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 "libavutil/opt.h"
00024 #include "libavutil/random_seed.h"
00025 #include "libavcodec/xiph.h"
00026 #include "libavcodec/bytestream.h"
00027 #include "libavcodec/flac.h"
00028 #include "avformat.h"
00029 #include "avio_internal.h"
00030 #include "internal.h"
00031 #include "vorbiscomment.h"
00032 
00033 #define MAX_PAGE_SIZE 65025
00034 
00035 typedef struct {
00036     int64_t granule;
00037     int stream_index;
00038     uint8_t flags;
00039     uint8_t segments_count;
00040     uint8_t segments[255];
00041     uint8_t data[MAX_PAGE_SIZE];
00042     uint16_t size;
00043 } OGGPage;
00044 
00045 typedef struct {
00046     unsigned page_counter;
00047     uint8_t *header[3];
00048     int header_len[3];
00050     int kfgshift;
00051     int64_t last_kf_pts;
00052     int vrev;
00053     int eos;
00054     unsigned page_count; 
00055     OGGPage page; 
00056     unsigned serial_num; 
00057     int64_t last_granule; 
00058 } OGGStreamContext;
00059 
00060 typedef struct OGGPageList {
00061     OGGPage page;
00062     struct OGGPageList *next;
00063 } OGGPageList;
00064 
00065 typedef struct {
00066     const AVClass *class;
00067     OGGPageList *page_list;
00068     int pref_size; 
00069 } OGGContext;
00070 
00071 
00072 static const AVOption options[] = {
00073     { "oggpagesize", "Set preferred Ogg page size.",
00074       offsetof(OGGContext, pref_size), FF_OPT_TYPE_INT, {.dbl=0}, 0, MAX_PAGE_SIZE, AV_OPT_FLAG_ENCODING_PARAM},
00075     { NULL },
00076 };
00077 
00078 static const AVClass ogg_muxer_class = {
00079     "Ogg muxer",
00080     av_default_item_name,
00081     options,
00082     LIBAVUTIL_VERSION_INT,
00083 };
00084 
00085 
00086 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
00087 {
00088     int64_t pos = avio_tell(pb);
00089     uint32_t checksum = ffio_get_checksum(pb);
00090     avio_seek(pb, crc_offset, SEEK_SET);
00091     avio_wb32(pb, checksum);
00092     avio_seek(pb, pos, SEEK_SET);
00093 }
00094 
00095 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
00096 {
00097     OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
00098     AVIOContext *pb;
00099     int64_t crc_offset;
00100     int ret, size;
00101     uint8_t *buf;
00102 
00103     ret = avio_open_dyn_buf(&pb);
00104     if (ret < 0)
00105         return ret;
00106     ffio_init_checksum(pb, ff_crc04C11DB7_update, 0);
00107     ffio_wfourcc(pb, "OggS");
00108     avio_w8(pb, 0);
00109     avio_w8(pb, page->flags | extra_flags);
00110     avio_wl64(pb, page->granule);
00111     avio_wl32(pb, oggstream->serial_num);
00112     avio_wl32(pb, oggstream->page_counter++);
00113     crc_offset = avio_tell(pb);
00114     avio_wl32(pb, 0); // crc
00115     avio_w8(pb, page->segments_count);
00116     avio_write(pb, page->segments, page->segments_count);
00117     avio_write(pb, page->data, page->size);
00118 
00119     ogg_update_checksum(s, pb, crc_offset);
00120     avio_flush(pb);
00121 
00122     size = avio_close_dyn_buf(pb, &buf);
00123     if (size < 0)
00124         return size;
00125 
00126     avio_write(s->pb, buf, size);
00127     avio_flush(s->pb);
00128     av_free(buf);
00129     oggstream->page_count--;
00130     return 0;
00131 }
00132 
00133 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
00134 {
00135     if (oggstream->kfgshift)
00136         return (granule>>oggstream->kfgshift) +
00137             (granule & ((1<<oggstream->kfgshift)-1));
00138     else
00139         return granule;
00140 }
00141 
00142 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
00143 {
00144     AVStream *st2 = s->streams[next->stream_index];
00145     AVStream *st  = s->streams[page->stream_index];
00146     int64_t next_granule, cur_granule;
00147 
00148     if (next->granule == -1 || page->granule == -1)
00149         return 0;
00150 
00151     next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
00152                                 st2->time_base, AV_TIME_BASE_Q);
00153     cur_granule  = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
00154                                 st ->time_base, AV_TIME_BASE_Q);
00155     return next_granule > cur_granule;
00156 }
00157 
00158 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
00159 {
00160     oggstream->page.granule = -1;
00161     oggstream->page.flags = 0;
00162     oggstream->page.segments_count = 0;
