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