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 Libav.
   6  *
   7  * Libav 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  * Libav 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 Libav; 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/mathematics.h"
  24 #include "libavutil/random_seed.h"
  25 #include "libavcodec/xiph.h"
  26 #include "libavcodec/bytestream.h"
  27 #include "libavcodec/flac.h"
  28 #include "avformat.h"
  29 #include "avio_internal.h"
  30 #include "internal.h"
  31 #include "vorbiscomment.h"
  32
  33 #define MAX_PAGE_SIZE 65025
  34
  35 typedef struct {
  36     int64_t granule;
  37     int stream_index;
  38     uint8_t flags;
  39     uint8_t segments_count;
  40     uint8_t segments[255];
  41     uint8_t data[MAX_PAGE_SIZE];
  42     uint16_t size;
  43 } OGGPage;
  44
  45 typedef struct {
  46     unsigned page_counter;
  47     uint8_t *header[3];
  48     int header_len[3];
  49     /** for theora granule */
  50     int kfgshift;
  51     int64_t last_kf_pts;
  52     int vrev;
  53     int eos;
  54     unsigned page_count; ///< number of page buffered
  55     OGGPage page; ///< current page
  56     unsigned serial_num; ///< serial number
  57     int64_t last_granule; ///< last packet granule
  58 } OGGStreamContext;
  59
  60 typedef struct OGGPageList {
  61     OGGPage page;
  62     struct OGGPageList *next;
  63 } OGGPageList;
  64
  65 typedef struct {
  66     OGGPageList *page_list;
  67 } OGGContext;
  68
  69 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
  70 {
  71     int64_t pos = avio_tell(pb);
  72     uint32_t checksum = ffio_get_checksum(pb);
  73     avio_seek(pb, crc_offset, SEEK_SET);
  74     avio_wb32(pb, checksum);
  75     avio_seek(pb, pos, SEEK_SET);
  76 }
  77
  78 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
  79 {
  80     OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
  81     AVIOContext *pb;
  82     int64_t crc_offset;
  83     int ret, size;
  84     uint8_t *buf;
  85
  86     ret = avio_open_dyn_buf(&pb);
  87     if (ret < 0)
  88         return ret;
  89     ffio_init_checksum(pb, ff_crc04C11DB7_update, 0);
  90     ffio_wfourcc(pb, "OggS");
  91     avio_w8(pb, 0);
  92     avio_w8(pb, page->flags | extra_flags);
  93     avio_wl64(pb, page->granule);
  94     avio_wl32(pb, oggstream->serial_num);
  95     avio_wl32(pb, oggstream->page_counter++);
  96     crc_offset = avio_tell(pb);
  97     avio_wl32(pb, 0); // crc
  98     avio_w8(pb, page->segments_count);
  99     avio_write(pb, page->segments, page->segments_count);
 100     avio_write(pb, page->data, page->size);
 101
 102     ogg_update_checksum(s, pb, crc_offset);
 103     avio_flush(pb);
 104
 105     size = avio_close_dyn_buf(pb, &buf);
 106     if (size < 0)
 107         return size;
 108
 109     avio_write(s->pb, buf, size);
 110     avio_flush(s->pb);
 111     av_free(buf);
 112     oggstream->page_count--;
 113     return 0;
 114 }
 115
 116 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
 117 {
 118     if (oggstream->kfgshift)
 119         return (granule>>oggstream->kfgshift) +
 120             (granule & ((1<<oggstream->kfgshift)-1));
 121     else
 122         return granule;
 123 }
 124
 125 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
 126 {
 127     AVStream *st2 = s->streams[next->stream_index];
 128     AVStream *st  = s->streams[page->stream_index];
 129     int64_t next_granule, cur_granule;
 130
 131     if (next->granule == -1 || page->granule == -1)
 132         return 0;
 133
 134     next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
 135                                 st2->time_base, AV_TIME_BASE_Q);
 136     cur_granule  = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
 137                                 st ->time_base, AV_TIME_BASE_Q);
 138     return next_granule > cur_granule;
 139 }
 140
 141 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
 142 {
 143     oggstream->page.granule = -1;
 144     oggstream->page.flags = 0;
 145     oggstream->page.segments_count = 0;
 146     oggstream->page.size = 0;
 147     return 0;
 148 }
 149
 150 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
 151 {
 152     OGGContext *ogg = s->priv_data;
 153     OGGPageList **p = &ogg->page_list;
 154     OGGPageList *l = av_mallocz(sizeof(*l));
 155
 156     if (!l)
 157         return AVERROR(ENOMEM);
 158     l->page = oggstream->page;
 159
 160     oggstream->page_count++;
 161     ogg_reset_cur_page(oggstream);
 162
