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 "libavcodec/xiph.h"
  24 #include "libavcodec/bytestream.h"
  25 #include "libavcodec/flac.h"
  26 #include "avformat.h"
  27 #include "internal.h"
  28
  29 typedef struct {
  30     int64_t duration;
  31     unsigned page_counter;
  32     uint8_t *header[3];
  33     int header_len[3];
  34     /** for theora granule */
  35     int kfgshift;
  36     int64_t last_kf_pts;
  37     int vrev;
  38     int eos;
  39     unsigned packet_count; ///< number of packet buffered
  40 } OGGStreamContext;
  41
  42 static void ogg_update_checksum(AVFormatContext *s, int64_t crc_offset)
  43 {
  44     int64_t pos = url_ftell(s->pb);
  45     uint32_t checksum = get_checksum(s->pb);
  46     url_fseek(s->pb, crc_offset, SEEK_SET);
  47     put_be32(s->pb, checksum);
  48     url_fseek(s->pb, pos, SEEK_SET);
  49 }
  50
  51 static int ogg_write_page(AVFormatContext *s, const uint8_t *data, int size,
  52                           int64_t granule, int stream_index, int flags)
  53 {
  54     OGGStreamContext *oggstream = s->streams[stream_index]->priv_data;
  55     int64_t crc_offset;
  56     int page_segments, i;
  57
  58     if (size >= 255*255) {
  59         granule = -1;
  60         size = 255*255;
  61     } else if (oggstream->eos)
  62         flags |= 4;
  63
  64     page_segments = FFMIN((size/255)+!!size, 255);
  65
  66     init_checksum(s->pb, ff_crc04C11DB7_update, 0);
  67     put_tag(s->pb, "OggS");
  68     put_byte(s->pb, 0);
  69     put_byte(s->pb, flags);
  70     put_le64(s->pb, granule);
  71     put_le32(s->pb, stream_index);
  72     put_le32(s->pb, oggstream->page_counter++);
  73     crc_offset = url_ftell(s->pb);
  74     put_le32(s->pb, 0); // crc
  75     put_byte(s->pb, page_segments);
  76     for (i = 0; i < page_segments-1; i++)
  77         put_byte(s->pb, 255);
  78     if (size) {
  79         put_byte(s->pb, size - (page_segments-1)*255);
  80         put_buffer(s->pb, data, size);
  81     }
  82     ogg_update_checksum(s, crc_offset);
  83     put_flush_packet(s->pb);
  84     return size;
  85 }
  86
  87 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
  88                                         int *header_len)
  89 {
  90     const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
  91     int size;
  92     uint8_t *p, *p0;
  93
  94     size = offset + 4 + strlen(vendor) + 4;
  95     p = av_mallocz(size);
  96     if (!p)
  97         return NULL;
  98     p0 = p;
  99
 100     p += offset;
 101     bytestream_put_le32(&p, strlen(vendor));
 102     bytestream_put_buffer(&p, vendor, strlen(vendor));
 103     bytestream_put_le32(&p, 0); // user comment list length
 104
 105     *header_len = size;
 106     return p0;
 107 }
 108
 109 static int ogg_build_flac_headers(AVCodecContext *avctx,
 110                                   OGGStreamContext *oggstream, int bitexact)
 111 {
 112     enum FLACExtradataFormat format;
 113     uint8_t *streaminfo;
 114     uint8_t *p;
 115
 116     if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo))
 117         return -1;
 118
 119     // first packet: STREAMINFO
 120     oggstream->header_len[0] = 51;
 121     oggstream->header[0] = av_mallocz(51); // per ogg flac specs
 122     p = oggstream->header[0];
 123     if (!p)
 124         return AVERROR_NOMEM;
 125     bytestream_put_byte(&p, 0x7F);
 126     bytestream_put_buffer(&p, "FLAC", 4);
 127     bytestream_put_byte(&p, 1); // major version
 128     bytestream_put_byte(&p, 0); // minor version
 129     bytestream_put_be16(&p, 1); // headers packets without this one
 130     bytestream_put_buffer(&p, "fLaC", 4);
 131     bytestream_put_byte(&p, 0x00); // streaminfo
 132     bytestream_put_be24(&p, 34);
 133     bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
 134
 135     // second packet: VorbisComment
 136     p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1]);
 137     if (!p)
 138         return AVERROR_NOMEM;
 139     oggstream->header[1] = p;
 140     bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
 141     bytestream_put_be24(&p, oggstream->header_len[1] - 4);
 142
 143     return 0;
 144 }
 145
 146 #define SPEEX_HEADER_SIZE 80
 147
