3 * Copyright (c) 2003 The FFmpeg Project
5 * This demuxer will generate a 1 byte extradata for VP6F content.
7 * - upper 4 bits: difference between encoded width and visible width
8 * - lower 4 bits: difference between encoded height and visible height
10 * This file is part of FFmpeg.
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.
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.
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
27 #include "libavutil/avstring.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/dict.h"
30 #include "libavutil/opt.h"
31 #include "libavutil/intfloat.h"
32 #include "libavutil/mathematics.h"
33 #include "libavcodec/bytestream.h"
34 #include "libavcodec/mpeg4audio.h"
37 #include "avio_internal.h"
40 #define VALIDATE_INDEX_TS_THRESH 2500
42 #define RESYNC_BUFFER_SIZE (1<<20)
44 typedef struct FLVContext {
45 const AVClass *class; ///< Class for private options.
46 int trust_metadata; ///< configure streams according onMetaData
47 int wrong_dts; ///< wrong dts due to negative cts
48 uint8_t *new_extradata[FLV_STREAM_TYPE_NB];
49 int new_extradata_size[FLV_STREAM_TYPE_NB];
60 uint8_t resync_buffer[2*RESYNC_BUFFER_SIZE];
65 int last_keyframe_stream_index;
67 int64_t video_bit_rate;
68 int64_t audio_bit_rate;
69 int64_t *keyframe_times;
70 int64_t *keyframe_filepositions;
75 static int probe(AVProbeData *p, int live)
77 const uint8_t *d = p->buf;
78 unsigned offset = AV_RB32(d + 5);
83 d[3] < 5 && d[5] == 0 &&
84 offset + 100 < p->buf_size &&
86 int is_live = !memcmp(d + offset + 40, "NGINX RTMP", 10);
89 return AVPROBE_SCORE_MAX;
94 static int flv_probe(AVProbeData *p)
99 static int live_flv_probe(AVProbeData *p)
104 static void add_keyframes_index(AVFormatContext *s)
106 FLVContext *flv = s->priv_data;
107 AVStream *stream = NULL;
110 if (flv->last_keyframe_stream_index < 0) {
111 av_log(s, AV_LOG_DEBUG, "keyframe stream hasn't been created\n");
115 av_assert0(flv->last_keyframe_stream_index <= s->nb_streams);
116 stream = s->streams[flv->last_keyframe_stream_index];
118 if (stream->nb_index_entries == 0) {
119 for (i = 0; i < flv->keyframe_count; i++) {
120 av_log(s, AV_LOG_TRACE, "keyframe filepositions = %"PRId64" times = %"PRId64"\n",
121 flv->keyframe_filepositions[i], flv->keyframe_times[i] * 1000);
122 av_add_index_entry(stream, flv->keyframe_filepositions[i],
123 flv->keyframe_times[i] * 1000, 0, 0, AVINDEX_KEYFRAME);
126 av_log(s, AV_LOG_WARNING, "Skipping duplicate index\n");
128 if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
129 av_freep(&flv->keyframe_times);
130 av_freep(&flv->keyframe_filepositions);
131 flv->keyframe_count = 0;
135 static AVStream *create_stream(AVFormatContext *s, int codec_type)
137 FLVContext *flv = s->priv_data;
138 AVStream *st = avformat_new_stream(s, NULL);
141 st->codecpar->codec_type = codec_type;
142 if (s->nb_streams>=3 ||( s->nb_streams==2
143 && s->streams[0]->codecpar->codec_type != AVMEDIA_TYPE_SUBTITLE
144 && s->streams[1]->codecpar->codec_type != AVMEDIA_TYPE_SUBTITLE))
145 s->ctx_flags &= ~AVFMTCTX_NOHEADER;
146 if (codec_type == AVMEDIA_TYPE_AUDIO)
147 st->codecpar->bit_rate = flv->audio_bit_rate;
148 flv->missing_streams &= ~FLV_HEADER_FLAG_HASAUDIO;
149 if (codec_type == AVMEDIA_TYPE_VIDEO) {
150 st->codecpar->bit_rate = flv->video_bit_rate;
151 flv->missing_streams &= ~FLV_HEADER_FLAG_HASVIDEO;
152 st->avg_frame_rate = flv->framerate;
156 avpriv_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
157 flv->last_keyframe_stream_index = s->nb_streams - 1;
158 add_keyframes_index(s);
162 static int flv_same_audio_codec(AVCodecParameters *apar, int flags)
164 int bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
165 int flv_codecid = flags & FLV_AUDIO_CODECID_MASK;
168 if (!apar->codec_id && !apar->codec_tag)
171 if (apar->bits_per_coded_sample != bits_per_coded_sample)
174 switch (flv_codecid) {
175 // no distinction between S16 and S8 PCM codec flags
176 case FLV_CODECID_PCM:
177 codec_id = bits_per_coded_sample == 8
180 : AV_CODEC_ID_PCM_S16BE;
182 : AV_CODEC_ID_PCM_S16LE;
184 return codec_id == apar->codec_id;
185 case FLV_CODECID_PCM_LE:
186 codec_id = bits_per_coded_sample == 8
188 : AV_CODEC_ID_PCM_S16LE;
189 return codec_id == apar->codec_id;
190 case FLV_CODECID_AAC:
191 return apar->codec_id == AV_CODEC_ID_AAC;
192 case FLV_CODECID_ADPCM:
