !strcmp(creator->value, "MEGA")) {
148 /* Files with this metadatacreator tag seem to have filepositions
149 * pointing at the 4 trailer bytes of the previous packet,
150 * which isn't the norm (nor what we expect here, nor what
151 * jwplayer + lighttpd expect, nor what flvtool2 produces).
152 * Just ignore the index in this case, instead of risking trying
153 * to adjust it to something that might or might not work. */
157 while (avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
158 int64_t* current_array;
160 // Expect array object in context
161 if (avio_r8(ioc) != AMF_DATA_TYPE_ARRAY)
164 arraylen = avio_rb32(ioc);
169 * Expect only 'times' or 'filepositions' sub-arrays in other case refuse to use such metadata
172 if (!strcmp(KEYFRAMES_TIMESTAMP_TAG, str_val) && !times) {
173 if (!(times = av_mallocz(sizeof(*times) * arraylen))) {
174 ret = AVERROR(ENOMEM);
178 current_array = times;
179 } else if (!strcmp(KEYFRAMES_BYTEOFFSET_TAG, str_val) && !filepositions) {
180 if (!(filepositions = av_mallocz(sizeof(*filepositions) * arraylen))) {
181 ret = AVERROR(ENOMEM);
184 fileposlen = arraylen;
185 current_array = filepositions;
186 } else // unexpected metatag inside keyframes, will not use such metadata for indexing
189 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
190 if (avio_r8(ioc) != AMF_DATA_TYPE_NUMBER)
192 num_val = av_int2dbl(avio_rb64(ioc));
193 current_array[i] = num_val;
195 if (times && filepositions) {
196 // All done, exiting at a position allowing amf_parse_object
197 // to finish parsing the object
203 if (!ret && timeslen == fileposlen)
204 for (i = 0; i < fileposlen; i++)
205 av_add_index_entry(vstream, filepositions[i], times[i]*1000, 0, 0, AVINDEX_KEYFRAME);
207 av_log(s, AV_LOG_WARNING, "Invalid keyframes object, skipping.\n");
211 av_freep(&filepositions);
212 // If we got unexpected data, but successfully reset back to
213 // the start pos, the caller can continue parsing
214 if (ret < 0 && avio_seek(ioc, initial_pos, SEEK_SET) > 0)
219 static int amf_parse_object(AVFormatContext *s, AVStream *astream, AVStream *vstream, const char *key, int64_t max_pos, int depth) {
220 AVCodecContext *acodec, *vcodec;
222 AMFDataType amf_type;
229 amf_type = avio_r8(ioc);
232 case AMF_DATA_TYPE_NUMBER:
233 num_val = av_int2dbl(avio_rb64(ioc)); break;
234 case AMF_DATA_TYPE_BOOL:
235 num_val = avio_r8(ioc); break;
236 case AMF_DATA_TYPE_STRING:
237 if(amf_get_string(ioc, str_val, sizeof(str_val)) < 0)
240 case AMF_DATA_TYPE_OBJECT: {
243 if ((vstream || astream) && key && !strcmp(KEYFRAMES_TAG, key) && depth == 1)
244 if (parse_keyframes_index(s, ioc, vstream ? vstream : astream,
248 while(avio_tell(ioc) < max_pos - 2 && (keylen = avio_rb16(ioc))) {
249 avio_skip(ioc, keylen); //skip key string
250 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
251 return -1; //if we couldn't skip, bomb out.
253 if(avio_r8(ioc) != AMF_END_OF_OBJECT)
257 case AMF_DATA_TYPE_NULL:
258 case AMF_DATA_TYPE_UNDEFINED:
259 case AMF_DATA_TYPE_UNSUPPORTED:
260 break; //these take up no additional space
261 case AMF_DATA_TYPE_MIXEDARRAY:
262 avio_skip(ioc, 4); //skip 32-bit max array index
263 while(avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
264 //this is the only case in which we would want a nested parse to not skip over the object
265 if(amf_parse_object(s, astream, vstream, str_val, max_pos, depth + 1) < 0)
268 if(avio_r8(ioc) != AMF_END_OF_OBJECT)
271 case AMF_DATA_TYPE_ARRAY: {
272 unsigned int arraylen, i;
274 arraylen = avio_rb32(ioc);
275 for(i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
276 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
277 return -1; //if we couldn't skip, bomb out.
