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"
34 #include "avio_internal.h"
37 #define KEYFRAMES_TAG "keyframes"
38 #define KEYFRAMES_TIMESTAMP_TAG "times"
39 #define KEYFRAMES_BYTEOFFSET_TAG "filepositions"
42 int wrong_dts; ///< wrong dts due to negative cts
45 static int flv_probe(AVProbeData *p)
50 if (d[0] == 'F' && d[1] == 'L' && d[2] == 'V' && d[3] < 5 && d[5]==0 && AV_RB32(d+5)>8) {
51 return AVPROBE_SCORE_MAX;
56 static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream, int flv_codecid) {
57 AVCodecContext *acodec = astream->codec;
59 //no distinction between S16 and S8 PCM codec flags
61 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 :
68 case FLV_CODECID_PCM_LE:
69 acodec->codec_id = acodec->bits_per_coded_sample == 8 ? CODEC_ID_PCM_U8 : CODEC_ID_PCM_S16LE; break;
70 case FLV_CODECID_AAC : acodec->codec_id = CODEC_ID_AAC; break;
71 case FLV_CODECID_ADPCM: acodec->codec_id = CODEC_ID_ADPCM_SWF; break;
72 case FLV_CODECID_SPEEX:
73 acodec->codec_id = CODEC_ID_SPEEX;
74 acodec->sample_rate = 16000;
76 case FLV_CODECID_MP3 : acodec->codec_id = CODEC_ID_MP3 ; astream->need_parsing = AVSTREAM_PARSE_FULL; break;
77 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
78 acodec->sample_rate = 8000; //in case metadata does not otherwise declare samplerate
79 acodec->codec_id = CODEC_ID_NELLYMOSER;
81 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
82 acodec->sample_rate = 16000;
83 acodec->codec_id = CODEC_ID_NELLYMOSER;
85 case FLV_CODECID_NELLYMOSER:
86 acodec->codec_id = CODEC_ID_NELLYMOSER;
89 av_log(s, AV_LOG_INFO, "Unsupported audio codec (%x)\n", flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
90 acodec->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
94 static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream, int flv_codecid) {
95 AVCodecContext *vcodec = vstream->codec;
97 case FLV_CODECID_H263 : vcodec->codec_id = CODEC_ID_FLV1 ; break;
98 case FLV_CODECID_SCREEN: vcodec->codec_id = CODEC_ID_FLASHSV; break;
99 case FLV_CODECID_SCREEN2: vcodec->codec_id = CODEC_ID_FLASHSV2; break;
100 case FLV_CODECID_VP6 : vcodec->codec_id = CODEC_ID_VP6F ;
101 case FLV_CODECID_VP6A :
102 if(flv_codecid == FLV_CODECID_VP6A)
103 vcodec->codec_id = CODEC_ID_VP6A;
104 if(vcodec->extradata_size != 1) {
105 vcodec->extradata_size = 1;
106 vcodec->extradata = av_malloc(1);
108 vcodec->extradata[0] = avio_r8(s->pb);
109 return 1; // 1 byte body size adjustment for flv_read_packet()
110 case FLV_CODECID_H264:
111 vcodec->codec_id = CODEC_ID_H264;
112 return 3; // not 4, reading packet type will consume one byte
114 av_log(s, AV_LOG_INFO, "Unsupported video codec (%x)\n", flv_codecid);
115 vcodec->codec_tag = flv_codecid;
121 static int amf_get_string(AVIOContext *ioc, char *buffer, int buffsize) {
122 int length = avio_rb16(ioc);
123 if(length >= buffsize) {
124 avio_skip(ioc, length);
128 avio_read(ioc, buffer, length);
130 buffer[length] = '\0';
135 static int parse_keyframes_index(AVFormatContext *s, AVIOContext *ioc, AVStream *vstream, int64_t max_pos) {
136 unsigned int arraylen = 0, timeslen = 0, fileposlen = 0, i;
139 int64_t *times = NULL;
140 int64_t *filepositions = NULL;
141 int ret = AVERROR(ENOSYS);
142 int64_t initial_pos = avio_tell(ioc);
143 AVDictionaryEntry *creator = av_dict_get(s->metadata, "metadatacreator",
146 if (creator && !strcmp(creator->value, "MEGA")) {
147 /* Files with this metadatacreator tag seem to have filepositions
148 * pointing at the 4 trailer bytes of the previous packet,
149 * which isn't the norm (nor what we expect here, nor what
150 * jwplayer + lighttpd expect, nor what flvtool2 produces).