 163     while (*p) {
 164         if (ogg_compare_granule(s, &(*p)->page, &l->page))
 165             break;
 166         p = &(*p)->next;
 167     }
 168     l->next = *p;
 169     *p = l;
 170
 171     return 0;
 172 }
 173
 174 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
 175                            uint8_t *data, unsigned size, int64_t granule)
 176 {
 177     OGGStreamContext *oggstream = st->priv_data;
 178     int total_segments = size / 255 + 1;
 179     uint8_t *p = data;
 180     int i, segments, len, flush = 0;
 181
 182     // Handles VFR by flushing page because this frame needs to have a timestamp
 183     if (st->codec->codec_id == CODEC_ID_THEORA &&
 184         ogg_granule_to_timestamp(oggstream, granule) >
 185         ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1) {
 186         if (oggstream->page.granule != -1)
 187             ogg_buffer_page(s, oggstream);
 188         flush = 1;
 189     }
 190
 191     for (i = 0; i < total_segments; ) {
 192         OGGPage *page = &oggstream->page;
 193
 194         segments = FFMIN(total_segments - i, 255 - page->segments_count);
 195
 196         if (i && !page->segments_count)
 197             page->flags |= 1; // continued packet
 198
 199         memset(page->segments+page->segments_count, 255, segments - 1);
 200         page->segments_count += segments - 1;
 201
 202         len = FFMIN(size, segments*255);
 203         page->segments[page->segments_count++] = len - (segments-1)*255;
 204         memcpy(page->data+page->size, p, len);
 205         p += len;
 206         size -= len;
 207         i += segments;
 208         page->size += len;
 209
 210         if (i == total_segments)
 211             page->granule = granule;
 212
 213         if (page->segments_count == 255) {
 214             ogg_buffer_page(s, oggstream);
 215         }
 216     }
 217
 218     if (flush && oggstream->page.granule != -1)
 219         ogg_buffer_page(s, oggstream);
 220
 221     return 0;
 222 }
 223
 224 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
 225                                         int *header_len, AVDictionary **m, int framing_bit)
 226 {
 227     const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
 228     int size;
 229     uint8_t *p, *p0;
 230     unsigned int count;
 231
 232     ff_metadata_conv(m, ff_vorbiscomment_metadata_conv, NULL);
 233
 234     size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
 235     p = av_mallocz(size);
 236     if (!p)
 237         return NULL;
 238     p0 = p;
 239
 240     p += offset;
 241     ff_vorbiscomment_write(&p, m, vendor, count);
 242     if (framing_bit)
 243         bytestream_put_byte(&p, 1);
 244
 245     *header_len = size;
 246     return p0;
 247 }
 248
 249 static int ogg_build_flac_headers(AVCodecContext *avctx,
 250                                   OGGStreamContext *oggstream, int bitexact,
 251                                   AVDictionary **m)
 252 {
 253     enum FLACExtradataFormat format;
 254     uint8_t *streaminfo;
 255     uint8_t *p;
 256
 257     if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo))
 258         return -1;
 259
 260     // first packet: STREAMINFO
 261     oggstream->header_len[0] = 51;
 262     oggstream->header[0] = av_mallocz(51); // per ogg flac specs
 263     p = oggstream->header[0];
 264     if (!p)
 265         return AVERROR(ENOMEM);
 266     bytestream_put_byte(&p, 0x7F);
 267     bytestream_put_buffer(&p, "FLAC", 4);
 268     bytestream_put_byte(&p, 1); // major version
 269     bytestream_put_byte(&p, 0); // minor version
 270     bytestream_put_be16(&p, 1); // headers packets without this one
 271     bytestream_put_buffer(&p, "fLaC", 4);
 272     bytestream_put_byte(&p, 0x00); // streaminfo
 273     bytestream_put_be24(&p, 34);
 274     bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
 275
 276     // second packet: VorbisComment
 277     p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
 278     if (!p)
 279         return AVERROR(ENOMEM);
 280     oggstream->header[1] = p;
 281     bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
 282     bytestream_put_be24(&p, oggstream->header_len[1] - 4);
 283
 284     return 0;
 285 }
 286
 287 #define SPEEX_HEADER_SIZE 80
 288
 289 static int ogg_build_speex_headers(AVCodecContext *avctx,
 290                                    OGGStreamContext *oggstream, int bitexact,
 291                                    AVDictionary **m)
 292 {
 293     uint8_t *p;
 294