193 return apar->codec_id == AV_CODEC_ID_ADPCM_SWF;
194 case FLV_CODECID_SPEEX:
195 return apar->codec_id == AV_CODEC_ID_SPEEX;
196 case FLV_CODECID_MP3:
197 return apar->codec_id == AV_CODEC_ID_MP3;
198 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
199 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
200 case FLV_CODECID_NELLYMOSER:
201 return apar->codec_id == AV_CODEC_ID_NELLYMOSER;
202 case FLV_CODECID_PCM_MULAW:
203 return apar->sample_rate == 8000 &&
204 apar->codec_id == AV_CODEC_ID_PCM_MULAW;
205 case FLV_CODECID_PCM_ALAW:
206 return apar->sample_rate == 8000 &&
207 apar->codec_id == AV_CODEC_ID_PCM_ALAW;
209 return apar->codec_tag == (flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
213 static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream,
214 AVCodecParameters *apar, int flv_codecid)
216 switch (flv_codecid) {
217 // no distinction between S16 and S8 PCM codec flags
218 case FLV_CODECID_PCM:
219 apar->codec_id = apar->bits_per_coded_sample == 8
222 : AV_CODEC_ID_PCM_S16BE;
224 : AV_CODEC_ID_PCM_S16LE;
227 case FLV_CODECID_PCM_LE:
228 apar->codec_id = apar->bits_per_coded_sample == 8
230 : AV_CODEC_ID_PCM_S16LE;
232 case FLV_CODECID_AAC:
233 apar->codec_id = AV_CODEC_ID_AAC;
235 case FLV_CODECID_ADPCM:
236 apar->codec_id = AV_CODEC_ID_ADPCM_SWF;
238 case FLV_CODECID_SPEEX:
239 apar->codec_id = AV_CODEC_ID_SPEEX;
240 apar->sample_rate = 16000;
242 case FLV_CODECID_MP3:
243 apar->codec_id = AV_CODEC_ID_MP3;
244 astream->need_parsing = AVSTREAM_PARSE_FULL;
246 case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
247 // in case metadata does not otherwise declare samplerate
248 apar->sample_rate = 8000;
249 apar->codec_id = AV_CODEC_ID_NELLYMOSER;
251 case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
252 apar->sample_rate = 16000;
253 apar->codec_id = AV_CODEC_ID_NELLYMOSER;
255 case FLV_CODECID_NELLYMOSER:
256 apar->codec_id = AV_CODEC_ID_NELLYMOSER;
258 case FLV_CODECID_PCM_MULAW:
259 apar->sample_rate = 8000;
260 apar->codec_id = AV_CODEC_ID_PCM_MULAW;
262 case FLV_CODECID_PCM_ALAW:
263 apar->sample_rate = 8000;
264 apar->codec_id = AV_CODEC_ID_PCM_ALAW;
267 avpriv_request_sample(s, "Audio codec (%x)",
268 flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
269 apar->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
273 static int flv_same_video_codec(AVCodecParameters *vpar, int flags)
275 int flv_codecid = flags & FLV_VIDEO_CODECID_MASK;
277 if (!vpar->codec_id && !vpar->codec_tag)
280 switch (flv_codecid) {
281 case FLV_CODECID_H263:
282 return vpar->codec_id == AV_CODEC_ID_FLV1;
283 case FLV_CODECID_SCREEN:
284 return vpar->codec_id == AV_CODEC_ID_FLASHSV;
285 case FLV_CODECID_SCREEN2:
286 return vpar->codec_id == AV_CODEC_ID_FLASHSV2;
287 case FLV_CODECID_VP6:
288 return vpar->codec_id == AV_CODEC_ID_VP6F;
289 case FLV_CODECID_VP6A:
290 return vpar->codec_id == AV_CODEC_ID_VP6A;
291 case FLV_CODECID_H264:
292 return vpar->codec_id == AV_CODEC_ID_H264;
294 return vpar->codec_tag == flv_codecid;
298 static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream,
299 int flv_codecid, int read)
302 AVCodecParameters *par = vstream->codecpar;
303 enum AVCodecID old_codec_id = vstream->codecpar->codec_id;
304 switch (flv_codecid) {
305 case FLV_CODECID_H263:
306 par->codec_id = AV_CODEC_ID_FLV1;
308 case FLV_CODECID_REALH263:
309 par->codec_id = AV_CODEC_ID_H263;
310 break; // Really mean it this time
311 case FLV_CODECID_SCREEN:
312 par->codec_id = AV_CODEC_ID_FLASHSV;
314 case FLV_CODECID_SCREEN2:
315 par->codec_id = AV_CODEC_ID_FLASHSV2;
317 case FLV_CODECID_VP6:
318 par->codec_id = AV_CODEC_ID_VP6F;
319 case FLV_CODECID_VP6A:
320 if (flv_codecid == FLV_CODECID_VP6A)
321 par->codec_id = AV_CODEC_ID_VP6A;
323 if (par->extradata_size != 1) {
324 ff_alloc_extradata(par, 1);
327 par->extradata[0] = avio_r8(s->pb);
331 ret = 1; // 1 byte body size adjustment for flv_read_packet()
333 case FLV_CODECID_H264:
334 par->codec_id = AV_CODEC_ID_H264;
335 vstream->need_parsing = AVSTREAM_PARSE_HEADERS;
336 ret = 3; // not 4, reading packet type will consume one byte
338 case FLV_CODECID_MPEG4:
339 par->codec_id = AV_CODEC_ID_MPEG4;
343 avpriv_request_sample(s, "Video codec (%x)", flv_codecid);
344 par->codec_tag = flv_codecid;
347 if (!vstream->internal->need_context_update && par->codec_id != old_codec_id) {
348 avpriv_request_sample(s, "Changing the codec id midstream");