 148 static int ogg_build_speex_headers(AVCodecContext *avctx,
 149                                    OGGStreamContext *oggstream, int bitexact)
 150 {
 151     uint8_t *p;
 152
 153     if (avctx->extradata_size < SPEEX_HEADER_SIZE)
 154         return -1;
 155
 156     // first packet: Speex header
 157     p = av_mallocz(SPEEX_HEADER_SIZE);
 158     if (!p)
 159         return AVERROR_NOMEM;
 160     oggstream->header[0] = p;
 161     oggstream->header_len[0] = SPEEX_HEADER_SIZE;
 162     bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
 163     AV_WL32(&oggstream->header[0][68], 0);  // set extra_headers to 0
 164
 165     // second packet: VorbisComment
 166     p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1]);
 167     if (!p)
 168         return AVERROR_NOMEM;
 169     oggstream->header[1] = p;
 170
 171     return 0;
 172 }
 173
 174 static int ogg_write_header(AVFormatContext *s)
 175 {
 176     OGGStreamContext *oggstream;
 177     int i, j;
 178     for (i = 0; i < s->nb_streams; i++) {
 179         AVStream *st = s->streams[i];
 180         if (st->codec->codec_type == CODEC_TYPE_AUDIO)
 181             av_set_pts_info(st, 64, 1, st->codec->sample_rate);
 182         else if (st->codec->codec_type == CODEC_TYPE_VIDEO)
 183             av_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
 184         if (st->codec->codec_id != CODEC_ID_VORBIS &&
 185             st->codec->codec_id != CODEC_ID_THEORA &&
 186             st->codec->codec_id != CODEC_ID_SPEEX  &&
 187             st->codec->codec_id != CODEC_ID_FLAC) {
 188             av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
 189             return -1;
 190         }
 191
 192         if (!st->codec->extradata || !st->codec->extradata_size) {
 193             av_log(s, AV_LOG_ERROR, "No extradata present\n");
 194             return -1;
 195         }
 196         oggstream = av_mallocz(sizeof(*oggstream));
 197         st->priv_data = oggstream;
 198         if (st->codec->codec_id == CODEC_ID_FLAC) {
 199             int err = ogg_build_flac_headers(st->codec, oggstream,
 200                                              st->codec->flags & CODEC_FLAG_BITEXACT);
 201             if (err) {
 202                 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
 203                 av_freep(&st->priv_data);
 204                 return err;
 205             }
 206         } else if (st->codec->codec_id == CODEC_ID_SPEEX) {
 207             int err = ogg_build_speex_headers(st->codec, oggstream,
 208                                               st->codec->flags & CODEC_FLAG_BITEXACT);
 209             if (err) {
 210                 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
 211                 av_freep(&st->priv_data);
 212                 return err;
 213             }
 214         } else {
 215             if (ff_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
 216                                       st->codec->codec_id == CODEC_ID_VORBIS ? 30 : 42,
 217                                       oggstream->header, oggstream->header_len) < 0) {
 218                 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
 219                 av_freep(&st->priv_data);
 220                 return -1;
 221             }
 222             if (st->codec->codec_id == CODEC_ID_THEORA) {
 223                 /** KFGSHIFT is the width of the less significant section of the granule position
 224                     The less significant section is the frame count since the last keyframe */
 225                 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
 226                 oggstream->vrev = oggstream->header[0][9];
 227                 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
 228                        oggstream->kfgshift, oggstream->vrev);
 229             }
 230         }
 231     }
 232     for (i = 0; i < 3; i++) {
 233         for (j = 0; j < s->nb_streams; j++) {
 234             AVStream *st = s->streams[j];
 235             OGGStreamContext *oggstream = st->priv_data;
 236             if (oggstream && oggstream->header_len[i]) {
 237                 ogg_write_page(s, oggstream->header[i], oggstream->header_len[i],
 238                                0, st->index, i ? 0 : 2); // bos
