libavformat/flvdec.c

   1 /*
   2  * FLV demuxer
   3  * Copyright (c) 2003 The Libav 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 Libav.
  11  *
  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.
  16  *
  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.
  21  *
  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
  25  */
  26
  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"
  35 #include "avformat.h"
  36 #include "internal.h"
  37 #include "avio_internal.h"
  38 #include "flv.h"
  39
  40 #define KEYFRAMES_TAG            "keyframes"
  41 #define KEYFRAMES_TIMESTAMP_TAG  "times"
  42 #define KEYFRAMES_BYTEOFFSET_TAG "filepositions"
  43
  44 #define VALIDATE_INDEX_TS_THRESH 2500
  45
  46 typedef struct {
  47     const AVClass *class; ///< Class for private options.
  48     int trust_metadata; ///< configure streams according onMetaData
  49     int wrong_dts; ///< wrong dts due to negative cts
  50     uint8_t *new_extradata[2];
  51     int      new_extradata_size[2];
  52     int      last_sample_rate;
  53     int      last_channels;
  54     struct {
  55         int64_t dts;
  56         int64_t pos;
  57     } validate_index[2];
  58     int validate_next;
  59     int validate_count;
  60 } FLVContext;
  61
  62 static int flv_probe(AVProbeData *p)
  63 {
  64     const uint8_t *d;
  65
  66     d = p->buf;
  67     if (d[0] == 'F' && d[1] == 'L' && d[2] == 'V' && d[3] < 5 && d[5]==0 && AV_RB32(d+5)>8) {
  68         return AVPROBE_SCORE_MAX;
  69     }
  70     return 0;
  71 }
  72
  73 static AVStream *create_stream(AVFormatContext *s, int codec_type)
  74 {
  75     AVStream *st = avformat_new_stream(s, NULL);
  76     if (!st)
  77         return NULL;
  78     st->codec->codec_type = codec_type;
  79     avpriv_set_pts_info(st, 32, 1, 1000); /* 32 bit pts in ms */
  80     return st;
  81 }
  82 static int flv_same_audio_codec(AVCodecContext *acodec, int flags)
  83 {
  84     int bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
  85     int flv_codecid = flags & FLV_AUDIO_CODECID_MASK;
  86     int codec_id;
  87
  88     if (!acodec->codec_id && !acodec->codec_tag)
  89         return 1;
  90
  91     if (acodec->bits_per_coded_sample != bits_per_coded_sample)
  92         return 0;
  93
  94     switch(flv_codecid) {
  95         //no distinction between S16 and S8 PCM codec flags
  96     case FLV_CODECID_PCM:
  97         codec_id = bits_per_coded_sample == 8 ? AV_CODEC_ID_PCM_U8 :
  98 #if HAVE_BIGENDIAN
  99                             AV_CODEC_ID_PCM_S16BE;
 100 #else
 101                             AV_CODEC_ID_PCM_S16LE;
 102 #endif
 103         return codec_id == acodec->codec_id;
 104     case FLV_CODECID_PCM_LE:
 105         codec_id = bits_per_coded_sample == 8 ? AV_CODEC_ID_PCM_U8 : AV_CODEC_ID_PCM_S16LE;
 106         return codec_id == acodec->codec_id;
 107     case FLV_CODECID_AAC:
 108         return acodec->codec_id == AV_CODEC_ID_AAC;
 109     case FLV_CODECID_ADPCM:
 110         return acodec->codec_id == AV_CODEC_ID_ADPCM_SWF;
 111     case FLV_CODECID_SPEEX:
 112         return acodec->codec_id == AV_CODEC_ID_SPEEX;
 113     case FLV_CODECID_MP3:
 114         return acodec->codec_id == AV_CODEC_ID_MP3;
 115     case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
 116     case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
 117     case FLV_CODECID_NELLYMOSER:
 118         return acodec->codec_id == AV_CODEC_ID_NELLYMOSER;
 119     case FLV_CODECID_PCM_MULAW:
 120         return acodec->sample_rate == 8000 &&
 121                acodec->codec_id == AV_CODEC_ID_PCM_MULAW;
 122     case FLV_CODECID_PCM_ALAW:
 123         return acodec->sample_rate = 8000 &&
 124                acodec->codec_id == AV_CODEC_ID_PCM_ALAW;
 125     default:
 126         return acodec->codec_tag == (flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
 127     }
 128
 129     return 0;
 130 }
 131
 132 static void flv_set_audio_codec(AVFormatContext *s, AVStream *astream, AVCodecContext *acodec, int flv_codecid) {
 133     switch(flv_codecid) {
 134         //no distinction between S16 and S8 PCM codec flags
 135         case FLV_CODECID_PCM:
 136             acodec->codec_id = acodec->bits_per_coded_sample == 8 ? AV_CODEC_ID_PCM_U8 :
 137 #if HAVE_BIGENDIAN
 138                                 AV_CODEC_ID_PCM_S16BE;
 139 #else
 140                                 AV_CODEC_ID_PCM_S16LE;
 141 #endif
 142             break;
 143         case FLV_CODECID_PCM_LE:
 144             acodec->codec_id = acodec->bits_per_coded_sample == 8 ? AV_CODEC_ID_PCM_U8 : AV_CODEC_ID_PCM_S16LE; break;