349 return AVERROR_PATCHWELCOME;
355 static int amf_get_string(AVIOContext *ioc, char *buffer, int buffsize)
357 int length = avio_rb16(ioc);
358 if (length >= buffsize) {
359 avio_skip(ioc, length);
363 avio_read(ioc, buffer, length);
365 buffer[length] = '\0';
370 static int parse_keyframes_index(AVFormatContext *s, AVIOContext *ioc, int64_t max_pos)
372 FLVContext *flv = s->priv_data;
373 unsigned int timeslen = 0, fileposlen = 0, i;
375 int64_t *times = NULL;
376 int64_t *filepositions = NULL;
377 int ret = AVERROR(ENOSYS);
378 int64_t initial_pos = avio_tell(ioc);
380 if (flv->keyframe_count > 0) {
381 av_log(s, AV_LOG_DEBUG, "keyframes have been paresed\n");
384 av_assert0(!flv->keyframe_times);
385 av_assert0(!flv->keyframe_filepositions);
387 if (s->flags & AVFMT_FLAG_IGNIDX)
390 while (avio_tell(ioc) < max_pos - 2 &&
391 amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
392 int64_t **current_array;
393 unsigned int arraylen;
395 // Expect array object in context
396 if (avio_r8(ioc) != AMF_DATA_TYPE_ARRAY)
399 arraylen = avio_rb32(ioc);
403 if (!strcmp(KEYFRAMES_TIMESTAMP_TAG , str_val) && !times) {
404 current_array = ×
406 } else if (!strcmp(KEYFRAMES_BYTEOFFSET_TAG, str_val) &&
408 current_array = &filepositions;
409 fileposlen = arraylen;
411 // unexpected metatag inside keyframes, will not use such
412 // metadata for indexing
415 if (!(*current_array = av_mallocz(sizeof(**current_array) * arraylen))) {
416 ret = AVERROR(ENOMEM);
420 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
421 if (avio_r8(ioc) != AMF_DATA_TYPE_NUMBER)
423 current_array[0][i] = av_int2double(avio_rb64(ioc));
425 if (times && filepositions) {
426 // All done, exiting at a position allowing amf_parse_object
427 // to finish parsing the object
433 if (timeslen == fileposlen && fileposlen>1 && max_pos <= filepositions[0]) {
434 for (i = 0; i < FFMIN(2,fileposlen); i++) {
435 flv->validate_index[i].pos = filepositions[i];
436 flv->validate_index[i].dts = times[i] * 1000;
437 flv->validate_count = i + 1;
439 flv->keyframe_times = times;
440 flv->keyframe_filepositions = filepositions;
441 flv->keyframe_count = timeslen;
443 filepositions = NULL;
446 av_log(s, AV_LOG_WARNING, "Invalid keyframes object, skipping.\n");
451 av_freep(&filepositions);
452 avio_seek(ioc, initial_pos, SEEK_SET);
456 static int amf_parse_object(AVFormatContext *s, AVStream *astream,
457 AVStream *vstream, const char *key,
458 int64_t max_pos, int depth)
460 AVCodecParameters *apar, *vpar;
461 FLVContext *flv = s->priv_data;
463 AMFDataType amf_type;
469 amf_type = avio_r8(ioc);
472 case AMF_DATA_TYPE_NUMBER:
473 num_val = av_int2double(avio_rb64(ioc));
475 case AMF_DATA_TYPE_BOOL:
476 num_val = avio_r8(ioc);
478 case AMF_DATA_TYPE_STRING:
479 if (amf_get_string(ioc, str_val, sizeof(str_val)) < 0) {
480 av_log(s, AV_LOG_ERROR, "AMF_DATA_TYPE_STRING parsing failed\n");
484 case AMF_DATA_TYPE_OBJECT:
487 !strcmp(KEYFRAMES_TAG, key) && depth == 1)
488 if (parse_keyframes_index(s, ioc,
490 av_log(s, AV_LOG_ERROR, "Keyframe index parsing failed\n");
492 add_keyframes_index(s);
493 while (avio_tell(ioc) < max_pos - 2 &&
494 amf_get_string(ioc, str_val, sizeof(str_val)) > 0)
495 if (amf_parse_object(s, astream, vstream, str_val, max_pos,
497 return -1; // if we couldn't skip, bomb out.
498 if (avio_r8(ioc) != AMF_END_OF_OBJECT) {
499 av_log(s, AV_LOG_ERROR, "Missing AMF_END_OF_OBJECT in AMF_DATA_TYPE_OBJECT\n");
503 case AMF_DATA_TYPE_NULL:
504 case AMF_DATA_TYPE_UNDEFINED:
505 case AMF_DATA_TYPE_UNSUPPORTED:
506 break; // these take up no additional space
507 case AMF_DATA_TYPE_MIXEDARRAY:
510 avio_skip(ioc, 4); // skip 32-bit max array index
511 while (avio_tell(ioc) < max_pos - 2 &&
512 amf_get_string(ioc, str_val, sizeof(str_val)) > 0)
513 // this is the only case in which we would want a nested
514 // parse to not skip over the object
515 if (amf_parse_object(s, astream, vstream, str_val, max_pos,
519 if (v != AMF_END_OF_OBJECT) {
520 av_log(s, AV_LOG_ERROR, "Missing AMF_END_OF_OBJECT in AMF_DATA_TYPE_MIXEDARRAY, found %d\n", v);
525 case AMF_DATA_TYPE_ARRAY:
527 unsigned int arraylen, i;
529 arraylen = avio_rb32(ioc);
530 for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++)
531 if (amf_parse_object(s, NULL, NULL, NULL, max_pos,
533 return -1; // if we couldn't skip, bomb out.
