git.sesse.net Git - ffmpeg/blob - libavformat/oggenc.c

   1 /*
   2  * Ogg muxer
   3  * Copyright (c) 2007 Baptiste Coudurier <baptiste dot coudurier at free dot fr>
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 #include "libavutil/crc.h"
  23 #include "libavutil/random_seed.h"
  24 #include "libavcodec/xiph.h"
  25 #include "libavcodec/bytestream.h"
  26 #include "libavcodec/flac.h"
  27 #include "avformat.h"
  28 #include "avio_internal.h"
  29 #include "internal.h"
  30 #include "vorbiscomment.h"
  31
  32 #define MAX_PAGE_SIZE 65025
  33
  34 typedef struct {
  35     int64_t granule;
  36     int stream_index;
  37     uint8_t flags;
  38     uint8_t segments_count;
  39     uint8_t segments[255];
  40     uint8_t data[MAX_PAGE_SIZE];
  41     uint16_t size;
  42 } OGGPage;
  43
  44 typedef struct {
  45     unsigned page_counter;
  46     uint8_t *header[3];
  47     int header_len[3];
  48     /** for theora granule */
  49     int kfgshift;
  50     int64_t last_kf_pts;
  51     int vrev;
  52     int eos;
  53     unsigned page_count; ///< number of page buffered
  54     OGGPage page; ///< current page
  55     unsigned serial_num; ///< serial number
  56     int64_t last_granule; ///< last packet granule
  57 } OGGStreamContext;
  58
  59 typedef struct OGGPageList {
  60     OGGPage page;
  61     struct OGGPageList *next;
  62 } OGGPageList;
  63
  64 typedef struct {
  65     OGGPageList *page_list;
  66 } OGGContext;
  67
  68 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
  69 {
  70     int64_t pos = avio_tell(pb);
  71     uint32_t checksum = get_checksum(pb);
  72     avio_seek(pb, crc_offset, SEEK_SET);
  73     avio_wb32(pb, checksum);
  74     avio_seek(pb, pos, SEEK_SET);
  75 }
  76
  77 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
  78 {
  79     OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
  80     AVIOContext *pb;
  81     int64_t crc_offset;
  82     int ret, size;
  83     uint8_t *buf;
  84
  85     ret = url_open_dyn_buf(&pb);
  86     if (ret < 0)
  87         return ret;
  88     init_checksum(pb, ff_crc04C11DB7_update, 0);
  89     ffio_wfourcc(pb, "OggS");
  90     avio_w8(pb, 0);
  91     avio_w8(pb, page->flags | extra_flags);
  92     avio_wl64(pb, page->granule);
  93     avio_wl32(pb, oggstream->serial_num);
  94     avio_wl32(pb, oggstream->page_counter++);
  95     crc_offset = avio_tell(pb);
  96     avio_wl32(pb, 0); // crc
  97     avio_w8(pb, page->segments_count);
  98     avio_write(pb, page->segments, page->segments_count);
  99     avio_write(pb, page->data, page->size);
 100
 101     ogg_update_checksum(s, pb, crc_offset);
 102     put_flush_packet(pb);
 103
 104     size = url_close_dyn_buf(pb, &buf);
 105     if (size < 0)
 106         return size;
 107
 108     avio_write(s->pb, buf, size);
 109     put_flush_packet(s->pb);
 110     av_free(buf);
 111     oggstream->page_count--;
 112     return 0;
 113 }
 114
 115 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
 116 {
 117     if (oggstream->kfgshift)
 118         return (granule>>oggstream->kfgshift) +
 119             (granule & ((1<<oggstream->kfgshift)-1));
 120     else
 121         return granule;
 122 }
 123
 124 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
 125 {
 126     AVStream *st2 = s->streams[next->stream_index];
 127     AVStream *st  = s->streams[page->stream_index];
 128     int64_t next_granule, cur_granule;
 129
 130     if (next->granule == -1 || page->granule == -1)
 131         return 0;
 132
 133     next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
 134                                 st2->time_base, AV_TIME_BASE_Q);
 135     cur_granule  = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
 136                                 st ->time_base, AV_TIME_BASE_Q);
 137     return next_granule > cur_granule;
 138 }
 139
 140 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
 141 {
 142     oggstream->page.granule = -1;
 143     oggstream->page.flags = 0;
 144     oggstream->page.segments_count = 0;
 145     oggstream->page.size = 0;
 146     return 0;
 147 }
 148
 149 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
 150 {
 151     OGGContext *ogg = s->priv_data;
 152     OGGPageList **p = &ogg->page_list;
 153     OGGPageList *l = av_mallocz(sizeof(*l));
 154
 155     if (!l)
 156         return AVERROR(ENOMEM);
 157     l->page = oggstream->page;
 158
 159     oggstream->page_count++;