 145         case FLV_CODECID_AAC  : acodec->codec_id = AV_CODEC_ID_AAC;                                    break;
 146         case FLV_CODECID_ADPCM: acodec->codec_id = AV_CODEC_ID_ADPCM_SWF;                              break;
 147         case FLV_CODECID_SPEEX:
 148             acodec->codec_id = AV_CODEC_ID_SPEEX;
 149             acodec->sample_rate = 16000;
 150             break;
 151         case FLV_CODECID_MP3  : acodec->codec_id = AV_CODEC_ID_MP3      ; astream->need_parsing = AVSTREAM_PARSE_FULL; break;
 152         case FLV_CODECID_NELLYMOSER_8KHZ_MONO:
 153             acodec->sample_rate = 8000; //in case metadata does not otherwise declare samplerate
 154             acodec->codec_id = AV_CODEC_ID_NELLYMOSER;
 155             break;
 156         case FLV_CODECID_NELLYMOSER_16KHZ_MONO:
 157             acodec->sample_rate = 16000;
 158             acodec->codec_id = AV_CODEC_ID_NELLYMOSER;
 159             break;
 160         case FLV_CODECID_NELLYMOSER:
 161             acodec->codec_id = AV_CODEC_ID_NELLYMOSER;
 162             break;
 163         case FLV_CODECID_PCM_MULAW:
 164             acodec->sample_rate = 8000;
 165             acodec->codec_id = AV_CODEC_ID_PCM_MULAW;
 166             break;
 167         case FLV_CODECID_PCM_ALAW:
 168             acodec->sample_rate = 8000;
 169             acodec->codec_id = AV_CODEC_ID_PCM_ALAW;
 170             break;
 171         default:
 172             av_log(s, AV_LOG_INFO, "Unsupported audio codec (%x)\n", flv_codecid >> FLV_AUDIO_CODECID_OFFSET);
 173             acodec->codec_tag = flv_codecid >> FLV_AUDIO_CODECID_OFFSET;
 174     }
 175 }
 176
 177 static int flv_same_video_codec(AVCodecContext *vcodec, int flags)
 178 {
 179     int flv_codecid = flags & FLV_VIDEO_CODECID_MASK;
 180
 181     if (!vcodec->codec_id && !vcodec->codec_tag)
 182         return 1;
 183
 184     switch (flv_codecid) {
 185         case FLV_CODECID_H263:
 186             return vcodec->codec_id == AV_CODEC_ID_FLV1;
 187         case FLV_CODECID_SCREEN:
 188             return vcodec->codec_id == AV_CODEC_ID_FLASHSV;
 189         case FLV_CODECID_SCREEN2:
 190             return vcodec->codec_id == AV_CODEC_ID_FLASHSV2;
 191         case FLV_CODECID_VP6:
 192             return vcodec->codec_id == AV_CODEC_ID_VP6F;
 193         case FLV_CODECID_VP6A:
 194             return vcodec->codec_id == AV_CODEC_ID_VP6A;
 195         case FLV_CODECID_H264:
 196             return vcodec->codec_id == AV_CODEC_ID_H264;
 197         default:
 198             return vcodec->codec_tag == flv_codecid;
 199     }
 200
 201     return 0;
 202 }
 203
 204 static int flv_set_video_codec(AVFormatContext *s, AVStream *vstream, int flv_codecid, int read) {
 205     AVCodecContext *vcodec = vstream->codec;
 206     switch(flv_codecid) {
 207         case FLV_CODECID_H263  : vcodec->codec_id = AV_CODEC_ID_FLV1   ; break;
 208         case FLV_CODECID_SCREEN: vcodec->codec_id = AV_CODEC_ID_FLASHSV; break;
 209         case FLV_CODECID_SCREEN2: vcodec->codec_id = AV_CODEC_ID_FLASHSV2; break;
 210         case FLV_CODECID_VP6   : vcodec->codec_id = AV_CODEC_ID_VP6F   ;
 211         case FLV_CODECID_VP6A  :
 212             if(flv_codecid == FLV_CODECID_VP6A)
 213                 vcodec->codec_id = AV_CODEC_ID_VP6A;
 214             if (read) {
 215                 if (vcodec->extradata_size != 1) {
 216                     vcodec->extradata = av_malloc(1);
 217                     if (vcodec->extradata)
 218                         vcodec->extradata_size = 1;
 219                 }
 220                 if (vcodec->extradata)
 221                     vcodec->extradata[0] = avio_r8(s->pb);
 222                 else
 223                     avio_skip(s->pb, 1);
 224             }
 225             return 1; // 1 byte body size adjustment for flv_read_packet()
 226         case FLV_CODECID_H264:
 227             vcodec->codec_id = AV_CODEC_ID_H264;
 228             return 3; // not 4, reading packet type will consume one byte
 229         default:
 230             av_log(s, AV_LOG_INFO, "Unsupported video codec (%x)\n", flv_codecid);
 231             vcodec->codec_tag = flv_codecid;
 232     }
 233
 234     return 0;
 235 }
 236
 237 static int amf_get_string(AVIOContext *ioc, char *buffer, int buffsize) {
 238     int length = avio_rb16(ioc);
 239     if(length >= buffsize) {
 240         avio_skip(ioc, length);
 241         return -1;
 242     }
 243
 244     avio_read(ioc, buffer, length);
 245
 246     buffer[length] = '\0';
 247
 248     return length;
 249 }
 250
