FFmpeg: libavformat/oggenc.c Source File

00001 /*
00002  * Ogg muxer
00003  * Copyright (c) 2007 Baptiste Coudurier <baptiste dot coudurier at free dot fr>
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00022 #include "libavutil/crc.h"
00023 #include "libavcodec/xiph.h"
00024 #include "libavcodec/bytestream.h"
00025 #include "libavcodec/flac.h"
00026 #include "avformat.h"
00027 #include "internal.h"
00028 #include "vorbiscomment.h"
00029 
00030 #define MAX_PAGE_SIZE 65025
00031 
00032 typedef struct {
00033     int64_t granule;
00034     int stream_index;
00035     uint8_t flags;
00036     uint8_t segments_count;
00037     uint8_t segments[255];
00038     uint8_t data[MAX_PAGE_SIZE];
00039     uint16_t size;
00040 } OGGPage;
00041 
00042 typedef struct {
00043     int64_t duration;
00044     unsigned page_counter;
00045     uint8_t *header[3];
00046     int header_len[3];
00048     int kfgshift;
00049     int64_t last_kf_pts;
00050     int vrev;
00051     int eos;
00052     unsigned page_count; 
00053     OGGPage page; 
00054 } OGGStreamContext;
00055 
00056 typedef struct OGGPageList {
00057     OGGPage page;
00058     struct OGGPageList *next;
00059 } OGGPageList;
00060 
00061 typedef struct {
00062     OGGPageList *page_list;
00063 } OGGContext;
00064 
00065 static void ogg_update_checksum(AVFormatContext *s, int64_t crc_offset)
00066 {
00067     int64_t pos = url_ftell(s->pb);
00068     uint32_t checksum = get_checksum(s->pb);
00069     url_fseek(s->pb, crc_offset, SEEK_SET);
00070     put_be32(s->pb, checksum);
00071     url_fseek(s->pb, pos, SEEK_SET);
00072 }
00073 
00074 static void ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
00075 {
00076     OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
00077     int64_t crc_offset;
00078 
00079     init_checksum(s->pb, ff_crc04C11DB7_update, 0);
00080     put_tag(s->pb, "OggS");
00081     put_byte(s->pb, 0);
00082     put_byte(s->pb, page->flags | extra_flags);
00083     put_le64(s->pb, page->granule);
00084     put_le32(s->pb, page->stream_index);
00085     put_le32(s->pb, oggstream->page_counter++);
00086     crc_offset = url_ftell(s->pb);
00087     put_le32(s->pb, 0); // crc
00088     put_byte(s->pb, page->segments_count);
00089     put_buffer(s->pb, page->segments, page->segments_count);
00090     put_buffer(s->pb, page->data, page->size);
00091 
00092     ogg_update_checksum(s, crc_offset);
00093     put_flush_packet(s->pb);
00094     oggstream->page_count--;
00095 }
00096 
00097 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, OGGPage *page)
00098 {
00099     if (oggstream->kfgshift)
00100         return (page->granule>>oggstream->kfgshift) +
00101             (page->granule & ((1<<oggstream->kfgshift)-1));
00102     else
00103         return page->granule;
00104 }
00105 
00106 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
00107 {
00108     AVStream *st2 = s->streams[next->stream_index];
00109     AVStream *st  = s->streams[page->stream_index];
00110     int64_t next_granule, cur_granule;
00111 
00112     if (next->granule == -1 || page->granule == -1)
00113         return 0;
00114 
00115     next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next),
00116                                 st2->time_base, AV_TIME_BASE_Q);
00117     cur_granule  = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page),
00118                                 st ->time_base, AV_TIME_BASE_Q);
00119     return next_granule > cur_granule;
00120 }
00121 
00122 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
00123 {
00124     oggstream->page.granule = -1;
00125     oggstream->page.flags = 0;
00126     oggstream->page.segments_count = 0;
00127     oggstream->page.size = 0;
00128     return 0;
00129 }
00130 
00131 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
00132 {
00133     OGGContext *ogg = s->priv_data;
00134     OGGPageList **p = &ogg->page_list;
00135     OGGPageList *l = av_mallocz(sizeof(*l));
00136 
00137     if (!l)
00138         return AVERROR(ENOMEM);
00139     l->page = oggstream->page;
00140 
00141     oggstream->page_count++;
00142     ogg_reset_cur_page(oggstream);
00143 
00144     while (*p) {
00145         if (ogg_compare_granule(s, &(*p)->page, &l->page))
00146             break;
00147         p = &(*p)->next;
00148     }
00149     l->next = *p;
00150     *p = l;
00151 
00152     return 0;
00153 }
00154 
00155 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
00156                            uint8_t *data, unsigned size, int64_t granule)
00157 {
00158     OGGStreamContext *oggstream = st->priv_data;
00159     int total_segments = size / 255 + 1;
00160     uint8_t *p = data;
00161     int i, segments, len;
00162 
00163     for (i = 0; i < total_segments; ) {
00164         OGGPage *page = &oggstream->page;
00165 
00166         segments = FFMIN(total_segments - i, 255 - page->segments_count);
00167 
00168         if (i && !page->segments_count)
00169             page->flags |= 1; // continued packet
00170 
00171         memset(page->segments+page->segments_count, 255, segments - 1);
00172         page->segments_count += segments - 1;
00173 
00174         len = FFMIN(size, segments*255);
00175         page->segments[page->segments_count++] = len - (segments-1)*255;