151 * Just ignore the index in this case, instead of risking trying
152 * to adjust it to something that might or might not work. */
156 while (avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
157 int64_t* current_array;
159 // Expect array object in context
160 if (avio_r8(ioc) != AMF_DATA_TYPE_ARRAY)
163 arraylen = avio_rb32(ioc);
168 * Expect only 'times' or 'filepositions' sub-arrays in other case refuse to use such metadata
171 if (!strcmp(KEYFRAMES_TIMESTAMP_TAG, str_val) && !times) {
172 if (!(times = av_mallocz(sizeof(*times) * arraylen))) {
173 ret = AVERROR(ENOMEM);
177 current_array = times;
178 } else if (!strcmp(KEYFRAMES_BYTEOFFSET_TAG, str_val) && !filepositions) {
179 if (!(filepositions = av_mallocz(sizeof(*filepositions) * arraylen))) {
180 ret = AVERROR(ENOMEM);
183 fileposlen = arraylen;
184 current_array = filepositions;
185 } else // unexpected metatag inside keyframes, will not use such metadata for indexing
188 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
189 if (avio_r8(ioc) != AMF_DATA_TYPE_NUMBER)
191 num_val = av_int2dbl(avio_rb64(ioc));
192 current_array[i] = num_val;
194 if (times && filepositions) {
195 // All done, exiting at a position allowing amf_parse_object
196 // to finish parsing the object
202 if (!ret && timeslen == fileposlen)
203 for (i = 0; i < fileposlen; i++)
204 av_add_index_entry(vstream, filepositions[i], times[i]*1000, 0, 0, AVINDEX_KEYFRAME);
206 av_log(s, AV_LOG_WARNING, "Invalid keyframes object, skipping.\n");
210 av_freep(&filepositions);
211 // If we got unexpected data, but successfully reset back to
212 // the start pos, the caller can continue parsing
213 if (ret < 0 && avio_seek(ioc, initial_pos, SEEK_SET) > 0)
218 static int amf_parse_object(AVFormatContext *s, AVStream *astream, AVStream *vstream, const char *key, int64_t max_pos, int depth) {
219 AVCodecContext *acodec, *vcodec;
221 AMFDataType amf_type;
228 amf_type = avio_r8(ioc);
231 case AMF_DATA_TYPE_NUMBER:
232 num_val = av_int2dbl(avio_rb64(ioc)); break;
233 case AMF_DATA_TYPE_BOOL:
234 num_val = avio_r8(ioc); break;
235 case AMF_DATA_TYPE_STRING:
236 if(amf_get_string(ioc, str_val, sizeof(str_val)) < 0)
239 case AMF_DATA_TYPE_OBJECT: {
242 if ((vstream || astream) && key && !strcmp(KEYFRAMES_TAG, key) && depth == 1)
243 if (parse_keyframes_index(s, ioc, vstream ? vstream : astream,
247 while(avio_tell(ioc) < max_pos - 2 && (keylen = avio_rb16(ioc))) {
248 avio_skip(ioc, keylen); //skip key string
249 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
250 return -1; //if we couldn't skip, bomb out.