 251 static int parse_keyframes_index(AVFormatContext *s, AVIOContext *ioc, AVStream *vstream, int64_t max_pos) {
 252     FLVContext *flv = s->priv_data;
 253     unsigned int arraylen = 0, timeslen = 0, fileposlen = 0, i;
 254     double num_val;
 255     char str_val[256];
 256     int64_t *times = NULL;
 257     int64_t *filepositions = NULL;
 258     int ret = AVERROR(ENOSYS);
 259     int64_t initial_pos = avio_tell(ioc);
 260
 261     if (s->flags & AVFMT_FLAG_IGNIDX)
 262         return 0;
 263
 264     while (avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
 265         int64_t* current_array;
 266
 267         // Expect array object in context
 268         if (avio_r8(ioc) != AMF_DATA_TYPE_ARRAY)
 269             break;
 270
 271         arraylen = avio_rb32(ioc);
 272         if (arraylen >> 28)
 273             break;
 274
 275         /*
 276          * Expect only 'times' or 'filepositions' sub-arrays in other case refuse to use such metadata
 277          * for indexing
 278          */
 279         if (!strcmp(KEYFRAMES_TIMESTAMP_TAG, str_val) && !times) {
 280             if (!(times = av_mallocz(sizeof(*times) * arraylen))) {
 281                 ret = AVERROR(ENOMEM);
 282                 goto finish;
 283             }
 284             timeslen = arraylen;
 285             current_array = times;
 286         } else if (!strcmp(KEYFRAMES_BYTEOFFSET_TAG, str_val) && !filepositions) {
 287             if (!(filepositions = av_mallocz(sizeof(*filepositions) * arraylen))) {
 288                 ret = AVERROR(ENOMEM);
 289                 goto finish;
 290             }
 291             fileposlen = arraylen;
 292             current_array = filepositions;
 293         } else // unexpected metatag inside keyframes, will not use such metadata for indexing
 294             break;
 295
 296         for (i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
 297             if (avio_r8(ioc) != AMF_DATA_TYPE_NUMBER)
 298                 goto finish;
 299             num_val = av_int2double(avio_rb64(ioc));
 300             current_array[i] = num_val;
 301         }
 302         if (times && filepositions) {
 303             // All done, exiting at a position allowing amf_parse_object
 304             // to finish parsing the object
 305             ret = 0;
 306             break;
 307         }
 308     }
 309
 310     if (!ret && timeslen == fileposlen) {
 311         for (i = 0; i < fileposlen; i++) {
 312             av_add_index_entry(vstream, filepositions[i], times[i]*1000,
 313                                0, 0, AVINDEX_KEYFRAME);
 314             if (i < 2) {
 315                 flv->validate_index[i].pos = filepositions[i];
 316                 flv->validate_index[i].dts = times[i] * 1000;
 317                 flv->validate_count = i + 1;
 318             }
 319         }
 320     } else
 321         av_log(s, AV_LOG_WARNING, "Invalid keyframes object, skipping.\n");
 322
 323 finish:
 324     av_freep(&times);
 325     av_freep(&filepositions);
 326     // If we got unexpected data, but successfully reset back to
 327     // the start pos, the caller can continue parsing
 328     if (ret < 0 && avio_seek(ioc, initial_pos, SEEK_SET) > 0)
 329         return 0;
 330     return ret;
 331 }
 332
 333 static int amf_parse_object(AVFormatContext *s, AVStream *astream, AVStream *vstream, const char *key, int64_t max_pos, int depth) {
 334     AVCodecContext *acodec, *vcodec;
 335     FLVContext *flv = s->priv_data;
 336     AVIOContext *ioc;
 337     AMFDataType amf_type;
 338     char str_val[256];
 339     double num_val;
 340
 341     num_val = 0;
 342     ioc = s->pb;
 343
 344     amf_type = avio_r8(ioc);
 345
 346     switch(amf_type) {
 347         case AMF_DATA_TYPE_NUMBER:
 348             num_val = av_int2double(avio_rb64(ioc)); break;
 349         case AMF_DATA_TYPE_BOOL:
 350             num_val = avio_r8(ioc); break;
 351         case AMF_DATA_TYPE_STRING:
 352             if(amf_get_string(ioc, str_val, sizeof(str_val)) < 0)
 353                 return -1;
 354             break;
 355         case AMF_DATA_TYPE_OBJECT:
 356             if ((vstream || astream) && key && !strcmp(KEYFRAMES_TAG, key) && depth == 1)
 357                 if (parse_keyframes_index(s, ioc, vstream ? vstream : astream,
 358                                           max_pos) < 0)
 359                     return -1;
 360
 361             while (avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
 362                 if (amf_parse_object(s, astream, vstream, str_val, max_pos, depth + 1) < 0)
 363                     return -1; //if we couldn't skip, bomb out.