536 case AMF_DATA_TYPE_DATE:
537 avio_skip(ioc, 8 + 2); // timestamp (double) and UTC offset (int16)
539 default: // unsupported type, we couldn't skip
540 av_log(s, AV_LOG_ERROR, "unsupported amf type %d\n", amf_type);
545 apar = astream ? astream->codecpar : NULL;
546 vpar = vstream ? vstream->codecpar : NULL;
548 // stream info doesn't live any deeper than the first object
550 if (amf_type == AMF_DATA_TYPE_NUMBER ||
551 amf_type == AMF_DATA_TYPE_BOOL) {
552 if (!strcmp(key, "duration"))
553 s->duration = num_val * AV_TIME_BASE;
554 else if (!strcmp(key, "videodatarate") &&
555 0 <= (int)(num_val * 1024.0))
556 flv->video_bit_rate = num_val * 1024.0;
557 else if (!strcmp(key, "audiodatarate") &&
558 0 <= (int)(num_val * 1024.0))
559 flv->audio_bit_rate = num_val * 1024.0;
560 else if (!strcmp(key, "datastream")) {
561 AVStream *st = create_stream(s, AVMEDIA_TYPE_SUBTITLE);
563 return AVERROR(ENOMEM);
564 st->codecpar->codec_id = AV_CODEC_ID_TEXT;
565 } else if (!strcmp(key, "framerate")) {
566 flv->framerate = av_d2q(num_val, 1000);
568 vstream->avg_frame_rate = flv->framerate;
569 } else if (flv->trust_metadata) {
570 if (!strcmp(key, "videocodecid") && vpar) {
571 int ret = flv_set_video_codec(s, vstream, num_val, 0);
574 } else if (!strcmp(key, "audiocodecid") && apar) {
575 int id = ((int)num_val) << FLV_AUDIO_CODECID_OFFSET;
576 flv_set_audio_codec(s, astream, apar, id);
577 } else if (!strcmp(key, "audiosamplerate") && apar) {
578 apar->sample_rate = num_val;
579 } else if (!strcmp(key, "audiosamplesize") && apar) {
580 apar->bits_per_coded_sample = num_val;
581 } else if (!strcmp(key, "stereo") && apar) {
582 apar->channels = num_val + 1;
583 apar->channel_layout = apar->channels == 2 ?
584 AV_CH_LAYOUT_STEREO :
586 } else if (!strcmp(key, "width") && vpar) {
587 vpar->width = num_val;
588 } else if (!strcmp(key, "height") && vpar) {
589 vpar->height = num_val;
593 if (amf_type == AMF_DATA_TYPE_STRING) {
594 if (!strcmp(key, "encoder")) {
596 if (1 == sscanf(str_val, "Open Broadcaster Software v0.%d", &version)) {
597 if (version > 0 && version <= 655)
598 flv->broken_sizes = 1;
600 } else if (!strcmp(key, "metadatacreator") && !strcmp(str_val, "MEGA")) {
601 flv->broken_sizes = 1;
606 if (amf_type == AMF_DATA_TYPE_OBJECT && s->nb_streams == 1 &&
607 ((!apar && !strcmp(key, "audiocodecid")) ||
608 (!vpar && !strcmp(key, "videocodecid"))))
609 s->ctx_flags &= ~AVFMTCTX_NOHEADER; //If there is either audio/video missing, codecid will be an empty object
611 if (!strcmp(key, "duration") ||
612 !strcmp(key, "filesize") ||
613 !strcmp(key, "width") ||
614 !strcmp(key, "height") ||
615 !strcmp(key, "videodatarate") ||
616 !strcmp(key, "framerate") ||
617 !strcmp(key, "videocodecid") ||
618 !strcmp(key, "audiodatarate") ||
619 !strcmp(key, "audiosamplerate") ||
620 !strcmp(key, "audiosamplesize") ||
621 !strcmp(key, "stereo") ||
622 !strcmp(key, "audiocodecid") ||
623 !strcmp(key, "datastream"))
626 s->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
627 if (amf_type == AMF_DATA_TYPE_BOOL) {
628 av_strlcpy(str_val, num_val > 0 ? "true" : "false",
630 av_dict_set(&s->metadata, key, str_val, 0);
631 } else if (amf_type == AMF_DATA_TYPE_NUMBER) {
632 snprintf(str_val, sizeof(str_val), "%.f", num_val);
633 av_dict_set(&s->metadata, key, str_val, 0);
634 } else if (amf_type == AMF_DATA_TYPE_STRING)
635 av_dict_set(&s->metadata, key, str_val, 0);
641 #define TYPE_ONTEXTDATA 1
642 #define TYPE_ONCAPTION 2
643 #define TYPE_ONCAPTIONINFO 3
644 #define TYPE_UNKNOWN 9
646 static int flv_read_metabody(AVFormatContext *s, int64_t next_pos)
648 FLVContext *flv = s->priv_data;
650 AVStream *stream, *astream, *vstream;
651 AVStream av_unused *dstream;
654 // only needs to hold the string "onMetaData".
655 // Anything longer is something we don't want.