 160     ogg_reset_cur_page(oggstream);
 161
 162     while (*p) {
 163         if (ogg_compare_granule(s, &(*p)->page, &l->page))
 164             break;
 165         p = &(*p)->next;
 166     }
 167     l->next = *p;
 168     *p = l;
 169
 170     return 0;
 171 }
 172
 173 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
 174                            uint8_t *data, unsigned size, int64_t granule)
 175 {
 176     OGGStreamContext *oggstream = st->priv_data;
 177     int total_segments = size / 255 + 1;
 178     uint8_t *p = data;
 179     int i, segments, len, flush = 0;
 180
 181     // Handles VFR by flushing page because this frame needs to have a timestamp
 182     if (st->codec->codec_id == CODEC_ID_THEORA &&
 183         ogg_granule_to_timestamp(oggstream, granule) >
 184         ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1) {
 185         if (oggstream->page.granule != -1)
 186             ogg_buffer_page(s, oggstream);
 187         flush = 1;
 188     }
 189
 190     for (i = 0; i < total_segments; ) {
 191         OGGPage *page = &oggstream->page;
 192
 193         segments = FFMIN(total_segments - i, 255 - page->segments_count);
 194
 195         if (i && !page->segments_count)
 196             page->flags |= 1; // continued packet
 197
 198         memset(page->segments+page->segments_count, 255, segments - 1);
 199         page->segments_count += segments - 1;
 200
 201         len = FFMIN(size, segments*255);
 202         page->segments[page->segments_count++] = len - (segments-1)*255;
 203         memcpy(page->data+page->size, p, len);
 204         p += len;
 205         size -= len;
 206         i += segments;
 207         page->size += len;
 208
 209         if (i == total_segments)
 210             page->granule = granule;
 211
 212         if (page->segments_count == 255) {
 213             ogg_buffer_page(s, oggstream);
 214         }
 215     }
 216
 217     if (flush && oggstream->page.granule != -1)
 218         ogg_buffer_page(s, oggstream);
 219
 220     return 0;
 221 }
 222
 223 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
 224                                         int *header_len, AVMetadata **m, int framing_bit)
 225 {
 226     const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
 227     int size;
 228     uint8_t *p, *p0;
 229     unsigned int count;
 230
 231     ff_metadata_conv(m, ff_vorbiscomment_metadata_conv, NULL);
 232
 233     size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
 234     p = av_mallocz(size);
 235     if (!p)
 236         return NULL;
 237     p0 = p;
 238
 239     p += offset;
 240     ff_vorbiscomment_write(&p, m, vendor, count);
 241     if (framing_bit)
 242         bytestream_put_byte(&p, 1);
 243
 244     *header_len = size;
 245     return p0;
 246 }
 247
 248 static int ogg_build_flac_headers(AVCodecContext *avctx,
 249                                   OGGStreamContext *oggstream, int bitexact,
 250                                   AVMetadata **m)
 251 {
 252     enum FLACExtradataFormat format;
 253     uint8_t *streaminfo;
 254     uint8_t *p;
 255
 256     if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo))
 257         return -1;
 258
 259     // first packet: STREAMINFO
 260     oggstream->header_len[0] = 51;
 261     oggstream->header[0] = av_mallocz(51); // per ogg flac specs
 262     p = oggstream->header[0];
 263     if (!p)
 264         return AVERROR(ENOMEM);
 265     bytestream_put_byte(&p, 0x7F);
 266     bytestream_put_buffer(&p, "FLAC", 4);
 267     bytestream_put_byte(&p, 1); // major version
 268     bytestream_put_byte(&p, 0); // minor version
 269     bytestream_put_be16(&p, 1); // headers packets without this one
 270     bytestream_put_buffer(&p, "fLaC", 4);
 271     bytestream_put_byte(&p, 0x00); // streaminfo
 272     bytestream_put_be24(&p, 34);
 273     bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
 274
 275     // second packet: VorbisComment
 276     p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
 277     if (!p)
 278         return AVERROR(ENOMEM);
 279     oggstream->header[1] = p;
 280     bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
 281     bytestream_put_be24(&p, oggstream->header_len[1] - 4);
 282
 283     return 0;
 284 }
 285
 286 #define SPEEX_HEADER_SIZE 80
 287
 288 static int ogg_build_speex_headers(AVCodecContext *avctx,
 289                                    OGGStreamContext *oggstream, int bitexact,