00163     oggstream->page.size = 0;
00164     return 0;
00165 }
00166 
00167 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
00168 {
00169     OGGContext *ogg = s->priv_data;
00170     OGGPageList **p = &ogg->page_list;
00171     OGGPageList *l = av_mallocz(sizeof(*l));
00172 
00173     if (!l)
00174         return AVERROR(ENOMEM);
00175     l->page = oggstream->page;
00176 
00177     oggstream->page_count++;
00178     ogg_reset_cur_page(oggstream);
00179 
00180     while (*p) {
00181         if (ogg_compare_granule(s, &(*p)->page, &l->page))
00182             break;
00183         p = &(*p)->next;
00184     }
00185     l->next = *p;
00186     *p = l;
00187 
00188     return 0;
00189 }
00190 
00191 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
00192                            uint8_t *data, unsigned size, int64_t granule)
00193 {
00194     OGGStreamContext *oggstream = st->priv_data;
00195     OGGContext *ogg = s->priv_data;
00196     int total_segments = size / 255 + 1;
00197     uint8_t *p = data;
00198     int i, segments, len, flush = 0;
00199 
00200     // Handles VFR by flushing page because this frame needs to have a timestamp
00201     if (st->codec->codec_id == CODEC_ID_THEORA &&
00202         ogg_granule_to_timestamp(oggstream, granule) >
00203         ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1) {
00204         if (oggstream->page.granule != -1)
00205             ogg_buffer_page(s, oggstream);
00206         flush = 1;
00207     }
00208 
00209     for (i = 0; i < total_segments; ) {
00210         OGGPage *page = &oggstream->page;
00211 
00212         segments = FFMIN(total_segments - i, 255 - page->segments_count);
00213 
00214         if (i && !page->segments_count)
00215             page->flags |= 1; // continued packet
00216 
00217         memset(page->segments+page->segments_count, 255, segments - 1);
00218         page->segments_count += segments - 1;
00219 
00220         len = FFMIN(size, segments*255);
00221         page->segments[page->segments_count++] = len - (segments-1)*255;
00222         memcpy(page->data+page->size, p, len);
00223         p += len;
00224         size -= len;
00225         i += segments;
00226         page->size += len;
00227 
00228         if (i == total_segments)
00229             page->granule = granule;
00230 
00231         if(page->segments_count == 255 ||
00232            (ogg->pref_size > 0 && page->size >= ogg->pref_size)) {
00233            ogg_buffer_page(s, oggstream);
00234         }
00235     }
00236 
00237     if (flush && oggstream->page.granule != -1)
00238         ogg_buffer_page(s, oggstream);
00239 
00240     return 0;
00241 }
00242 
00243 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
00244                                         int *header_len, AVDictionary **m, int framing_bit)
00245 {
00246     const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
00247     int size;
00248     uint8_t *p, *p0;
00249     unsigned int count;
00250 
00251     ff_metadata_conv(m, ff_vorbiscomment_metadata_conv, NULL);
00252 
00253     size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
00254     p = av_mallocz(size);
00255     if (!p)
00256         return NULL;
00257     p0 = p;
00258 
00259     p += offset;
00260     ff_vorbiscomment_write(&p, m, vendor, count);
00261     if (framing_bit)
00262         bytestream_put_byte(&p, 1);
00263 
00264     *header_len = size;
00265     return p0;
00266 }
00267 
00268 static int ogg_build_flac_headers(AVCodecContext *avctx,
00269                                   OGGStreamContext *oggstream, int bitexact,
00270                                   AVDictionary **m)
00271 {
00272     enum FLACExtradataFormat format;
00273     uint8_t *streaminfo;
00274     uint8_t *p;
00275 
00276     if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo))
00277         return -1;
00278 
00279     // first packet: STREAMINFO
00280     oggstream->header_len[0] = 51;
00281     oggstream->header[0] = av_mallocz(51); // per ogg flac specs
00282     p = oggstream->header[0];
00283     if (!p)
00284         return AVERROR(ENOMEM);
00285     bytestream_put_byte(&p, 0x7F);
00286     bytestream_put_buffer(&p, "FLAC", 4);
00287     bytestream_put_byte(&p, 1); // major version
00288     bytestream_put_byte(&p, 0); // minor version
00289     bytestream_put_be16(&p, 1); // headers packets without this one
00290     bytestream_put_buffer(&p, "fLaC", 4);
