libavformat/oggparsevorbis.c

   1 /*
   2  * Copyright (C) 2005  Michael Ahlberg, Måns Rullgård
   3  *
   4  * Permission is hereby granted, free of charge, to any person
   5  * obtaining a copy of this software and associated documentation
   6  * files (the "Software"), to deal in the Software without
   7  * restriction, including without limitation the rights to use, copy,
   8  * modify, merge, publish, distribute, sublicense, and/or sell copies
   9  * of the Software, and to permit persons to whom the Software is
  10  * furnished to do so, subject to the following conditions:
  11  *
  12  *  The above copyright notice and this permission notice shall be
  13  *  included in all copies or substantial portions of the Software.
  14  *
  15  *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  16  *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  17  *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  18  *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
  19  *  HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
  20  *  WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  *  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
  22  *  DEALINGS IN THE SOFTWARE.
  23  */
  24
  25 #include <stdlib.h>
  26
  27 #include "libavutil/avstring.h"
  28 #include "libavutil/bswap.h"
  29 #include "libavutil/dict.h"
  30 #include "libavcodec/bytestream.h"
  31 #include "libavcodec/get_bits.h"
  32 #include "libavcodec/vorbis_parser.h"
  33 #include "avformat.h"
  34 #include "internal.h"
  35 #include "oggdec.h"
  36 #include "vorbiscomment.h"
  37
  38 static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
  39 {
  40     int i, cnum, h, m, s, ms, keylen = strlen(key);
  41     AVChapter *chapter = NULL;
  42
  43     if (keylen < 9 || sscanf(key, "CHAPTER%02d", &cnum) != 1)
  44         return 0;
  45
  46     if (keylen == 9) {
  47         if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
  48             return 0;
  49
  50         avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 },
  51                            ms + 1000 * (s + 60 * (m + 60 * h)),
  52                            AV_NOPTS_VALUE, NULL);
  53         av_free(val);
  54     } else if (!strcmp(key + 9, "NAME")) {
  55         for (i = 0; i < as->nb_chapters; i++)
  56             if (as->chapters[i]->id == cnum) {
  57                 chapter = as->chapters[i];
  58                 break;
  59             }
  60         if (!chapter)
  61             return 0;
  62
  63         av_dict_set(&chapter->metadata, "title", val, AV_DICT_DONT_STRDUP_VAL);
  64     } else
  65         return 0;
  66
  67     av_free(key);
  68     return 1;
  69 }
  70
  71 int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
  72                       const uint8_t *buf, int size)
  73 {
  74     const uint8_t *p   = buf;
  75     const uint8_t *end = buf + size;
  76     unsigned n, j;
  77     int s;
  78
  79     /* must have vendor_length and user_comment_list_length */
  80     if (size < 8)
  81         return AVERROR_INVALIDDATA;
  82
  83     s = bytestream_get_le32(&p);
  84
  85     if (end - p - 4 < s || s < 0)
  86         return AVERROR_INVALIDDATA;
  87
  88     p += s;
  89
  90     n = bytestream_get_le32(&p);
  91
  92     while (end - p >= 4 && n > 0) {
  93         const char *t, *v;
  94         int tl, vl;
  95
  96         s = bytestream_get_le32(&p);
  97
  98         if (end - p < s || s < 0)
  99             break;
 100
 101         t  = p;
 102         p += s;
 103         n--;
 104
 105         v = memchr(t, '=', s);
 106         if (!v)
 107             continue;
 108
 109         tl = v - t;
 110         vl = s - tl - 1;
 111         v++;
 112
 113         if (tl && vl) {
 114             char *tt, *ct;
 115
 116             tt = av_malloc(tl + 1);
 117             ct = av_malloc(vl + 1);
 118             if (!tt || !ct) {
 119                 av_freep(&tt);
 120                 av_freep(&ct);
 121                 av_log(as, AV_LOG_WARNING,
 122                        "out-of-memory error. skipping VorbisComment tag.\n");
 123                 continue;
 124             }
 125
 126             for (j = 0; j < tl; j++)
 127                 tt[j] = av_toupper(t[j]);
 128             tt[tl] = 0;
 129
 130             memcpy(ct, v, vl);
 131             ct[vl] = 0;
 132
 133             if (!ogm_chapter(as, tt, ct))
 134                 av_dict_set(m, tt, ct,
 135                             AV_DICT_DONT_STRDUP_KEY |
 136                             AV_DICT_DONT_STRDUP_VAL);
 137         }
 138     }
 139