663 // first object needs to be "onMetaData" string
665 if (type != AMF_DATA_TYPE_STRING ||
666 amf_get_string(ioc, buffer, sizeof(buffer)) < 0)
669 if (!strcmp(buffer, "onTextData"))
670 return TYPE_ONTEXTDATA;
672 if (!strcmp(buffer, "onCaption"))
673 return TYPE_ONCAPTION;
675 if (!strcmp(buffer, "onCaptionInfo"))
676 return TYPE_ONCAPTIONINFO;
678 if (strcmp(buffer, "onMetaData") && strcmp(buffer, "onCuePoint")) {
679 av_log(s, AV_LOG_DEBUG, "Unknown type %s\n", buffer);
683 // find the streams now so that amf_parse_object doesn't need to do
684 // the lookup every time it is called.
685 for (i = 0; i < s->nb_streams; i++) {
686 stream = s->streams[i];
687 if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
689 flv->last_keyframe_stream_index = i;
690 } else if (stream->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
692 if (flv->last_keyframe_stream_index == -1)
693 flv->last_keyframe_stream_index = i;
695 else if (stream->codecpar->codec_type == AVMEDIA_TYPE_SUBTITLE)
699 // parse the second object (we want a mixed array)
700 if (amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
706 static int flv_read_header(AVFormatContext *s)
709 FLVContext *flv = s->priv_data;
713 flags = avio_r8(s->pb);
715 flv->missing_streams = flags & (FLV_HEADER_FLAG_HASVIDEO | FLV_HEADER_FLAG_HASAUDIO);
717 s->ctx_flags |= AVFMTCTX_NOHEADER;
719 offset = avio_rb32(s->pb);
720 avio_seek(s->pb, offset, SEEK_SET);
724 flv->sum_flv_tag_size = 0;
725 flv->last_keyframe_stream_index = -1;
730 static int flv_read_close(AVFormatContext *s)
733 FLVContext *flv = s->priv_data;
734 for (i=0; i<FLV_STREAM_TYPE_NB; i++)
735 av_freep(&flv->new_extradata[i]);
736 av_freep(&flv->keyframe_times);
737 av_freep(&flv->keyframe_filepositions);
741 static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
743 av_freep(&st->codecpar->extradata);
744 if (ff_get_extradata(s, st->codecpar, s->pb, size) < 0)
745 return AVERROR(ENOMEM);
749 static int flv_queue_extradata(FLVContext *flv, AVIOContext *pb, int stream,
752 av_free(flv->new_extradata[stream]);
753 flv->new_extradata[stream] = av_mallocz(size +
754 AV_INPUT_BUFFER_PADDING_SIZE);
755 if (!flv->new_extradata[stream])
756 return AVERROR(ENOMEM);
757 flv->new_extradata_size[stream] = size;
758 avio_read(pb, flv->new_extradata[stream], size);
762 static void clear_index_entries(AVFormatContext *s, int64_t pos)
765 av_log(s, AV_LOG_WARNING,
766 "Found invalid index entries, clearing the index.\n");
767 for (i = 0; i < s->nb_streams; i++) {
768 AVStream *st = s->streams[i];
769 /* Remove all index entries that point to >= pos */
771 for (j = 0; j < st->nb_index_entries; j++)
772 if (st->index_entries[j].pos < pos)
773 st->index_entries[out++] = st->index_entries[j];
774 st->nb_index_entries = out;
778 static int amf_skip_tag(AVIOContext *pb, AMFDataType type)
780 int nb = -1, ret, parse_name = 1;
783 case AMF_DATA_TYPE_NUMBER:
786 case AMF_DATA_TYPE_BOOL:
789 case AMF_DATA_TYPE_STRING:
790 avio_skip(pb, avio_rb16(pb));
792 case AMF_DATA_TYPE_ARRAY:
794 case AMF_DATA_TYPE_MIXEDARRAY:
796 case AMF_DATA_TYPE_OBJECT:
797 while(!pb->eof_reached && (nb-- > 0 || type != AMF_DATA_TYPE_ARRAY)) {
799 int size = avio_rb16(pb);
806 if ((ret = amf_skip_tag(pb, avio_r8(pb))) < 0)
810 case AMF_DATA_TYPE_NULL:
811 case AMF_DATA_TYPE_OBJECT_END:
814 return AVERROR_INVALIDDATA;
819 static int flv_data_packet(AVFormatContext *s, AVPacket *pkt,
820 int64_t dts, int64_t next)
822 AVIOContext *pb = s->pb;
825 int ret = AVERROR_INVALIDDATA;
829 switch (avio_r8(pb)) {
830 case AMF_DATA_TYPE_ARRAY:
832 case AMF_DATA_TYPE_MIXEDARRAY:
833 avio_seek(pb, 4, SEEK_CUR);
834 case AMF_DATA_TYPE_OBJECT:
840 while (array || (ret = amf_get_string(pb, buf, sizeof(buf))) > 0) {
841 AMFDataType type = avio_r8(pb);
842 if (type == AMF_DATA_TYPE_STRING && (array || !strcmp(buf, "text"))) {
843 length = avio_rb16(pb);
844 ret = av_get_packet(pb, pkt, length);
850 if ((ret = amf_skip_tag(pb, type)) < 0)
856 ret = AVERROR_INVALIDDATA;
860 for (i = 0; i < s->nb_streams; i++) {
862 if (st->codecpar->codec_type == AVMEDIA_TYPE_SUBTITLE)
866 if (i == s->nb_streams) {
867 st = create_stream(s, AVMEDIA_TYPE_SUBTITLE);
869 return AVERROR(ENOMEM);