252 if(avio_r8(ioc) != AMF_END_OF_OBJECT)
256 case AMF_DATA_TYPE_NULL:
257 case AMF_DATA_TYPE_UNDEFINED:
258 case AMF_DATA_TYPE_UNSUPPORTED:
259 break; //these take up no additional space
260 case AMF_DATA_TYPE_MIXEDARRAY:
261 avio_skip(ioc, 4); //skip 32-bit max array index
262 while(avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
263 //this is the only case in which we would want a nested parse to not skip over the object
264 if(amf_parse_object(s, astream, vstream, str_val, max_pos, depth + 1) < 0)
267 if(avio_r8(ioc) != AMF_END_OF_OBJECT)
270 case AMF_DATA_TYPE_ARRAY: {
271 unsigned int arraylen, i;
273 arraylen = avio_rb32(ioc);
274 for(i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
275 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
276 return -1; //if we couldn't skip, bomb out.
280 case AMF_DATA_TYPE_DATE:
281 avio_skip(ioc, 8 + 2); //timestamp (double) and UTC offset (int16)
283 default: //unsupported type, we couldn't skip
287 if(depth == 1 && key) { //only look for metadata values when we are not nested and key != NULL
288 acodec = astream ? astream->codec : NULL;
289 vcodec = vstream ? vstream->codec : NULL;
291 if (amf_type == AMF_DATA_TYPE_NUMBER) {
292 if (!strcmp(key, "duration"))
293 s->duration = num_val * AV_TIME_BASE;
294 else if (!strcmp(key, "videodatarate") && vcodec && 0 <= (int)(num_val * 1024.0))
295 vcodec->bit_rate = num_val * 1024.0;
296 else if (!strcmp(key, "audiodatarate") && acodec && 0 <= (int)(num_val * 1024.0))
297 acodec->bit_rate = num_val * 1024.0;
300 if (!strcmp(key, "duration") ||
301 !strcmp(key, "filesize") ||
302 !strcmp(key, "width") ||
303 !strcmp(key, "height") ||
304 !strcmp(key, "videodatarate") ||
305 !strcmp(key, "framerate") ||
306 !strcmp(key, "videocodecid") ||
307 !strcmp(key, "audiodatarate") ||
308 !strcmp(key, "audiosamplerate") ||
309 !strcmp(key, "audiosamplesize") ||
310 !strcmp(key, "stereo") ||
311 !strcmp(key, "audiocodecid"))
314 if(amf_type == AMF_DATA_TYPE_BOOL) {
315 av_strlcpy(str_val, num_val > 0 ? "true" : "false", sizeof(str_val));
316 av_dict_set(&s->metadata, key, str_val, 0);
317 } else if(amf_type == AMF_DATA_TYPE_NUMBER) {
318 snprintf(str_val, sizeof(str_val), "%.f", num_val);
319 av_dict_set(&s->metadata, key, str_val, 0);
320 } else if (amf_type == AMF_DATA_TYPE_STRING)
321 av_dict_set(&s->metadata, key, str_val, 0);
327 static int flv_read_metabody(AVFormatContext *s, int64_t next_pos) {
329 AVStream *stream, *astream, *vstream;
332 char buffer[11]; //only needs to hold the string "onMetaData". Anything longer is something we don't want.
338 //first object needs to be "onMetaData" string
340 if(type != AMF_DATA_TYPE_STRING || amf_get_string(ioc, buffer, sizeof(buffer)) < 0 || strcmp(buffer, "onMetaData"))
343 //find the streams now so that amf_parse_object doesn't need to do the lookup every time it is called.