 364             }
 365             if(avio_r8(ioc) != AMF_END_OF_OBJECT)
 366                 return -1;
 367             break;
 368         case AMF_DATA_TYPE_NULL:
 369         case AMF_DATA_TYPE_UNDEFINED:
 370         case AMF_DATA_TYPE_UNSUPPORTED:
 371             break; //these take up no additional space
 372         case AMF_DATA_TYPE_MIXEDARRAY:
 373             avio_skip(ioc, 4); //skip 32-bit max array index
 374             while(avio_tell(ioc) < max_pos - 2 && amf_get_string(ioc, str_val, sizeof(str_val)) > 0) {
 375                 //this is the only case in which we would want a nested parse to not skip over the object
 376                 if(amf_parse_object(s, astream, vstream, str_val, max_pos, depth + 1) < 0)
 377                     return -1;
 378             }
 379             if(avio_r8(ioc) != AMF_END_OF_OBJECT)
 380                 return -1;
 381             break;
 382         case AMF_DATA_TYPE_ARRAY: {
 383             unsigned int arraylen, i;
 384
 385             arraylen = avio_rb32(ioc);
 386             for(i = 0; i < arraylen && avio_tell(ioc) < max_pos - 1; i++) {
 387                 if(amf_parse_object(s, NULL, NULL, NULL, max_pos, depth + 1) < 0)
 388                     return -1; //if we couldn't skip, bomb out.
 389             }
 390         }
 391             break;
 392         case AMF_DATA_TYPE_DATE:
 393             avio_skip(ioc, 8 + 2); //timestamp (double) and UTC offset (int16)
 394             break;
 395         default: //unsupported type, we couldn't skip
 396             return -1;
 397     }
 398
 399     if(depth == 1 && key) { //only look for metadata values when we are not nested and key != NULL
 400         acodec = astream ? astream->codec : NULL;
 401         vcodec = vstream ? vstream->codec : NULL;
 402
 403         if (amf_type == AMF_DATA_TYPE_NUMBER) {
 404             if (!strcmp(key, "duration"))
 405                 s->duration = num_val * AV_TIME_BASE;
 406             else if (!strcmp(key, "videodatarate") && vcodec && 0 <= (int)(num_val * 1024.0))
 407                 vcodec->bit_rate = num_val * 1024.0;
 408             else if (!strcmp(key, "audiodatarate") && acodec && 0 <= (int)(num_val * 1024.0))
 409                 acodec->bit_rate = num_val * 1024.0;
 410             else if (!strcmp(key, "datastream")) {
 411                 AVStream *st = create_stream(s, AVMEDIA_TYPE_DATA);
 412                 if (!st)
 413                     return AVERROR(ENOMEM);
 414                 st->codec->codec_id = AV_CODEC_ID_TEXT;
 415             } else if (flv->trust_metadata) {
 416                 if (!strcmp(key, "videocodecid") && vcodec) {
 417                     flv_set_video_codec(s, vstream, num_val, 0);
 418                 } else
 419                 if (!strcmp(key, "audiocodecid") && acodec) {
 420                     flv_set_audio_codec(s, astream, acodec, num_val);
 421                 } else
 422                 if (!strcmp(key, "audiosamplerate") && acodec) {
 423                     acodec->sample_rate = num_val;
 424                 } else
 425                 if (!strcmp(key, "width") && vcodec) {
 426                     vcodec->width = num_val;
 427                 } else
 428                 if (!strcmp(key, "height") && vcodec) {
 429                     vcodec->height = num_val;
 430                 }
 431             }
 432         }
 433
 434         if (!strcmp(key, "duration")        ||
 435             !strcmp(key, "filesize")        ||
 436             !strcmp(key, "width")           ||
 437             !strcmp(key, "height")          ||
 438             !strcmp(key, "videodatarate")   ||
 439             !strcmp(key, "framerate")       ||
 440             !strcmp(key, "videocodecid")    ||
 441             !strcmp(key, "audiodatarate")   ||
 442             !strcmp(key, "audiosamplerate") ||
 443             !strcmp(key, "audiosamplesize") ||
 444             !strcmp(key, "stereo")          ||
 445             !strcmp(key, "audiocodecid"))
 446             return 0;
 447
 448         if(amf_type == AMF_DATA_TYPE_BOOL) {
 449             av_strlcpy(str_val, num_val > 0 ? "true" : "false", sizeof(str_val));
 450             av_dict_set(&s->metadata, key, str_val, 0);
 451         } else if(amf_type == AMF_DATA_TYPE_NUMBER) {
 452             snprintf(str_val, sizeof(str_val), "%.f", num_val);
 453             av_dict_set(&s->metadata, key, str_val, 0);
 454         } else if (amf_type == AMF_DATA_TYPE_STRING)
 455             av_dict_set(&s->metadata, key, str_val, 0);
 456     }
 457
 458     return 0;
 459 }
 460
 461 static int flv_read_metabody(AVFormatContext *s, int64_t next_pos) {
 462     AMFDataType type;
 463     AVStream *stream, *astream, *vstream;
 464     AVIOContext *ioc;
 465     int i;
 466     char buffer[11]; //only needs to hold the string "onMetaData". Anything longer is something we don't want.