00176         memcpy(page->data+page->size, p, len);
00177         p += len;
00178         size -= len;
00179         i += segments;
00180         page->size += len;
00181 
00182         if (i == total_segments)
00183             page->granule = granule;
00184 
00185         if (page->segments_count == 255) {
00186             ogg_buffer_page(s, oggstream);
00187         }
00188     }
00189     return 0;
00190 }
00191 
00192 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
00193                                         int *header_len, AVMetadata *m)
00194 {
00195     const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
00196     int size;
00197     uint8_t *p, *p0;
00198     unsigned int count;
00199 
00200     size = offset + ff_vorbiscomment_length(m, vendor, &count);
00201     p = av_mallocz(size);
00202     if (!p)
00203         return NULL;
00204     p0 = p;
00205 
00206     p += offset;
00207     ff_vorbiscomment_write(&p, m, vendor, count);
00208 
00209     *header_len = size;
00210     return p0;
00211 }
00212 
00213 static int ogg_build_flac_headers(AVCodecContext *avctx,
00214                                   OGGStreamContext *oggstream, int bitexact,
00215                                   AVMetadata *m)
00216 {
00217     enum FLACExtradataFormat format;
00218     uint8_t *streaminfo;
00219     uint8_t *p;
00220 
00221     if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo))
00222         return -1;
00223 
00224     // first packet: STREAMINFO
00225     oggstream->header_len[0] = 51;
00226     oggstream->header[0] = av_mallocz(51); // per ogg flac specs
00227     p = oggstream->header[0];
00228     if (!p)
00229         return AVERROR(ENOMEM);
00230     bytestream_put_byte(&p, 0x7F);
00231     bytestream_put_buffer(&p, "FLAC", 4);
00232     bytestream_put_byte(&p, 1); // major version
00233     bytestream_put_byte(&p, 0); // minor version
00234     bytestream_put_be16(&p, 1); // headers packets without this one
00235     bytestream_put_buffer(&p, "fLaC", 4);
00236     bytestream_put_byte(&p, 0x00); // streaminfo
00237     bytestream_put_be24(&p, 34);
00238     bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
00239 
00240     // second packet: VorbisComment
00241     p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m);
00242     if (!p)
00243         return AVERROR(ENOMEM);
00244     oggstream->header[1] = p;
00245     bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
00246     bytestream_put_be24(&p, oggstream->header_len[1] - 4);
00247 
00248     return 0;
00249 }
00250 
00251 #define SPEEX_HEADER_SIZE 80
00252 
00253 static int ogg_build_speex_headers(AVCodecContext *avctx,
00254                                    OGGStreamContext *oggstream, int bitexact,
00255                                    AVMetadata *m)
00256 {
00257     uint8_t *p;
00258 
00259     if (avctx->extradata_size < SPEEX_HEADER_SIZE)
00260         return -1;
00261 
00262     // first packet: Speex header
00263     p = av_mallocz(SPEEX_HEADER_SIZE);
00264     if (!p)
00265         return AVERROR(ENOMEM);
00266     oggstream->header[0] = p;
00267     oggstream->header_len[0] = SPEEX_HEADER_SIZE;
00268     bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
00269     AV_WL32(&oggstream->header[0][68], 0);  // set extra_headers to 0
00270 
00271     // second packet: VorbisComment
00272     p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m);
00273     if (!p)
00274         return AVERROR(ENOMEM);
00275     oggstream->header[1] = p;
00276 
00277     return 0;
00278 }
00279 
00280 static int ogg_write_header(AVFormatContext *s)
00281 {
00282     OGGStreamContext *oggstream;
00283     int i, j;
00284     for (i = 0; i < s->nb_streams; i++) {
00285         AVStream *st = s->streams[i];
00286         if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
00287             av_set_pts_info(st, 64, 1, st->codec->sample_rate);
00288         else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
00289             av_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
00290         if (st->codec->codec_id != CODEC_ID_VORBIS &&
00291             st->codec->codec_id != CODEC_ID_THEORA &&
00292             st->codec->codec_id != CODEC_ID_SPEEX  &&
00293             st->codec->codec_id != CODEC_ID_FLAC) {
00294             av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
00295             return -1;
00296         }
00297 
00298         if (!st->codec->extradata || !st->codec->extradata_size) {
00299             av_log(s, AV_LOG_ERROR, "No extradata present\n");
00300             return -1;
00301         }
00302         oggstream = av_mallocz(sizeof(*oggstream));
00303         oggstream->page.stream_index = i;
00304         st->priv_data = oggstream;
00305         if (st->codec->codec_id == CODEC_ID_FLAC) {
00306             int err = ogg_build_flac_headers(st->codec, oggstream,
00307                                              st->codec->flags & CODEC_FLAG_BITEXACT,
00308                                              s->metadata);
00309             if (err) {
00310                 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
00311                 av_freep(&st->priv_data);
00312                 return err;