 140     if (p != end)
 141         av_log(as, AV_LOG_INFO,
 142                "%ti bytes of comment header remain\n", end - p);
 143     if (n > 0)
 144         av_log(as, AV_LOG_INFO,
 145                "truncated comment header, %i comments not found\n", n);
 146
 147     ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
 148
 149     return 0;
 150 }
 151
 152 /*
 153  * Parse the vorbis header
 154  *
 155  * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
 156  * [vorbis_version] = read 32 bits as unsigned integer | Not used
 157  * [audio_channels] = read 8 bit integer as unsigned | Used
 158  * [audio_sample_rate] = read 32 bits as unsigned integer | Used
 159  * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
 160  * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
 161  * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
 162  * [blocksize_0] = read 4 bits as unsigned integer | Not Used
 163  * [blocksize_1] = read 4 bits as unsigned integer | Not Used
 164  * [framing_flag] = read one bit | Not Used
 165  */
 166
 167 struct oggvorbis_private {
 168     unsigned int len[3];
 169     unsigned char *packet[3];
 170     VorbisParseContext vp;
 171     int64_t final_pts;
 172     int final_duration;
 173 };
 174
 175 static unsigned int fixup_vorbis_headers(AVFormatContext *as,
 176                                          struct oggvorbis_private *priv,
 177                                          uint8_t **buf)
 178 {
 179     int i, offset, len, err;
 180     unsigned char *ptr;
 181
 182     len = priv->len[0] + priv->len[1] + priv->len[2];
 183     ptr = *buf = av_mallocz(len + len / 255 + 64);
 184
 185     ptr[0]  = 2;
 186     offset  = 1;
 187     offset += av_xiphlacing(&ptr[offset], priv->len[0]);
 188     offset += av_xiphlacing(&ptr[offset], priv->len[1]);
 189     for (i = 0; i < 3; i++) {
 190         memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
 191         offset += priv->len[i];
 192         av_freep(&priv->packet[i]);
 193     }
 194     if ((err = av_reallocp(buf, offset + FF_INPUT_BUFFER_PADDING_SIZE)) < 0)
 195         return err;
 196     return offset;
 197 }
 198
 199 static void vorbis_cleanup(AVFormatContext *s, int idx)
 200 {
 201     struct ogg *ogg = s->priv_data;
 202     struct ogg_stream *os = ogg->streams + idx;
 203     struct oggvorbis_private *priv = os->private;
 204     int i;
 205     if (os->private)
 206         for (i = 0; i < 3; i++)
 207             av_freep(&priv->packet[i]);
 208 }
 209
 210 static int vorbis_header(AVFormatContext *s, int idx)
 211 {
 212     struct ogg *ogg = s->priv_data;
 213     AVStream *st    = s->streams[idx];
 214     struct ogg_stream *os = ogg->streams + idx;
 215     struct oggvorbis_private *priv;
 216     int pkt_type = os->buf[os->pstart];
 217
 218     if (!os->private) {
 219         os->private = av_mallocz(sizeof(struct oggvorbis_private));
 220         if (!os->private)
 221             return AVERROR(ENOMEM);
 222     }
 223
 224     if (!(pkt_type & 1))
 225         return 0;
 226
 227     if (os->psize < 1 || pkt_type > 5)
 228         return AVERROR_INVALIDDATA;
 229
 230     priv = os->private;
 231
 232     if (priv->packet[pkt_type >> 1])
 233         return AVERROR_INVALIDDATA;
 234     if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
 235         return AVERROR_INVALIDDATA;
 236
 237     priv->len[pkt_type >> 1]    = os->psize;
 238     priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
 239     memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
 240     if (os->buf[os->pstart] == 1) {
 241         const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
 242         unsigned blocksize, bs0, bs1;
 243         int srate;
 244
 245         if (os->psize != 30)
 246             return AVERROR_INVALIDDATA;
 247
 248         if (bytestream_get_le32(&p) != 0) /* vorbis_version */
 249             return AVERROR_INVALIDDATA;
 250
 251         st->codec->channels = bytestream_get_byte(&p);
 252         srate               = bytestream_get_le32(&p);
 253         p += 4; // skip maximum bitrate
 254         st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
 255         p += 4; // skip minimum bitrate
 256
 257         blocksize = bytestream_get_byte(&p);
 258         bs0       = blocksize & 15;
 259         bs1       = blocksize >> 4;
 260
 261         if (bs0 > bs1)
 262             return AVERROR_INVALIDDATA;
 263         if (bs0 < 6 || bs1 > 13)
 264             return AVERROR_INVALIDDATA;
 265
 266         if (bytestream_get_byte(&p) != 1) /* framing_flag */