 467
 468     astream = NULL;
 469     vstream = NULL;
 470     ioc = s->pb;
 471
 472     //first object needs to be "onMetaData" string
 473     type = avio_r8(ioc);
 474     if (type != AMF_DATA_TYPE_STRING ||
 475         amf_get_string(ioc, buffer, sizeof(buffer)) < 0)
 476         return -1;
 477
 478     if (!strcmp(buffer, "onTextData"))
 479         return 1;
 480
 481     if (strcmp(buffer, "onMetaData"))
 482         return -1;
 483
 484     //find the streams now so that amf_parse_object doesn't need to do the lookup every time it is called.
 485     for(i = 0; i < s->nb_streams; i++) {
 486         stream = s->streams[i];
 487         if     (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO) astream = stream;
 488         else if(stream->codec->codec_type == AVMEDIA_TYPE_VIDEO) vstream = stream;
 489     }
 490
 491     //parse the second object (we want a mixed array)
 492     if(amf_parse_object(s, astream, vstream, buffer, next_pos, 0) < 0)
 493         return -1;
 494
 495     return 0;
 496 }
 497
 498 static int flv_read_header(AVFormatContext *s)
 499 {
 500     int offset, flags;
 501
 502     avio_skip(s->pb, 4);
 503     flags = avio_r8(s->pb);
 504     /* old flvtool cleared this field */
 505     /* FIXME: better fix needed */
 506     if (!flags) {
 507         flags = FLV_HEADER_FLAG_HASVIDEO | FLV_HEADER_FLAG_HASAUDIO;
 508         av_log(s, AV_LOG_WARNING, "Broken FLV file, which says no streams present, this might fail\n");
 509     }
 510
 511     s->ctx_flags |= AVFMTCTX_NOHEADER;
 512
 513     if(flags & FLV_HEADER_FLAG_HASVIDEO){
 514         if(!create_stream(s, AVMEDIA_TYPE_VIDEO))
 515             return AVERROR(ENOMEM);
 516     }
 517     if(flags & FLV_HEADER_FLAG_HASAUDIO){
 518         if(!create_stream(s, AVMEDIA_TYPE_AUDIO))
 519             return AVERROR(ENOMEM);
 520     }
 521
 522     offset = avio_rb32(s->pb);
 523     avio_seek(s->pb, offset, SEEK_SET);
 524     avio_skip(s->pb, 4);
 525
 526     s->start_time = 0;
 527
 528     return 0;
 529 }
 530
 531 static int flv_read_close(AVFormatContext *s)
 532 {
 533     FLVContext *flv = s->priv_data;
 534     av_freep(&flv->new_extradata[0]);
 535     av_freep(&flv->new_extradata[1]);
 536     return 0;
 537 }
 538
 539 static int flv_get_extradata(AVFormatContext *s, AVStream *st, int size)
 540 {
 541     av_free(st->codec->extradata);
 542     st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
 543     if (!st->codec->extradata)
 544         return AVERROR(ENOMEM);
 545     st->codec->extradata_size = size;
 546     avio_read(s->pb, st->codec->extradata, st->codec->extradata_size);
 547     return 0;
 548 }
 549
 550 static int flv_queue_extradata(FLVContext *flv, AVIOContext *pb, int stream,
 551                                int size)
 552 {
 553     av_free(flv->new_extradata[stream]);
 554     flv->new_extradata[stream] = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
 555     if (!flv->new_extradata[stream])
 556         return AVERROR(ENOMEM);
 557     flv->new_extradata_size[stream] = size;
 558     avio_read(pb, flv->new_extradata[stream], size);
 559     return 0;
 560 }
 561
 562 static void clear_index_entries(AVFormatContext *s, int64_t pos)
 563 {
 564     int i, j, out;
 565     av_log(s, AV_LOG_WARNING, "Found invalid index entries, clearing the index.\n");
 566     for (i = 0; i < s->nb_streams; i++) {
 567         AVStream *st = s->streams[i];
 568         /* Remove all index entries that point to >= pos */
 569         out = 0;
 570         for (j = 0; j < st->nb_index_entries; j++) {
 571             if (st->index_entries[j].pos < pos)
 572                 st->index_entries[out++] = st->index_entries[j];
 573         }
 574         st->nb_index_entries = out;
 575     }
 576 }
 577
 578
 579 static int flv_data_packet(AVFormatContext *s, AVPacket *pkt,
 580                            int64_t dts, int64_t next)
 581 {
 582     int ret = AVERROR_INVALIDDATA, i;
 583     AVIOContext *pb = s->pb;
 584     AVStream *st = NULL;
 585     AMFDataType type;
 586     char buf[20];