344 for(i = 0; i < s->nb_streams; i++) {
345 stream = s->streams[i];
346 if (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO) astream = stream;
347 else if(stream->codec->codec_type == AVMEDIA_TYPE_VIDEO) vstream = stream;
350 //parse the second object (we want a mixed array)
351 if(amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
357 static AVStream *create_stream(AVFormatContext *s, int is_audio){
358 AVStream *st = avformat_new_stream(s, NULL);
362 st->codec->codec_type = is_audio ? AVMEDIA_TYPE_AUDIO : AVMEDIA_TYPE_VIDEO;
363 av_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
367 static int flv_read_header(AVFormatContext *s,
368 AVFormatParameters *ap)
373 flags = avio_r8(s->pb);
374 /* old flvtool cleared this field */
375 /* FIXME: better fix needed */
377 flags = FLV_HEADER_FLAG_HASVIDEO | FLV_HEADER_FLAG_HASAUDIO;
378 av_log(s, AV_LOG_WARNING, "Broken FLV file, which says no streams present, this might fail\n");
381 if((flags & (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
382 != (FLV_HEADER_FLAG_HASVIDEO|FLV_HEADER_FLAG_HASAUDIO))
383 s->ctx_flags |= AVFMTCTX_NOHEADER;
385 if(flags & FLV_HEADER_FLAG_HASVIDEO){
386 if(!create_stream(s, 0))
387 return AVERROR(ENOMEM);
389 if(flags & FLV_HEADER_FLAG_HASAUDIO){
390 if(!create_stream(s, 1))
391 return AVERROR(ENOMEM);
394 offset = avio_rb32(s->pb);
395 avio_seek(s->pb, offset, SEEK_SET);
403 static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
405 av_free(st->codec->extradata);
406 st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
407 if (!st->codec->extradata)
408 return AVERROR(ENOMEM);
409 st->codec->extradata_size = size;
410 avio_read(s->pb, st->codec->extradata, st->codec->extradata_size);
414 static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
416 FLVContext *flv = s->priv_data;
417 int ret, i, type, size, flags, is_audio;
419 int64_t dts, pts = AV_NOPTS_VALUE;
422 for(;;avio_skip(s->pb, 4)){ /* pkt size is repeated at end. skip it */
423 pos = avio_tell(s->pb);
424 type = avio_r8(s->pb);
425 size = avio_rb24(s->pb);
426 dts = avio_rb24(s->pb);
427 dts |= avio_r8(s->pb) << 24;
428 av_dlog(s, "type:%d, size:%d, dts:%"PRId64"\n", type, size, dts);
429 if (s->pb->eof_reached)
431 avio_skip(s->pb, 3); /* stream id, always 0 */
437 next= size + avio_tell(s->pb);
439 if (type == FLV_TAG_TYPE_AUDIO) {
441 flags = avio_r8(s->pb);
443 } else if (type == FLV_TAG_TYPE_VIDEO) {
445 flags = avio_r8(s->pb);
447 if ((flags & 0xf0) == 0x50) /* video info / command frame */
450 if (type == FLV_TAG_TYPE_META && size > 13+1+4)
451 flv_read_metabody(s, next);
452 else /* skip packet */
453 av_log(s, AV_LOG_DEBUG, "skipping flv packet: type %d, size %d, flags %d\n", type, size, flags);
455 avio_seek(s->pb, next, SEEK_SET);
459 /* skip empty data packets */
463 /* now find stream */
464 for(i=0;i<s->nb_streams;i++) {
466 if (st->id == is_audio)
469 if(i == s->nb_streams){
470 av_log(s, AV_LOG_ERROR, "invalid stream\n");
471 st= create_stream(s, is_audio);
472 s->ctx_flags &= ~AVFMTCTX_NOHEADER;
474 av_dlog(s, "%d %X %d \n", is_audio, flags, st->discard);
475 if( (st->discard >= AVDISCARD_NONKEY && !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY || is_audio))
476 ||(st->discard >= AVDISCARD_BIDIR && ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && !is_audio))
477 || st->discard >= AVDISCARD_ALL
479 avio_seek(s->pb, next, SEEK_SET);
482 if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY)
483 av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