 290                                    AVMetadata **m)
 291 {
 292     uint8_t *p;
 293
 294     if (avctx->extradata_size < SPEEX_HEADER_SIZE)
 295         return -1;
 296
 297     // first packet: Speex header
 298     p = av_mallocz(SPEEX_HEADER_SIZE);
 299     if (!p)
 300         return AVERROR(ENOMEM);
 301     oggstream->header[0] = p;
 302     oggstream->header_len[0] = SPEEX_HEADER_SIZE;
 303     bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
 304     AV_WL32(&oggstream->header[0][68], 0);  // set extra_headers to 0
 305
 306     // second packet: VorbisComment
 307     p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
 308     if (!p)
 309         return AVERROR(ENOMEM);
 310     oggstream->header[1] = p;
 311
 312     return 0;
 313 }
 314
 315 static int ogg_write_header(AVFormatContext *s)
 316 {
 317     OGGStreamContext *oggstream;
 318     int i, j;
 319
 320     for (i = 0; i < s->nb_streams; i++) {
 321         AVStream *st = s->streams[i];
 322         unsigned serial_num = i;
 323
 324         if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
 325             av_set_pts_info(st, 64, 1, st->codec->sample_rate);
 326         else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
 327             av_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
 328         if (st->codec->codec_id != CODEC_ID_VORBIS &&
 329             st->codec->codec_id != CODEC_ID_THEORA &&
 330             st->codec->codec_id != CODEC_ID_SPEEX  &&
 331             st->codec->codec_id != CODEC_ID_FLAC) {
 332             av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
 333             return -1;
 334         }
 335
 336         if (!st->codec->extradata || !st->codec->extradata_size) {
 337             av_log(s, AV_LOG_ERROR, "No extradata present\n");
 338             return -1;
 339         }
 340         oggstream = av_mallocz(sizeof(*oggstream));
 341         oggstream->page.stream_index = i;
 342
 343         if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
 344             do {
 345                 serial_num = av_get_random_seed();
 346                 for (j = 0; j < i; j++) {
 347                     OGGStreamContext *sc = s->streams[j]->priv_data;
 348                     if (serial_num == sc->serial_num)
 349                         break;
 350                 }
 351             } while (j < i);
 352         oggstream->serial_num = serial_num;
 353
 354         st->priv_data = oggstream;
 355         if (st->codec->codec_id == CODEC_ID_FLAC) {
 356             int err = ogg_build_flac_headers(st->codec, oggstream,
 357                                              st->codec->flags & CODEC_FLAG_BITEXACT,
 358                                              &s->metadata);
 359             if (err) {
 360                 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
 361                 av_freep(&st->priv_data);
 362                 return err;
 363             }
 364         } else if (st->codec->codec_id == CODEC_ID_SPEEX) {
 365             int err = ogg_build_speex_headers(st->codec, oggstream,
 366                                               st->codec->flags & CODEC_FLAG_BITEXACT,
 367                                               &s->metadata);
 368             if (err) {
 369                 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
 370                 av_freep(&st->priv_data);
 371                 return err;
 372             }
 373         } else {
 374             uint8_t *p;
 375             const char *cstr = st->codec->codec_id == CODEC_ID_VORBIS ? "vorbis" : "theora";
 376             int header_type = st->codec->codec_id == CODEC_ID_VORBIS ? 3 : 0x81;
 377             int framing_bit = st->codec->codec_id == CODEC_ID_VORBIS ? 1 : 0;
 378
 379             if (ff_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
 380                                       st->codec->codec_id == CODEC_ID_VORBIS ? 30 : 42,
 381                                       oggstream->header, oggstream->header_len) < 0) {
 382                 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
 383                 av_freep(&st->priv_data);
 384                 return -1;
 385             }
 386
 387             p = ogg_write_vorbiscomment(7, st->codec->flags & CODEC_FLAG_BITEXACT,
 388                                         &oggstream->header_len[1], &s->metadata,
 389                                         framing_bit);
 390             if (!p)