00313             }
00314         } else if (st->codec->codec_id == CODEC_ID_SPEEX) {
00315             int err = ogg_build_speex_headers(st->codec, oggstream,
00316                                               st->codec->flags & CODEC_FLAG_BITEXACT,
00317                                               s->metadata);
00318             if (err) {
00319                 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
00320                 av_freep(&st->priv_data);
00321                 return err;
00322             }
00323         } else {
00324             if (ff_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
00325                                       st->codec->codec_id == CODEC_ID_VORBIS ? 30 : 42,
00326                                       oggstream->header, oggstream->header_len) < 0) {
00327                 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
00328                 av_freep(&st->priv_data);
00329                 return -1;
00330             }
00331             if (st->codec->codec_id == CODEC_ID_THEORA) {
00334                 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
00335                 oggstream->vrev = oggstream->header[0][9];
00336                 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
00337                        oggstream->kfgshift, oggstream->vrev);
00338             }
00339         }
00340     }
00341 
00342     for (j = 0; j < s->nb_streams; j++) {
00343         OGGStreamContext *oggstream = s->streams[j]->priv_data;
00344         ogg_buffer_data(s, s->streams[j], oggstream->header[0],
00345                         oggstream->header_len[0], 0);
00346         oggstream->page.flags |= 2; // bos
00347         ogg_buffer_page(s, oggstream);
00348     }
00349     for (j = 0; j < s->nb_streams; j++) {
00350         AVStream *st = s->streams[j];
00351         OGGStreamContext *oggstream = st->priv_data;
00352         for (i = 1; i < 3; i++) {
00353             if (oggstream && oggstream->header_len[i])
00354                 ogg_buffer_data(s, st, oggstream->header[i],
00355                                 oggstream->header_len[i], 0);
00356         }
00357         ogg_buffer_page(s, oggstream);
00358     }
00359     return 0;
00360 }
00361 
00362 static void ogg_write_pages(AVFormatContext *s, int flush)
00363 {
00364     OGGContext *ogg = s->priv_data;
00365     OGGPageList *next, *p;
00366 
00367     if (!ogg->page_list)
00368         return;
00369 
00370     for (p = ogg->page_list; p; ) {
00371         OGGStreamContext *oggstream =
00372             s->streams[p->page.stream_index]->priv_data;
00373         if (oggstream->page_count < 2 && !flush)
00374             break;
00375         ogg_write_page(s, &p->page,
00376                        flush && oggstream->page_count == 1 ? 4 : 0); // eos
00377         next = p->next;
00378         av_freep(&p);
00379         p = next;
00380     }
00381     ogg->page_list = p;
00382 }
00383 
00384 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
00385 {
00386     AVStream *st = s->streams[pkt->stream_index];
00387     OGGStreamContext *oggstream = st->priv_data;
00388     int ret;
00389     int64_t granule;
00390 
00391     if (st->codec->codec_id == CODEC_ID_THEORA) {
00392         int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
00393         int pframe_count;
00394         if (pkt->flags & AV_PKT_FLAG_KEY)
00395             oggstream->last_kf_pts = pts;
00396         pframe_count = pts - oggstream->last_kf_pts;
00397         // prevent frame count from overflow if key frame flag is not set
00398         if (pframe_count >= (1<<oggstream->kfgshift)) {
00399             oggstream->last_kf_pts += pframe_count;
00400             pframe_count = 0;
00401         }
00402         granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
00403     } else
00404         granule = pkt->pts + pkt->duration;
00405     oggstream->duration = granule;
00406 
00407     ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule);
00408     if (ret < 0)
00409         return ret;
00410 
00411     ogg_write_pages(s, 0);
00412 
00413     return 0;
00414 }
00415 
00416 static int ogg_write_trailer(AVFormatContext *s)
00417 {
00418     int i;
00419 
00420     /* flush current page */
00421     for (i = 0; i < s->nb_streams; i++)
00422         ogg_buffer_page(s, s->streams[i]->priv_data);
00423 
00424     ogg_write_pages(s, 1);
00425 
00426     for (i = 0; i < s->nb_streams; i++) {
00427         AVStream *st = s->streams[i];
00428         OGGStreamContext *oggstream = st->priv_data;
00429         if (st->codec->codec_id == CODEC_ID_FLAC ||
00430             st->codec->codec_id == CODEC_ID_SPEEX) {
00431             av_free(oggstream->header[0]);
00432             av_free(oggstream->header[1]);
00433         }
00434         av_freep(&st->priv_data);
00435     }
00436     return 0;
00437 }
00438 
00439 AVOutputFormat ogg_muxer = {
00440     "ogg",
00441     NULL_IF_CONFIG_SMALL("Ogg"),
00442     "application/ogg",
00443     "ogg,ogv,spx",
00444     sizeof(OGGContext),
00445     CODEC_ID_FLAC,
00446     CODEC_ID_THEORA,
00447     ogg_write_header,
00448     ogg_write_packet,
00449     ogg_write_trailer,
00450     .metadata_conv = ff_vorbiscomment_metadata_conv,
00451 };