487 // if not streamed and no duration from metadata then seek to end to find the duration from the timestamps
488 if(s->pb->seekable && (!s->duration || s->duration==AV_NOPTS_VALUE)){
490 const int64_t pos= avio_tell(s->pb);
491 const int64_t fsize= avio_size(s->pb);
492 avio_seek(s->pb, fsize-4, SEEK_SET);
493 size= avio_rb32(s->pb);
494 avio_seek(s->pb, fsize-3-size, SEEK_SET);
495 if(size == avio_rb24(s->pb) + 11){
496 uint32_t ts = avio_rb24(s->pb);
497 ts |= avio_r8(s->pb) << 24;
498 s->duration = ts * (int64_t)AV_TIME_BASE / 1000;
500 avio_seek(s->pb, pos, SEEK_SET);
504 if(!st->codec->channels || !st->codec->sample_rate || !st->codec->bits_per_coded_sample) {
505 st->codec->channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
506 st->codec->sample_rate = (44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3);
507 st->codec->bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
509 if(!st->codec->codec_id){
510 flv_set_audio_codec(s, st, flags & FLV_AUDIO_CODECID_MASK);
513 size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK);
516 if (st->codec->codec_id == CODEC_ID_AAC ||
517 st->codec->codec_id == CODEC_ID_H264) {
518 int type = avio_r8(s->pb);
520 if (st->codec->codec_id == CODEC_ID_H264) {
521 int32_t cts = (avio_rb24(s->pb)+0xff800000)^0xff800000; // sign extension
523 if (cts < 0) { // dts are wrong
525 av_log(s, AV_LOG_WARNING, "negative cts, previous timestamps might be wrong\n");
528 dts = AV_NOPTS_VALUE;
531 if ((ret = flv_get_extradata(s, st, size)) < 0)
533 if (st->codec->codec_id == CODEC_ID_AAC) {
534 MPEG4AudioConfig cfg;
535 avpriv_mpeg4audio_get_config(&cfg, st->codec->extradata,
536 st->codec->extradata_size);
537 st->codec->channels = cfg.channels;
538 if (cfg.ext_sample_rate)
539 st->codec->sample_rate = cfg.ext_sample_rate;
541 st->codec->sample_rate = cfg.sample_rate;
542 av_dlog(s, "mp4a config channels %d sample rate %d\n",
543 st->codec->channels, st->codec->sample_rate);
546 ret = AVERROR(EAGAIN);
551 /* skip empty data packets */
553 ret = AVERROR(EAGAIN);
557 ret= av_get_packet(s->pb, pkt, size);
561 /* note: we need to modify the packet size here to handle the last
565 pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
566 pkt->stream_index = st->index;
568 if (is_audio || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY))
569 pkt->flags |= AV_PKT_FLAG_KEY;
576 static int flv_read_seek(AVFormatContext *s, int stream_index,
577 int64_t ts, int flags)
579 return avio_seek_time(s->pb, stream_index, ts, flags);
582 #if 0 /* don't know enough to implement this */
583 static int flv_read_seek2(AVFormatContext *s, int stream_index,
584 int64_t min_ts, int64_t ts, int64_t max_ts, int flags)
586 int ret = AVERROR(ENOSYS);
588 if (ts - min_ts > (uint64_t)(max_ts - ts)) flags |= AVSEEK_FLAG_BACKWARD;
590 if (!s->pb->seekable) {
591 if (stream_index < 0) {
592 stream_index = av_find_default_stream_index(s);
593 if (stream_index < 0)
596 /* timestamp for default must be expressed in AV_TIME_BASE units */
597 ts = av_rescale_rnd(ts, 1000, AV_TIME_BASE,
598 flags & AVSEEK_FLAG_BACKWARD ? AV_ROUND_DOWN : AV_ROUND_UP);
600 ret = avio_seek_time(s->pb, stream_index, ts, flags);
603 if (ret == AVERROR(ENOSYS))
604 ret = av_seek_frame(s, stream_index, ts, flags);
609 AVInputFormat ff_flv_demuxer = {
611 .long_name = NULL_IF_CONFIG_SMALL("FLV format"),
612 .priv_data_size = sizeof(FLVContext),
613 .read_probe = flv_probe,
614 .read_header = flv_read_header,
615 .read_packet = flv_read_packet,
616 .read_seek = flv_read_seek,
618 .read_seek2 = flv_read_seek2,
621 .value = CODEC_ID_FLV1,