 295     if (avctx->extradata_size < SPEEX_HEADER_SIZE)
 296         return -1;
 297
 298     // first packet: Speex header
 299     p = av_mallocz(SPEEX_HEADER_SIZE);
 300     if (!p)
 301         return AVERROR(ENOMEM);
 302     oggstream->header[0] = p;
 303     oggstream->header_len[0] = SPEEX_HEADER_SIZE;
 304     bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
 305     AV_WL32(&oggstream->header[0][68], 0);  // set extra_headers to 0
 306
 307     // second packet: VorbisComment
 308     p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
 309     if (!p)
 310         return AVERROR(ENOMEM);
 311     oggstream->header[1] = p;
 312
 313     return 0;
 314 }
 315
 316 static int ogg_write_header(AVFormatContext *s)
 317 {
 318     OGGStreamContext *oggstream;
 319     int i, j;
 320
 321     for (i = 0; i < s->nb_streams; i++) {
 322         AVStream *st = s->streams[i];
 323         unsigned serial_num = i;
 324
 325         if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
 326             av_set_pts_info(st, 64, 1, st->codec->sample_rate);
 327         else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
 328             av_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
 329         if (st->codec->codec_id != CODEC_ID_VORBIS &&
 330             st->codec->codec_id != CODEC_ID_THEORA &&
 331             st->codec->codec_id != CODEC_ID_SPEEX  &&
 332             st->codec->codec_id != CODEC_ID_FLAC) {
 333             av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
 334             return -1;
 335         }
 336
 337         if (!st->codec->extradata || !st->codec->extradata_size) {
 338             av_log(s, AV_LOG_ERROR, "No extradata present\n");
 339             return -1;
 340         }
 341         oggstream = av_mallocz(sizeof(*oggstream));
 342         oggstream->page.stream_index = i;
 343
 344         if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
 345             do {
 346                 serial_num = av_get_random_seed();
 347                 for (j = 0; j < i; j++) {
 348                     OGGStreamContext *sc = s->streams[j]->priv_data;
 349                     if (serial_num == sc->serial_num)
 350                         break;
 351                 }
 352             } while (j < i);
 353         oggstream->serial_num = serial_num;
 354
 355         st->priv_data = oggstream;
 356         if (st->codec->codec_id == CODEC_ID_FLAC) {
 357             int err = ogg_build_flac_headers(st->codec, oggstream,
 358                                              st->codec->flags & CODEC_FLAG_BITEXACT,
 359                                              &s->metadata);
 360             if (err) {
 361                 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
 362                 av_freep(&st->priv_data);
 363                 return err;
 364             }
 365         } else if (st->codec->codec_id == CODEC_ID_SPEEX) {
 366             int err = ogg_build_speex_headers(st->codec, oggstream,
 367                                               st->codec->flags & CODEC_FLAG_BITEXACT,
 368                                               &s->metadata);
 369             if (err) {
 370                 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
 371                 av_freep(&st->priv_data);
 372                 return err;
 373             }
 374         } else {
 375             uint8_t *p;
 376             const char *cstr = st->codec->codec_id == CODEC_ID_VORBIS ? "vorbis" : "theora";
 377             int header_type = st->codec->codec_id == CODEC_ID_VORBIS ? 3 : 0x81;
 378             int framing_bit = st->codec->codec_id == CODEC_ID_VORBIS ? 1 : 0;
 379
 380             if (ff_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
 381                                       st->codec->codec_id == CODEC_ID_VORBIS ? 30 : 42,
 382                                       oggstream->header, oggstream->header_len) < 0) {
 383                 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
 384                 av_freep(&st->priv_data);
 385                 return -1;
 386             }
 387
 388             p = ogg_write_vorbiscomment(7, st->codec->flags & CODEC_FLAG_BITEXACT,
 389                                         &oggstream->header_len[1], &s->metadata,
 390                                         framing_bit);
 391             if (!p)
 392                 return AVERROR(ENOMEM);
 393
 394             oggstream->header[1] = p;
 395             bytestream_put_byte(&p, header_type);