281 case AMF_DATA_TYPE_DATE:
282 avio_skip(ioc, 8 + 2); //timestamp (double) and UTC offset (int16)
284 default: //unsupported type, we couldn't skip
288 if(depth == 1 && key) { //only look for metadata values when we are not nested and key != NULL
289 acodec = astream ? astream->codec : NULL;
290 vcodec = vstream ? vstream->codec : NULL;
292 if (amf_type == AMF_DATA_TYPE_NUMBER) {
293 if (!strcmp(key, "duration"))
294 s->duration = num_val * AV_TIME_BASE;
295 else if (!strcmp(key, "videodatarate") && vcodec && 0 <= (int)(num_val * 1024.0))
296 vcodec->bit_rate = num_val * 1024.0;
297 else if (!strcmp(key, "audiodatarate") && acodec && 0 <= (int)(num_val * 1024.0))
298 acodec->bit_rate = num_val * 1024.0;
301 if (!strcmp(key, "duration") ||
302 !strcmp(key, "filesize") ||
303 !strcmp(key, "width") ||
304 !strcmp(key, "height") ||
305 !strcmp(key, "videodatarate") ||
306 !strcmp(key, "framerate") ||
307 !strcmp(key, "videocodecid") ||
308 !strcmp(key, "audiodatarate") ||
309 !strcmp(key, "audiosamplerate") ||
310 !strcmp(key, "audiosamplesize") ||
311 !strcmp(key, "stereo") ||
312 !strcmp(key, "audiocodecid"))
315 if(amf_type == AMF_DATA_TYPE_BOOL) {
316 av_strlcpy(str_val, num_val > 0 ? "true" : "false", sizeof(str_val));
317 av_dict_set(&s->metadata, key, str_val, 0);
318 } else if(amf_type == AMF_DATA_TYPE_NUMBER) {
319 snprintf(str_val, sizeof(str_val), "%.f", num_val);
320 av_dict_set(&s->metadata, key, str_val, 0);
321 } else if (amf_type == AMF_DATA_TYPE_STRING)
322 av_dict_set(&s->metadata, key, str_val, 0);
328 static int flv_read_metabody(AVFormatContext *s, int64_t next_pos) {
330 AVStream *stream, *astream, *vstream;
333 char buffer[11]; //only needs to hold the string "onMetaData". Anything longer is something we don't want.
339 //first object needs to be "onMetaData" string
341 if(type != AMF_DATA_TYPE_STRING || amf_get_string(ioc, buffer, sizeof(buffer)) < 0 || strcmp(buffer, "onMetaData"))
344 //find the streams now so that amf_parse_object doesn't need to do the lookup every time it is called.