 391                 return AVERROR(ENOMEM);
 392
 393             oggstream->header[1] = p;
 394             bytestream_put_byte(&p, header_type);
 395             bytestream_put_buffer(&p, cstr, 6);
 396
 397             if (st->codec->codec_id == CODEC_ID_THEORA) {
 398                 /** KFGSHIFT is the width of the less significant section of the granule position
 399                     The less significant section is the frame count since the last keyframe */
 400                 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
 401                 oggstream->vrev = oggstream->header[0][9];
 402                 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
 403                        oggstream->kfgshift, oggstream->vrev);
 404             }
 405         }
 406     }
 407
 408     for (j = 0; j < s->nb_streams; j++) {
 409         OGGStreamContext *oggstream = s->streams[j]->priv_data;
 410         ogg_buffer_data(s, s->streams[j], oggstream->header[0],
 411                         oggstream->header_len[0], 0);
 412         oggstream->page.flags |= 2; // bos
 413         ogg_buffer_page(s, oggstream);
 414     }
 415     for (j = 0; j < s->nb_streams; j++) {
 416         AVStream *st = s->streams[j];
 417         OGGStreamContext *oggstream = st->priv_data;
 418         for (i = 1; i < 3; i++) {
 419             if (oggstream && oggstream->header_len[i])
 420                 ogg_buffer_data(s, st, oggstream->header[i],
 421                                 oggstream->header_len[i], 0);
 422         }
 423         ogg_buffer_page(s, oggstream);
 424     }
 425     return 0;
 426 }
 427
 428 static void ogg_write_pages(AVFormatContext *s, int flush)
 429 {
 430     OGGContext *ogg = s->priv_data;
 431     OGGPageList *next, *p;
 432
 433     if (!ogg->page_list)
 434         return;
 435
 436     for (p = ogg->page_list; p; ) {
 437         OGGStreamContext *oggstream =
 438             s->streams[p->page.stream_index]->priv_data;
 439         if (oggstream->page_count < 2 && !flush)
 440             break;
 441         ogg_write_page(s, &p->page,
 442                        flush && oggstream->page_count == 1 ? 4 : 0); // eos
 443         next = p->next;
 444         av_freep(&p);
 445         p = next;
 446     }
 447     ogg->page_list = p;
 448 }
 449
 450 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
 451 {
 452     AVStream *st = s->streams[pkt->stream_index];
 453     OGGStreamContext *oggstream = st->priv_data;
 454     int ret;
 455     int64_t granule;
 456
 457     if (st->codec->codec_id == CODEC_ID_THEORA) {
 458         int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
 459         int pframe_count;
 460         if (pkt->flags & AV_PKT_FLAG_KEY)
 461             oggstream->last_kf_pts = pts;
 462         pframe_count = pts - oggstream->last_kf_pts;
 463         // prevent frame count from overflow if key frame flag is not set
 464         if (pframe_count >= (1<<oggstream->kfgshift)) {
 465             oggstream->last_kf_pts += pframe_count;
 466             pframe_count = 0;
 467         }
 468         granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
 469     } else
 470         granule = pkt->pts + pkt->duration;
 471
 472     ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule);
 473     if (ret < 0)
 474         return ret;
 475
 476     ogg_write_pages(s, 0);
 477
 478     oggstream->last_granule = granule;
 479
 480     return 0;
 481 }
 482
 483 static int ogg_write_trailer(AVFormatContext *s)
 484 {
 485     int i;
 486
 487     /* flush current page */
 488     for (i = 0; i < s->nb_streams; i++)
 489         ogg_buffer_page(s, s->streams[i]->priv_data);
 490
 491     ogg_write_pages(s, 1);
 492
 493     for (i = 0; i < s->nb_streams; i++) {
 494         AVStream *st = s->streams[i];
 495         OGGStreamContext *oggstream = st->priv_data;
 496         if (st->codec->codec_id == CODEC_ID_FLAC ||
 497             st->codec->codec_id == CODEC_ID_SPEEX) {
 498             av_free(oggstream->header[0]);
 499             av_free(oggstream->header[1]);
 500         }
 501         av_freep(&st->priv_data);
 502     }
 503     return 0;
 504 }
 505
 506 AVOutputFormat ff_ogg_muxer = {
 507     "ogg",
 508     NULL_IF_CONFIG_SMALL("Ogg"),
 509     "application/ogg",
 510     "ogg,ogv,spx",
 511     sizeof(OGGContext),
 512     CODEC_ID_FLAC,
 513     CODEC_ID_THEORA,
 514     ogg_write_header,
 515     ogg_write_packet,
 516     ogg_write_trailer,
 517 };