 239             }
 240         }
 241     }
 242     return 0;
 243 }
 244
 245 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
 246 {
 247     AVStream *st = s->streams[pkt->stream_index];
 248     OGGStreamContext *oggstream = st->priv_data;
 249     uint8_t *ptr = pkt->data;
 250     int ret, size = pkt->size;
 251     int64_t granule;
 252
 253     if (st->codec->codec_id == CODEC_ID_THEORA) {
 254         int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
 255         int pframe_count;
 256         if (pkt->flags & PKT_FLAG_KEY)
 257             oggstream->last_kf_pts = pts;
 258         pframe_count = pts - oggstream->last_kf_pts;
 259         // prevent frame count from overflow if key frame flag is not set
 260         if (pframe_count >= (1<<oggstream->kfgshift)) {
 261             oggstream->last_kf_pts += pframe_count;
 262             pframe_count = 0;
 263         }
 264         granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
 265     } else
 266         granule = pkt->pts + pkt->duration;
 267     oggstream->duration = granule;
 268     do {
 269         ret = ogg_write_page(s, ptr, size, granule, pkt->stream_index, ptr != pkt->data);
 270         ptr  += ret; size -= ret;
 271     } while (size > 0 || ret == 255*255); // need to output a last nil page
 272
 273     return 0;
 274 }
 275
 276 static int ogg_compare_granule(AVFormatContext *s, AVPacket *next, AVPacket *pkt)
 277 {
 278     AVStream *st2 = s->streams[next->stream_index];
 279     AVStream *st  = s->streams[pkt ->stream_index];
 280
 281     int64_t next_granule = av_rescale_q(next->pts + next->duration,
 282                                         st2->time_base, AV_TIME_BASE_Q);
 283     int64_t cur_granule  = av_rescale_q(pkt ->pts + pkt ->duration,
 284                                         st ->time_base, AV_TIME_BASE_Q);
 285     return next_granule > cur_granule;
 286 }
 287
 288 static int ogg_interleave_per_granule(AVFormatContext *s, AVPacket *out, AVPacket *pkt, int flush)
 289 {
 290     OGGStreamContext *ogg;
 291     int i, stream_count = 0;
 292     int interleaved = 0;
 293
 294     if (pkt) {
 295         ff_interleave_add_packet(s, pkt, ogg_compare_granule);
 296         ogg = s->streams[pkt->stream_index]->priv_data;
 297         ogg->packet_count++;
 298     }
 299
 300     for (i = 0; i < s->nb_streams; i++) {
 301         ogg = s->streams[i]->priv_data;
 302         stream_count += !!ogg->packet_count;
 303         interleaved += ogg->packet_count > 1;
 304     }
 305
 306     if ((s->nb_streams == stream_count && interleaved == stream_count) ||
 307         (flush && stream_count)) {
 308         AVPacketList *pktl= s->packet_buffer;
 309         *out= pktl->pkt;
 310         s->packet_buffer = pktl->next;
 311
 312         ogg = s->streams[out->stream_index]->priv_data;
 313         if (flush && ogg->packet_count == 1)
 314             ogg->eos = 1;
 315         ogg->packet_count--;
 316
 317         if(!s->packet_buffer)
 318             s->packet_buffer_end= NULL;
 319
 320         if(s->streams[out->stream_index]->last_in_packet_buffer == pktl)
 321             s->streams[out->stream_index]->last_in_packet_buffer= NULL;
 322
 323         av_freep(&pktl);
 324         return 1;
 325     } else {
 326         av_init_packet(out);
 327         return 0;
 328     }
 329 }
 330
 331 static int ogg_write_trailer(AVFormatContext *s)
 332 {
 333     int i;
 334     for (i = 0; i < s->nb_streams; i++) {
 335         AVStream *st = s->streams[i];
 336         OGGStreamContext *oggstream = st->priv_data;
 337         if (st->codec->codec_id == CODEC_ID_FLAC ||
 338             st->codec->codec_id == CODEC_ID_SPEEX) {
 339             av_free(oggstream->header[0]);
 340             av_free(oggstream->header[1]);
 341         }
 342         av_freep(&st->priv_data);
 343     }
 344     return 0;
 345 }
 346
 347 AVOutputFormat ogg_muxer = {
 348     "ogg",
 349     NULL_IF_CONFIG_SMALL("Ogg"),
 350     "application/ogg",
 351     "ogg,ogv,spx",
 352     0,
 353     CODEC_ID_FLAC,
 354     CODEC_ID_THEORA,
 355     ogg_write_header,
 356     ogg_write_packet,
 357     ogg_write_trailer,
 358     .interleave_packet = ogg_interleave_per_granule,
 359 };