345 for(i = 0; i < s->nb_streams; i++) {
346 stream = s->streams[i];
347 if (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO) astream = stream;
348 else if(stream->codec->codec_type == AVMEDIA_TYPE_VIDEO) vstream = stream;
351 //parse the second object (we want a mixed array)
352 if(amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
358 static AVStream *create_stream(AVFormatContext *s, int is_audio){
359 AVStream *st = avformat_new_stream(s, NULL);
363 st->codec->codec_type = is_audio ? AVMEDIA_TYPE_AUDIO : AVMEDIA_TYPE_VIDEO;
364 avpriv_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
368 static int flv_read_header(AVFormatContext *s,
369 AVFormatParameters *ap)
374 flags = avio_r8(s->pb);
375 /* old flvtool cleared this field */
376 /* FIXME: better fix needed */
378 flags = FLV_HEADER_FLAG_HASVIDEO | FLV_HEADER_FLAG_HASAUDIO;
379 av_log(s, AV_LOG_WARNING, "Broken FLV file, which says no streams present, this might fail\n");
382 if((flags & (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
383 != (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
384 s->ctx_flags |= AVFMTCTX_NOHEADER;
386 if(flags & FLV_HEADER_FLAG_HASVIDEO){
387 if(!create_stream(s, 0))
388 return AVERROR(ENOMEM);
390 if(flags & FLV_HEADER_FLAG_HASAUDIO){
391 if(!create_stream(s, 1))
392 return AVERROR(ENOMEM);
395 offset = avio_rb32(s->pb);
396 avio_seek(s->pb, offset, SEEK_SET);
404 static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
406 av_free(st->codec->extradata);
407 st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
408 if (!st->codec->extradata)
409 return AVERROR(ENOMEM);
410 st->codec->extradata_size = size;
411 avio_read(s->pb, st->codec->extradata, st->codec->extradata_size);
415 static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
417 FLVContext *flv = s->priv_data;
418 int ret, i, type, size, flags, is_audio;
420 int64_t dts, pts = AV_NOPTS_VALUE;
423 for(;;avio_skip(s->pb, 4)){ /* pkt size is repeated at end. skip it */
424 pos = avio_tell(s->pb);
425 type = avio_r8(s->pb);
426 size = avio_rb24(s->pb);
427 dts = avio_rb24(s->pb);
428 dts |= avio_r8(s->pb) << 24;
429 av_dlog(s, "type:%d, size:%d, dts:%"PRId64"\n", type, size, dts);
430 if (s->pb->eof_reached)
432 avio_skip(s->pb, 3); /* stream id, always 0 */
438 next= size + avio_tell(s->pb);
440 if (type == FLV_TAG_TYPE_AUDIO) {
442 flags = avio_r8(s->pb);
444 } else if (type == FLV_TAG_TYPE_VIDEO) {
446 flags = avio_r8(s->pb);
448 if ((flags & 0xf0) == 0x50) /* video info / command frame */
451 if (type == FLV_TAG_TYPE_META && size > 13+1+4)
452 flv_read_metabody(s, next);
453 else /* skip packet */
454 av_log(s, AV_LOG_DEBUG, "skipping flv packet: type %d, size %d, flags %d\n", type, size, flags);
456 avio_seek(s->pb, next, SEEK_SET);
460 /* skip empty data packets */
464 /* now find stream */
465 for(i=0;i<s->nb_streams;i++) {
467 if (st->id == is_audio)
470 if(i == s->nb_streams){
471 av_log(s, AV_LOG_ERROR, "invalid stream\n");
472 st= create_stream(s, is_audio);
473 s->ctx_flags &= ~AVFMTCTX_NOHEADER;
475 av_dlog(s, "%d %X %d \n", is_audio, flags, st->discard);
476 if( (st->discard >= AVDISCARD_NONKEY && !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY || is_audio))
477 ||(st->discard >= AVDISCARD_BIDIR && ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && !is_audio))
478 || st->discard >= AVDISCARD_ALL
480 avio_seek(s->pb, next, SEEK_SET);
483 if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY)
484 av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