 587     int length;
 588
 589     type = avio_r8(pb);
 590     if (type == AMF_DATA_TYPE_MIXEDARRAY)
 591         avio_seek(pb, 4, SEEK_CUR);
 592     else if (type != AMF_DATA_TYPE_OBJECT)
 593         goto out;
 594
 595     amf_get_string(pb, buf, sizeof(buf));
 596     if (strcmp(buf, "type") || avio_r8(pb) != AMF_DATA_TYPE_STRING)
 597         goto out;
 598
 599     amf_get_string(pb, buf, sizeof(buf));
 600     //FIXME parse it as codec_id
 601     amf_get_string(pb, buf, sizeof(buf));
 602     if (strcmp(buf, "text") || avio_r8(pb) != AMF_DATA_TYPE_STRING)
 603         goto out;
 604
 605     length = avio_rb16(pb);
 606     ret = av_get_packet(s->pb, pkt, length);
 607     if (ret < 0) {
 608         ret = AVERROR(EIO);
 609         goto out;
 610     }
 611
 612     for (i = 0; i < s->nb_streams; i++) {
 613         st = s->streams[i];
 614         if (st->codec->codec_type == AVMEDIA_TYPE_DATA)
 615             break;
 616     }
 617
 618     if (i == s->nb_streams) {
 619         st = create_stream(s, AVMEDIA_TYPE_DATA);
 620         if (!st)
 621             goto out;
 622         st->codec->codec_id = AV_CODEC_ID_TEXT;
 623     }
 624
 625     pkt->dts  = dts;
 626     pkt->pts  = dts;
 627     pkt->size = ret;
 628
 629     pkt->stream_index = st->index;
 630     pkt->flags |= AV_PKT_FLAG_KEY;
 631
 632     avio_seek(s->pb, next + 4, SEEK_SET);
 633 out:
 634     return ret;
 635 }
 636
 637 static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
 638 {
 639     FLVContext *flv = s->priv_data;
 640     int ret, i, type, size, flags, is_audio;
 641     int64_t next, pos;
 642     int64_t dts, pts = AV_NOPTS_VALUE;
 643     int sample_rate = 0, channels = 0;
 644     AVStream *st = NULL;
 645
 646  for(;;avio_skip(s->pb, 4)){ /* pkt size is repeated at end. skip it */
 647     pos = avio_tell(s->pb);
 648     type = avio_r8(s->pb);
 649     size = avio_rb24(s->pb);
 650     dts = avio_rb24(s->pb);
 651     dts |= avio_r8(s->pb) << 24;
 652     av_dlog(s, "type:%d, size:%d, dts:%"PRId64"\n", type, size, dts);
 653     if (s->pb->eof_reached)
 654         return AVERROR_EOF;
 655     avio_skip(s->pb, 3); /* stream id, always 0 */
 656     flags = 0;
 657
 658     if (flv->validate_next < flv->validate_count) {
 659         int64_t validate_pos = flv->validate_index[flv->validate_next].pos;
 660         if (pos == validate_pos) {
 661             if (FFABS(dts - flv->validate_index[flv->validate_next].dts) <=
 662                 VALIDATE_INDEX_TS_THRESH) {
 663                 flv->validate_next++;
 664             } else {
 665                 clear_index_entries(s, validate_pos);
 666                 flv->validate_count = 0;
 667             }
 668         } else if (pos > validate_pos) {
 669             clear_index_entries(s, validate_pos);
 670             flv->validate_count = 0;
 671         }
 672     }
 673
 674     if(size == 0)
 675         continue;
 676
 677     next= size + avio_tell(s->pb);
 678
 679     if (type == FLV_TAG_TYPE_AUDIO) {
 680         is_audio=1;
 681         flags = avio_r8(s->pb);
 682         size--;
 683     } else if (type == FLV_TAG_TYPE_VIDEO) {
 684         is_audio=0;
 685         flags = avio_r8(s->pb);
 686         size--;
 687         if ((flags & 0xf0) == 0x50) /* video info / command frame */
 688             goto skip;
 689     } else {
 690         if (type == FLV_TAG_TYPE_META && size > 13+1+4)
 691             if (flv_read_metabody(s, next) > 0) {
 692                 return flv_data_packet(s, pkt, dts, next);
 693             }
 694         else /* skip packet */
 695             av_log(s, AV_LOG_DEBUG, "skipping flv packet: type %d, size %d, flags %d\n", type, size, flags);
 696     skip:
 697         avio_seek(s->pb, next, SEEK_SET);
 698         continue;
 699     }
 700
 701     /* skip empty data packets */
 702     if (!size)
 703         continue;
 704
 705     /* now find stream */
 706     for(i=0;i<s->nb_streams;i++) {
 707         st = s->streams[i];
 708         if (is_audio && st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