870 st->codecpar->codec_id = AV_CODEC_ID_TEXT;
877 pkt->stream_index = st->index;
878 pkt->flags |= AV_PKT_FLAG_KEY;
881 avio_seek(s->pb, next + 4, SEEK_SET);
886 static int resync(AVFormatContext *s)
888 FLVContext *flv = s->priv_data;
890 int64_t pos = avio_tell(s->pb);
892 for (i=0; !avio_feof(s->pb); i++) {
893 int j = i & (RESYNC_BUFFER_SIZE-1);
894 int j1 = j + RESYNC_BUFFER_SIZE;
895 flv->resync_buffer[j ] =
896 flv->resync_buffer[j1] = avio_r8(s->pb);
899 unsigned lsize2 = AV_RB32(flv->resync_buffer + j1 - 4);
900 if (lsize2 >= 11 && lsize2 + 8LL < FFMIN(i, RESYNC_BUFFER_SIZE)) {
901 unsigned size2 = AV_RB24(flv->resync_buffer + j1 - lsize2 + 1 - 4);
902 unsigned lsize1 = AV_RB32(flv->resync_buffer + j1 - lsize2 - 8);
903 if (lsize1 >= 11 && lsize1 + 8LL + lsize2 < FFMIN(i, RESYNC_BUFFER_SIZE)) {
904 unsigned size1 = AV_RB24(flv->resync_buffer + j1 - lsize1 + 1 - lsize2 - 8);
905 if (size1 == lsize1 - 11 && size2 == lsize2 - 11) {
906 avio_seek(s->pb, pos + i - lsize1 - lsize2 - 8, SEEK_SET);
916 static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
918 FLVContext *flv = s->priv_data;
919 int ret, i, size, flags;
920 enum FlvTagType type;
922 int64_t next, pos, meta_pos;
923 int64_t dts, pts = AV_NOPTS_VALUE;
924 int av_uninit(channels);
925 int av_uninit(sample_rate);
931 /* pkt size is repeated at end. skip it */
932 pos = avio_tell(s->pb);
933 type = (avio_r8(s->pb) & 0x1F);
935 size = avio_rb24(s->pb);
936 flv->sum_flv_tag_size += size + 11;
937 dts = avio_rb24(s->pb);
938 dts |= (unsigned)avio_r8(s->pb) << 24;
939 av_log(s, AV_LOG_TRACE, "type:%d, size:%d, last:%d, dts:%"PRId64" pos:%"PRId64"\n", type, size, last, dts, avio_tell(s->pb));
940 if (avio_feof(s->pb))
942 avio_skip(s->pb, 3); /* stream id, always 0 */
945 if (flv->validate_next < flv->validate_count) {
946 int64_t validate_pos = flv->validate_index[flv->validate_next].pos;
947 if (pos == validate_pos) {
948 if (FFABS(dts - flv->validate_index[flv->validate_next].dts) <=
949 VALIDATE_INDEX_TS_THRESH) {
950 flv->validate_next++;
952 clear_index_entries(s, validate_pos);
953 flv->validate_count = 0;
955 } else if (pos > validate_pos) {
956 clear_index_entries(s, validate_pos);
957 flv->validate_count = 0;
966 next = size + avio_tell(s->pb);
968 if (type == FLV_TAG_TYPE_AUDIO) {
969 stream_type = FLV_STREAM_TYPE_AUDIO;
970 flags = avio_r8(s->pb);
972 } else if (type == FLV_TAG_TYPE_VIDEO) {
973 stream_type = FLV_STREAM_TYPE_VIDEO;
974 flags = avio_r8(s->pb);
976 if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_VIDEO_INFO_CMD)
978 } else if (type == FLV_TAG_TYPE_META) {
979 stream_type=FLV_STREAM_TYPE_DATA;
980 if (size > 13 + 1 + 4) { // Header-type metadata stuff
982 meta_pos = avio_tell(s->pb);
983 type = flv_read_metabody(s, next);
984 if (type == 0 && dts == 0 || type < 0 || type == TYPE_UNKNOWN) {
985 if (type < 0 && flv->validate_count &&
986 flv->validate_index[0].pos > next &&
987 flv->validate_index[0].pos - 4 < next
989 av_log(s, AV_LOG_WARNING, "Adjusting next position due to index mismatch\n");
990 next = flv->validate_index[0].pos - 4;
993 } else if (type == TYPE_ONTEXTDATA) {
994 avpriv_request_sample(s, "OnTextData packet");
995 return flv_data_packet(s, pkt, dts, next);
996 } else if (type == TYPE_ONCAPTION) {
997 return flv_data_packet(s, pkt, dts, next);
999 avio_seek(s->pb, meta_pos, SEEK_SET);
1002 av_log(s, AV_LOG_DEBUG,
1003 "Skipping flv packet: type %d, size %d, flags %d.\n",
1006 avio_seek(s->pb, next, SEEK_SET);
1011 /* skip empty data packets */
1017 /* now find stream */
1018 for (i = 0; i < s->nb_streams; i++) {
1020 if (stream_type == FLV_STREAM_TYPE_AUDIO) {
1021 if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO &&
1022 (s->audio_codec_id || flv_same_audio_codec(st->codecpar, flags)))
1024 } else if (stream_type == FLV_STREAM_TYPE_VIDEO) {
1025 if (st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO &&
1026 (s->video_codec_id || flv_same_video_codec(st->codecpar, flags)))
1028 } else if (stream_type == FLV_STREAM_TYPE_DATA) {
1029 if (st->codecpar->codec_type == AVMEDIA_TYPE_SUBTITLE)
1033 if (i == s->nb_streams) {
1034 static const enum AVMediaType stream_types[] = {AVMEDIA_TYPE_VIDEO, AVMEDIA_TYPE_AUDIO, AVMEDIA_TYPE_SUBTITLE};