 396             bytestream_put_buffer(&p, cstr, 6);
 397
 398             if (st->codec->codec_id == CODEC_ID_THEORA) {
 399                 /** KFGSHIFT is the width of the less significant section of the granule position
 400                     The less significant section is the frame count since the last keyframe */
 401                 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
 402                 oggstream->vrev = oggstream->header[0][9];
 403                 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
 404                        oggstream->kfgshift, oggstream->vrev);
 405             }
 406         }
 407     }
 408
 409     for (j = 0; j < s->nb_streams; j++) {
 410         OGGStreamContext *oggstream = s->streams[j]->priv_data;
 411         ogg_buffer_data(s, s->streams[j], oggstream->header[0],
 412                         oggstream->header_len[0], 0);
 413         oggstream->page.flags |= 2; // bos
 414         ogg_buffer_page(s, oggstream);
 415     }
 416     for (j = 0; j < s->nb_streams; j++) {
 417         AVStream *st = s->streams[j];
 418         OGGStreamContext *oggstream = st->priv_data;
 419         for (i = 1; i < 3; i++) {
 420             if (oggstream && oggstream->header_len[i])
 421                 ogg_buffer_data(s, st, oggstream->header[i],
 422                                 oggstream->header_len[i], 0);
 423         }
 424         ogg_buffer_page(s, oggstream);
 425     }
 426     return 0;
 427 }
 428
 429 static void ogg_write_pages(AVFormatContext *s, int flush)
 430 {
 431     OGGContext *ogg = s->priv_data;
 432     OGGPageList *next, *p;
 433
 434     if (!ogg->page_list)
 435         return;
 436
 437     for (p = ogg->page_list; p; ) {
 438         OGGStreamContext *oggstream =
 439             s->streams[p->page.stream_index]->priv_data;
 440         if (oggstream->page_count < 2 && !flush)
 441             break;
 442         ogg_write_page(s, &p->page,
 443                        flush && oggstream->page_count == 1 ? 4 : 0); // eos
 444         next = p->next;
 445         av_freep(&p);
 446         p = next;
 447     }
 448     ogg->page_list = p;
 449 }
 450
 451 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
 452 {
 453     AVStream *st = s->streams[pkt->stream_index];
 454     OGGStreamContext *oggstream = st->priv_data;
 455     int ret;
 456     int64_t granule;
 457
 458     if (st->codec->codec_id == CODEC_ID_THEORA) {
 459         int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
 460         int pframe_count;
 461         if (pkt->flags & AV_PKT_FLAG_KEY)
 462             oggstream->last_kf_pts = pts;
 463         pframe_count = pts - oggstream->last_kf_pts;
 464         // prevent frame count from overflow if key frame flag is not set
 465         if (pframe_count >= (1<<oggstream->kfgshift)) {
 466             oggstream->last_kf_pts += pframe_count;
 467             pframe_count = 0;
 468         }
 469         granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
 470     } else
 471         granule = pkt->pts + pkt->duration;
 472
 473     ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule);
 474     if (ret < 0)
 475         return ret;
 476
 477     ogg_write_pages(s, 0);
 478
 479     oggstream->last_granule = granule;
 480
 481     return 0;
 482 }
 483
 484 static int ogg_write_trailer(AVFormatContext *s)
 485 {
 486     int i;
 487
 488     /* flush current page */
 489     for (i = 0; i < s->nb_streams; i++)
 490         ogg_buffer_page(s, s->streams[i]->priv_data);
 491
 492     ogg_write_pages(s, 1);
 493
 494     for (i = 0; i < s->nb_streams; i++) {
 495         AVStream *st = s->streams[i];
 496         OGGStreamContext *oggstream = st->priv_data;
 497         if (st->codec->codec_id == CODEC_ID_FLAC ||
 498             st->codec->codec_id == CODEC_ID_SPEEX) {
 499             av_free(oggstream->header[0]);
 500             av_free(oggstream->header[1]);
 501         }
 502         av_freep(&st->priv_data);
 503     }
 504     return 0;
 505 }
 506
 507 AVOutputFormat ff_ogg_muxer = {
 508     .name              = "ogg",
 509     .long_name         = NULL_IF_CONFIG_SMALL("Ogg"),
 510     .mime_type         = "application/ogg",
 511     .extensions        = "ogg,ogv,spx",
 512     .priv_data_size    = sizeof(OGGContext),
 513     .audio_codec       = CODEC_ID_FLAC,
 514     .video_codec       = CODEC_ID_THEORA,
 515     .write_header      = ogg_write_header,
 516     .write_packet      = ogg_write_packet,
 517     .write_trailer     = ogg_write_trailer,
 518 };