 709             if (flv_same_audio_codec(st->codec, flags)) {
 710                 break;
 711             }
 712         } else
 713         if (!is_audio && st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
 714             if (flv_same_video_codec(st->codec, flags)) {
 715                 break;
 716             }
 717         }
 718     }
 719     if(i == s->nb_streams){
 720         st = create_stream(s,
 721             is_audio ? AVMEDIA_TYPE_AUDIO : AVMEDIA_TYPE_VIDEO);
 722     }
 723     av_dlog(s, "%d %X %d \n", is_audio, flags, st->discard);
 724     if(  (st->discard >= AVDISCARD_NONKEY && !((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY ||         is_audio))
 725        ||(st->discard >= AVDISCARD_BIDIR  &&  ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_DISP_INTER && !is_audio))
 726        || st->discard >= AVDISCARD_ALL
 727        ){
 728         avio_seek(s->pb, next, SEEK_SET);
 729         continue;
 730     }
 731     if ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY)
 732         av_add_index_entry(st, pos, dts, size, 0, AVINDEX_KEYFRAME);
 733     break;
 734  }
 735
 736     // if not streamed and no duration from metadata then seek to end to find the duration from the timestamps
 737     if(s->pb->seekable && (!s->duration || s->duration==AV_NOPTS_VALUE)){
 738         int size;
 739         const int64_t pos= avio_tell(s->pb);
 740         const int64_t fsize= avio_size(s->pb);
 741         avio_seek(s->pb, fsize-4, SEEK_SET);
 742         size= avio_rb32(s->pb);
 743         avio_seek(s->pb, fsize-3-size, SEEK_SET);
 744         if(size == avio_rb24(s->pb) + 11){
 745             uint32_t ts = avio_rb24(s->pb);
 746             ts |= avio_r8(s->pb) << 24;
 747             s->duration = ts * (int64_t)AV_TIME_BASE / 1000;
 748         }
 749         avio_seek(s->pb, pos, SEEK_SET);
 750     }
 751
 752     if(is_audio){
 753         int bits_per_coded_sample;
 754         channels    = (flags & FLV_AUDIO_CHANNEL_MASK) == FLV_STEREO ? 2 : 1;
 755         sample_rate = (44100 << ((flags & FLV_AUDIO_SAMPLERATE_MASK) >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3);
 756         bits_per_coded_sample = (flags & FLV_AUDIO_SAMPLESIZE_MASK) ? 16 : 8;
 757         if(!st->codec->channels || !st->codec->sample_rate || !st->codec->bits_per_coded_sample) {
 758             st->codec->channels              = channels;
 759             st->codec->channel_layout        = channels == 1 ? AV_CH_LAYOUT_MONO :
 760                                                                AV_CH_LAYOUT_STEREO;
 761             st->codec->sample_rate           = sample_rate;
 762             st->codec->bits_per_coded_sample = bits_per_coded_sample;
 763         }
 764         if(!st->codec->codec_id){
 765             flv_set_audio_codec(s, st, st->codec, flags & FLV_AUDIO_CODECID_MASK);
 766             flv->last_sample_rate = sample_rate = st->codec->sample_rate;
 767             flv->last_channels    = channels    = st->codec->channels;
 768         } else {
 769             AVCodecContext ctx;
 770             ctx.sample_rate = sample_rate;
 771             flv_set_audio_codec(s, st, &ctx, flags & FLV_AUDIO_CODECID_MASK);
 772             sample_rate = ctx.sample_rate;
 773         }
 774     }else{
 775         size -= flv_set_video_codec(s, st, flags & FLV_VIDEO_CODECID_MASK, 1);
 776     }
 777
 778     if (st->codec->codec_id == AV_CODEC_ID_AAC ||
 779         st->codec->codec_id == AV_CODEC_ID_H264) {
 780         int type = avio_r8(s->pb);
 781         size--;
 782         if (st->codec->codec_id == AV_CODEC_ID_H264) {
 783             int32_t cts = (avio_rb24(s->pb)+0xff800000)^0xff800000; // sign extension
 784             pts = dts + cts;
 785             if (cts < 0) { // dts are wrong
 786                 flv->wrong_dts = 1;
 787                 av_log(s, AV_LOG_WARNING, "negative cts, previous timestamps might be wrong\n");
 788             }
 789             if (flv->wrong_dts)
 790                 dts = AV_NOPTS_VALUE;