1035 av_log(s, AV_LOG_WARNING, "%s stream discovered after head already parsed\n", av_get_media_type_string(stream_types[stream_type]));
1036 st = create_stream(s, stream_types[stream_type]);
1038 return AVERROR(ENOMEM);
1041 av_log(s, AV_LOG_TRACE, "%d %X %d \n", stream_type, flags, st->discard);
1043 if (s->pb->seekable &&
1044 ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY ||
1045 stream_type == FLV_STREAM_TYPE_AUDIO))
1046 av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
1048 if ( (st->discard >= AVDISCARD_NONKEY && !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY || (stream_type == FLV_STREAM_TYPE_AUDIO)))
1049 ||(st->discard >= AVDISCARD_BIDIR && ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && (stream_type == FLV_STREAM_TYPE_VIDEO)))
1050 || st->discard >= AVDISCARD_ALL
1052 avio_seek(s->pb, next, SEEK_SET);
1057 // if not streamed and no duration from metadata then seek to end to find
1058 // the duration from the timestamps
1059 if (s->pb->seekable && (!s->duration || s->duration == AV_NOPTS_VALUE) &&
1060 !flv->searched_for_end) {
1062 const int64_t pos = avio_tell(s->pb);
1063 // Read the last 4 bytes of the file, this should be the size of the
1064 // previous FLV tag. Use the timestamp of its payload as duration.
1065 int64_t fsize = avio_size(s->pb);
1067 avio_seek(s->pb, fsize - 4, SEEK_SET);
1068 size = avio_rb32(s->pb);
1069 if (size > 0 && size < fsize) {
1070 // Seek to the start of the last FLV tag at position (fsize - 4 - size)
1071 // but skip the byte indicating the type.
1072 avio_seek(s->pb, fsize - 3 - size, SEEK_SET);
1073 if (size == avio_rb24(s->pb) + 11) {
1074 uint32_t ts = avio_rb24(s->pb);
1075 ts |= avio_r8(s->pb) << 24;
1077 s->duration = ts * (int64_t)AV_TIME_BASE / 1000;
1078 else if (fsize >= 8 && fsize - 8 >= size) {
1080 goto retry_duration;
1085 avio_seek(s->pb, pos, SEEK_SET);
1086 flv->searched_for_end = 1;
1089 if (stream_type == FLV_STREAM_TYPE_AUDIO) {
1090 int bits_per_coded_sample;
1091 channels = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
1092 sample_rate = 44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >>
1093 FLV_AUDIO_SAMPLERATE_OFFSET) >> 3;
1094 bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
1095 if (!st->codecpar->channels || !st->codecpar->sample_rate ||
1096 !st->codecpar->bits_per_coded_sample) {
1097 st->codecpar->channels = channels;
1098 st->codecpar->channel_layout = channels == 1
1100 : AV_CH_LAYOUT_STEREO;
1101 st->codecpar->sample_rate = sample_rate;
1102 st->codecpar->bits_per_coded_sample = bits_per_coded_sample;
1104 if (!st->codecpar->codec_id) {
1105 flv_set_audio_codec(s, st, st->codecpar,
1106 flags & FLV_AUDIO_CODECID_MASK);
1107 flv->last_sample_rate =
1108 sample_rate = st->codecpar->sample_rate;
1109 flv->last_channels =
1110 channels = st->codecpar->channels;
1112 AVCodecParameters *par = avcodec_parameters_alloc();
1114 ret = AVERROR(ENOMEM);
1117 par->sample_rate = sample_rate;
1118 par->bits_per_coded_sample = bits_per_coded_sample;
1119 flv_set_audio_codec(s, st, par, flags & FLV_AUDIO_CODECID_MASK);
1120 sample_rate = par->sample_rate;
1121 avcodec_parameters_free(&par);
1123 } else if (stream_type == FLV_STREAM_TYPE_VIDEO) {
1124 int ret = flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK, 1);
1128 } else if (stream_type == FLV_STREAM_TYPE_DATA) {
1129 st->codecpar->codec_id = AV_CODEC_ID_TEXT;
1132 if (st->codecpar->codec_id == AV_CODEC_ID_AAC ||
1133 st->codecpar->codec_id == AV_CODEC_ID_H264 ||
1134 st->codecpar->codec_id == AV_CODEC_ID_MPEG4) {
1135 int type = avio_r8(s->pb);
1137 if (st->codecpar->codec_id == AV_CODEC_ID_H264 || st->codecpar->codec_id == AV_CODEC_ID_MPEG4) {
1139 int32_t cts = (avio_rb24(s->pb) + 0xff800000) ^ 0xff800000;
1141 if (cts < 0) { // dts might be wrong
1142 if (!flv->wrong_dts)
1143 av_log(s, AV_LOG_WARNING,
1144 "Negative cts, previous timestamps might be wrong.\n");
1146 } else if (FFABS(dts - pts) > 1000*60*15) {
1147 av_log(s, AV_LOG_WARNING,
1148 "invalid timestamps %"PRId64" %"PRId64"\n", dts, pts);
1149 dts = pts = AV_NOPTS_VALUE;
1152 if (type == 0 && (!st->codecpar->extradata || st->codecpar->codec_id == AV_CODEC_ID_AAC ||