488 // if not streamed and no duration from metadata then seek to end to find the duration from the timestamps
489 if(s->pb->seekable && (!s->duration || s->duration==AV_NOPTS_VALUE)){
491 const int64_t pos= avio_tell(s->pb);
492 const int64_t fsize= avio_size(s->pb);
493 avio_seek(s->pb, fsize-4, SEEK_SET);
494 size= avio_rb32(s->pb);
495 avio_seek(s->pb, fsize-3-size, SEEK_SET);
496 if(size == avio_rb24(s->pb) + 11){
497 uint32_t ts = avio_rb24(s->pb);
498 ts |= avio_r8(s->pb) << 24;
499 s->duration = ts * (int64_t)AV_TIME_BASE / 1000;
501 avio_seek(s->pb, pos, SEEK_SET);
505 if(!st->codec->channels || !st->codec->sample_rate || !st->codec->bits_per_coded_sample) {
506 st->codec->channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
507 st->codec->sample_rate = (44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3);
508 st->codec->bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
510 if(!st->codec->codec_id){
511 flv_set_audio_codec(s, st, flags & FLV_AUDIO_CODECID_MASK);
514 size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK);
517 if (st->codec->codec_id == CODEC_ID_AAC ||
518 st->codec->codec_id == CODEC_ID_H264) {
519 int type = avio_r8(s->pb);
521 if (st->codec->codec_id == CODEC_ID_H264) {
522 int32_t cts = (avio_rb24(s->pb)+0xff800000)^0xff800000; // sign extension
524 if (cts < 0) { // dts are wrong
526 av_log(s, AV_LOG_WARNING, "negative cts, previous timestamps might be wrong\n");
529 dts = AV_NOPTS_VALUE;
532 if ((ret = flv_get_extradata(s, st, size)) < 0)
534 if (st->codec->codec_id == CODEC_ID_AAC) {
535 MPEG4AudioConfig cfg;
536 avpriv_mpeg4audio_get_config(&cfg, st->codec->extradata,
537 st->codec->extradata_size * 8, 1);
538 st->codec->channels = cfg.channels;
539 if (cfg.ext_sample_rate)
540 st->codec->sample_rate = cfg.ext_sample_rate;
542 st->codec->sample_rate = cfg.sample_rate;
543 av_dlog(s, "mp4a config channels %d sample rate %d\n",
544 st->codec->channels, st->codec->sample_rate);
547 ret = AVERROR(EAGAIN);
552 /* skip empty data packets */
554 ret = AVERROR(EAGAIN);
558 ret= av_get_packet(s->pb, pkt, size);
562 /* note: we need to modify the packet size here to handle the last
566 pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
567 pkt->stream_index = st->index;
569 if (is_audio || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY))
570 pkt->flags |= AV_PKT_FLAG_KEY;
577 static int flv_read_seek(AVFormatContext *s, int stream_index,
578 int64_t ts, int flags)
580 return avio_seek_time(s->pb, stream_index, ts, flags);
583 #if 0 /* don't know enough to implement this */
584 static int flv_read_seek2(AVFormatContext *s, int stream_index,
585 int64_t min_ts, int64_t ts, int64_t max_ts, int flags)
587 int ret = AVERROR(ENOSYS);
589 if (ts - min_ts > (uint64_t)(max_ts - ts)) flags |= AVSEEK_FLAG_BACKWARD;
591 if (!s->pb->seekable) {
592 if (stream_index < 0) {
593 stream_index = av_find_default_stream_index(s);
594 if (stream_index < 0)
597 /* timestamp for default must be expressed in AV_TIME_BASE units */
598 ts = av_rescale_rnd(ts, 1000, AV_TIME_BASE,
599 flags & AVSEEK_FLAG_BACKWARD ? AV_ROUND_DOWN : AV_ROUND_UP);
601 ret = avio_seek_time(s->pb, stream_index, ts, flags);
604 if (ret == AVERROR(ENOSYS))
605 ret = av_seek_frame(s, stream_index, ts, flags);
610 AVInputFormat ff_flv_demuxer = {
612 .long_name = NULL_IF_CONFIG_SMALL("FLV format"),
613 .priv_data_size = sizeof(FLVContext),
614 .read_probe = flv_probe,
615 .read_header = flv_read_header,
616 .read_packet = flv_read_packet,
617 .read_seek = flv_read_seek,
619 .read_seek2 = flv_read_seek2,
622 .value = CODEC_ID_FLV1,