 791         }
 792         if (type == 0) {
 793             if (st->codec->extradata) {
 794                 if ((ret = flv_queue_extradata(flv, s->pb, is_audio, size)) < 0)
 795                     return ret;
 796                 ret = AVERROR(EAGAIN);
 797                 goto leave;
 798             }
 799             if ((ret = flv_get_extradata(s, st, size)) < 0)
 800                 return ret;
 801             if (st->codec->codec_id == AV_CODEC_ID_AAC) {
 802                 MPEG4AudioConfig cfg;
 803                 avpriv_mpeg4audio_get_config(&cfg, st->codec->extradata,
 804                                              st->codec->extradata_size * 8, 1);
 805                 st->codec->channels = cfg.channels;
 806                 st->codec->channel_layout = 0;
 807                 if (cfg.ext_sample_rate)
 808                     st->codec->sample_rate = cfg.ext_sample_rate;
 809                 else
 810                     st->codec->sample_rate = cfg.sample_rate;
 811                 av_dlog(s, "mp4a config channels %d sample rate %d\n",
 812                         st->codec->channels, st->codec->sample_rate);
 813             }
 814
 815             ret = AVERROR(EAGAIN);
 816             goto leave;
 817         }
 818     }
 819
 820     /* skip empty data packets */
 821     if (!size) {
 822         ret = AVERROR(EAGAIN);
 823         goto leave;
 824     }
 825
 826     ret= av_get_packet(s->pb, pkt, size);
 827     if (ret < 0) {
 828         return AVERROR(EIO);
 829     }
 830     /* note: we need to modify the packet size here to handle the last
 831        packet */
 832     pkt->size = ret;
 833     pkt->dts = dts;
 834     pkt->pts = pts == AV_NOPTS_VALUE ? dts : pts;
 835     pkt->stream_index = st->index;
 836     if (flv->new_extradata[is_audio]) {
 837         uint8_t *side = av_packet_new_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA,
 838                                                 flv->new_extradata_size[is_audio]);
 839         if (side) {
 840             memcpy(side, flv->new_extradata[is_audio],
 841                    flv->new_extradata_size[is_audio]);
 842             av_freep(&flv->new_extradata[is_audio]);
 843             flv->new_extradata_size[is_audio] = 0;
 844         }
 845     }
 846     if (is_audio && (sample_rate != flv->last_sample_rate ||
 847                      channels != flv->last_channels)) {
 848         flv->last_sample_rate = sample_rate;
 849         flv->last_channels    = channels;
 850         ff_add_param_change(pkt, channels, 0, sample_rate, 0, 0);
 851     }
 852
 853     if (is_audio || ((flags & FLV_VIDEO_FRAMETYPE_MASK) == FLV_FRAME_KEY))
 854         pkt->flags |= AV_PKT_FLAG_KEY;
 855
 856 leave:
 857     avio_skip(s->pb, 4);
 858     return ret;
 859 }
 860
 861 static int flv_read_seek(AVFormatContext *s, int stream_index,
 862     int64_t ts, int flags)
 863 {
 864     FLVContext *flv = s->priv_data;
 865     flv->validate_count = 0;
 866     return avio_seek_time(s->pb, stream_index, ts, flags);
 867 }
 868
 869 #define OFFSET(x) offsetof(FLVContext, x)
 870 #define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM
 871 static const AVOption options[] = {
 872     { "flv_metadata", "Allocate streams according the onMetaData array",      OFFSET(trust_metadata), AV_OPT_TYPE_INT,    { .i64 = 0 }, 0, 1, VD},
 873     { NULL }
 874 };
 875
 876 static const AVClass class = {
 877     .class_name = "flvdec",
 878     .item_name  = av_default_item_name,
 879     .option     = options,
 880     .version    = LIBAVUTIL_VERSION_INT,
 881 };
 882
 883 AVInputFormat ff_flv_demuxer = {
 884     .name           = "flv",
 885     .long_name      = NULL_IF_CONFIG_SMALL("FLV (Flash Video)"),
 886     .priv_data_size = sizeof(FLVContext),
 887     .read_probe     = flv_probe,
 888     .read_header    = flv_read_header,
 889     .read_packet    = flv_read_packet,
 890     .read_seek      = flv_read_seek,
 891     .read_close     = flv_read_close,
 892     .extensions     = "flv",
 893     .priv_class     = &class,
 894 };