 267             return AVERROR_INVALIDDATA;
 268
 269         st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
 270         st->codec->codec_id   = AV_CODEC_ID_VORBIS;
 271
 272         if (srate > 0) {
 273             st->codec->sample_rate = srate;
 274             avpriv_set_pts_info(st, 64, 1, srate);
 275         }
 276     } else if (os->buf[os->pstart] == 3) {
 277         if (os->psize > 8 &&
 278             ff_vorbis_comment(s, &st->metadata, os->buf + os->pstart + 7,
 279                               os->psize - 8) >= 0) {
 280             // drop all metadata we parsed and which is not required by libvorbis
 281             unsigned new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
 282             if (new_len >= 16 && new_len < os->psize) {
 283                 AV_WL32(priv->packet[1] + new_len - 5, 0);
 284                 priv->packet[1][new_len - 1] = 1;
 285                 priv->len[1]                 = new_len;
 286             }
 287         }
 288     } else {
 289         int ret = fixup_vorbis_headers(s, priv, &st->codec->extradata);
 290         if (ret < 0) {
 291             st->codec->extradata_size = 0;
 292             return ret;
 293         }
 294         st->codec->extradata_size = ret;
 295         if ((ret = avpriv_vorbis_parse_extradata(st->codec, &priv->vp))) {
 296             av_freep(&st->codec->extradata);
 297             st->codec->extradata_size = 0;
 298             return ret;
 299         }
 300     }
 301
 302     return 1;
 303 }
 304
 305 static int vorbis_packet(AVFormatContext *s, int idx)
 306 {
 307     struct ogg *ogg = s->priv_data;
 308     struct ogg_stream *os = ogg->streams + idx;
 309     struct oggvorbis_private *priv = os->private;
 310     int duration;
 311
 312     /* first packet handling
 313      * here we parse the duration of each packet in the first page and compare
 314      * the total duration to the page granule to find the encoder delay and
 315      * set the first timestamp */
 316     if (!os->lastpts) {
 317         int seg;
 318         uint8_t *last_pkt  = os->buf + os->pstart;
 319         uint8_t *next_pkt  = last_pkt;
 320         int first_duration = 0;
 321
 322         avpriv_vorbis_parse_reset(&priv->vp);
 323         duration = 0;
 324         for (seg = 0; seg < os->nsegs; seg++) {
 325             if (os->segments[seg] < 255) {
 326                 int d = avpriv_vorbis_parse_frame(&priv->vp, last_pkt, 1);
 327                 if (d < 0) {
 328                     duration = os->granule;
 329                     break;
 330                 }
 331                 if (!duration)
 332                     first_duration = d;
 333                 duration += d;
 334                 last_pkt  = next_pkt + os->segments[seg];
 335             }
 336             next_pkt += os->segments[seg];
 337         }
 338         os->lastpts                 =
 339         os->lastdts                 = os->granule - duration;
 340         s->streams[idx]->start_time = os->lastpts + first_duration;
 341         if (s->streams[idx]->duration)
 342             s->streams[idx]->duration -= s->streams[idx]->start_time;
 343         s->streams[idx]->cur_dts = AV_NOPTS_VALUE;
 344         priv->final_pts          = AV_NOPTS_VALUE;
 345         avpriv_vorbis_parse_reset(&priv->vp);
 346     }
 347
 348     /* parse packet duration */
 349     if (os->psize > 0) {
 350         duration = avpriv_vorbis_parse_frame(&priv->vp, os->buf + os->pstart, 1);
 351         if (duration <= 0) {
 352             os->pflags |= AV_PKT_FLAG_CORRUPT;
 353             return 0;
 354         }
 355         os->pduration = duration;
 356     }
 357
 358     /* final packet handling
 359      * here we save the pts of the first packet in the final page, sum up all
 360      * packet durations in the final page except for the last one, and compare
 361      * to the page granule to find the duration of the final packet */
 362     if (os->flags & OGG_FLAG_EOS) {
 363         if (os->lastpts != AV_NOPTS_VALUE) {
 364             priv->final_pts      = os->lastpts;
 365             priv->final_duration = 0;
 366         }
 367         if (os->segp == os->nsegs)
 368             os->pduration = os->granule - priv->final_pts - priv->final_duration;
 369         priv->final_duration += os->pduration;
 370     }
 371
 372     return 0;
 373 }
 374
 375 const struct ogg_codec ff_vorbis_codec = {
 376     .magic     = "\001vorbis",
 377     .magicsize = 7,
 378     .header    = vorbis_header,
 379     .packet    = vorbis_packet,
 380     .cleanup   = vorbis_cleanup,
 381     .nb_header = 3,
 382 };