1153 st->codecpar->codec_id == AV_CODEC_ID_H264)) {
1154 AVDictionaryEntry *t;
1156 if (st->codecpar->extradata) {
1157 if ((ret = flv_queue_extradata(flv, s->pb, stream_type, size)) < 0)
1162 if ((ret = flv_get_extradata(s, st, size)) < 0)
1165 /* Workaround for buggy Omnia A/XE encoder */
1166 t = av_dict_get(s->metadata, "Encoder", NULL, 0);
1167 if (st->codecpar->codec_id == AV_CODEC_ID_AAC && t && !strcmp(t->value, "Omnia A/XE"))
1168 st->codecpar->extradata_size = 2;
1170 if (st->codecpar->codec_id == AV_CODEC_ID_AAC && 0) {
1171 MPEG4AudioConfig cfg;
1173 if (avpriv_mpeg4audio_get_config(&cfg, st->codecpar->extradata,
1174 st->codecpar->extradata_size * 8, 1) >= 0) {
1175 st->codecpar->channels = cfg.channels;
1176 st->codecpar->channel_layout = 0;
1177 if (cfg.ext_sample_rate)
1178 st->codecpar->sample_rate = cfg.ext_sample_rate;
1180 st->codecpar->sample_rate = cfg.sample_rate;
1181 av_log(s, AV_LOG_TRACE, "mp4a config channels %d sample rate %d\n",
1182 st->codecpar->channels, st->codecpar->sample_rate);
1191 /* skip empty data packets */
1197 ret = av_get_packet(s->pb, pkt, size);
1201 pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
1202 pkt->stream_index = st->index;
1204 if (flv->new_extradata[stream_type]) {
1205 uint8_t *side = av_packet_new_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA,
1206 flv->new_extradata_size[stream_type]);
1208 memcpy(side, flv->new_extradata[stream_type],
1209 flv->new_extradata_size[stream_type]);
1210 av_freep(&flv->new_extradata[stream_type]);
1211 flv->new_extradata_size[stream_type] = 0;
1214 if (stream_type == FLV_STREAM_TYPE_AUDIO &&
1215 (sample_rate != flv->last_sample_rate ||
1216 channels != flv->last_channels)) {
1217 flv->last_sample_rate = sample_rate;
1218 flv->last_channels = channels;
1219 ff_add_param_change(pkt, channels, 0, sample_rate, 0, 0);
1222 if ( stream_type == FLV_STREAM_TYPE_AUDIO ||
1223 ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY) ||
1224 stream_type == FLV_STREAM_TYPE_DATA)
1225 pkt->flags |= AV_PKT_FLAG_KEY;
1228 last = avio_rb32(s->pb);
1229 if (last != orig_size + 11 && last != orig_size + 10 &&
1230 !avio_feof(s->pb) &&
1231 (last != orig_size || !last) && last != flv->sum_flv_tag_size &&
1232 !flv->broken_sizes) {
1233 av_log(s, AV_LOG_ERROR, "Packet mismatch %d %d %d\n", last, orig_size + 11, flv->sum_flv_tag_size);
1234 avio_seek(s->pb, pos + 1, SEEK_SET);
1236 av_packet_unref(pkt);
1244 static int flv_read_seek(AVFormatContext *s, int stream_index,
1245 int64_t ts, int flags)
1247 FLVContext *flv = s->priv_data;
1248 flv->validate_count = 0;
1249 return avio_seek_time(s->pb, stream_index, ts, flags);
1252 #define OFFSET(x) offsetof(FLVContext, x)
1253 #define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM
1254 static const AVOption options[] = {
1255 { "flv_metadata", "Allocate streams according to the onMetaData array", OFFSET(trust_metadata), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VD },
1256 { "missing_streams", "", OFFSET(missing_streams), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 0xFF, VD | AV_OPT_FLAG_EXPORT | AV_OPT_FLAG_READONLY },
1260 static const AVClass flv_class = {
1261 .class_name = "flvdec",
1262 .item_name = av_default_item_name,
1264 .version = LIBAVUTIL_VERSION_INT,
1267 AVInputFormat ff_flv_demuxer = {
1269 .long_name = NULL_IF_CONFIG_SMALL("FLV (Flash Video)"),
1270 .priv_data_size = sizeof(FLVContext),
1271 .read_probe = flv_probe,
1272 .read_header = flv_read_header,
1273 .read_packet = flv_read_packet,
1274 .read_seek = flv_read_seek,
1275 .read_close = flv_read_close,
1276 .extensions = "flv",
1277 .priv_class = &flv_class,
1280 static const AVClass live_flv_class = {
1281 .class_name = "live_flvdec",
1282 .item_name = av_default_item_name,
1284 .version = LIBAVUTIL_VERSION_INT,
1287 AVInputFormat ff_live_flv_demuxer = {
1289 .long_name = NULL_IF_CONFIG_SMALL("live RTMP FLV (Flash Video)"),
1290 .priv_data_size = sizeof(FLVContext),
1291 .read_probe = live_flv_probe,
1292 .read_header = flv_read_header,
1293 .read_packet = flv_read_packet,
1294 .read_seek = flv_read_seek,
1295 .read_close = flv_read_close,
1296 .extensions = "flv",
1297 .priv_class = &live_flv_class,
1298 .flags = AVFMT_TS_DISCONT