00291     bytestream_put_byte(&p, 0x00); // streaminfo
00292     bytestream_put_be24(&p, 34);
00293     bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
00294 
00295     // second packet: VorbisComment
00296     p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
00297     if (!p)
00298         return AVERROR(ENOMEM);
00299     oggstream->header[1] = p;
00300     bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
00301     bytestream_put_be24(&p, oggstream->header_len[1] - 4);
00302 
00303     return 0;
00304 }
00305 
00306 #define SPEEX_HEADER_SIZE 80
00307 
00308 static int ogg_build_speex_headers(AVCodecContext *avctx,
00309                                    OGGStreamContext *oggstream, int bitexact,
00310                                    AVDictionary **m)
00311 {
00312     uint8_t *p;
00313 
00314     if (avctx->extradata_size < SPEEX_HEADER_SIZE)
00315         return -1;
00316 
00317     // first packet: Speex header
00318     p = av_mallocz(SPEEX_HEADER_SIZE);
00319     if (!p)
00320         return AVERROR(ENOMEM);
00321     oggstream->header[0] = p;
00322     oggstream->header_len[0] = SPEEX_HEADER_SIZE;
00323     bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
00324     AV_WL32(&oggstream->header[0][68], 0);  // set extra_headers to 0
00325 
00326     // second packet: VorbisComment
00327     p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
00328     if (!p)
00329         return AVERROR(ENOMEM);
00330     oggstream->header[1] = p;
00331 
00332     return 0;
00333 }
00334 
00335 static int ogg_write_header(AVFormatContext *s)
00336 {
00337     OGGStreamContext *oggstream;
00338     int i, j;
00339 
00340     for (i = 0; i < s->nb_streams; i++) {
00341         AVStream *st = s->streams[i];
00342         unsigned serial_num = i;
00343 
00344         if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
00345             av_set_pts_info(st, 64, 1, st->codec->sample_rate);
00346         else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
00347             av_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
00348         if (st->codec->codec_id != CODEC_ID_VORBIS &&
00349             st->codec->codec_id != CODEC_ID_THEORA &&
00350             st->codec->codec_id != CODEC_ID_SPEEX  &&
00351             st->codec->codec_id != CODEC_ID_FLAC) {
00352             av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
00353             return -1;
00354         }
00355 
00356         if (!st->codec->extradata || !st->codec->extradata_size) {
00357             av_log(s, AV_LOG_ERROR, "No extradata present\n");
00358             return -1;
00359         }
00360         oggstream = av_mallocz(sizeof(*oggstream));
00361         oggstream->page.stream_index = i;
00362 
00363         if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
00364             do {
00365                 serial_num = av_get_random_seed();
00366                 for (j = 0; j < i; j++) {
00367                     OGGStreamContext *sc = s->streams[j]->priv_data;
00368                     if (serial_num == sc->serial_num)
00369                         break;
00370                 }
00371             } while (j < i);
00372         oggstream->serial_num = serial_num;
00373 
00374         st->priv_data = oggstream;
00375         if (st->codec->codec_id == CODEC_ID_FLAC) {
00376             int err = ogg_build_flac_headers(st->codec, oggstream,
00377                                              st->codec->flags & CODEC_FLAG_BITEXACT,
00378                                              &s->metadata);
00379             if (err) {
00380                 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
00381                 av_freep(&st->priv_data);
00382                 return err;
00383             }
00384         } else if (st->codec->codec_id == CODEC_ID_SPEEX) {
00385             int err = ogg_build_speex_headers(st->codec, oggstream,
00386                                               st->codec->flags & CODEC_FLAG_BITEXACT,
00387                                               &s->metadata);
00388             if (err) {
00389                 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
00390                 av_freep(&st->priv_data);
00391                 return err;
00392             }
00393         } else {
00394             uint8_t *p;
00395             const char *cstr = st->codec->codec_id == CODEC_ID_VORBIS ? "vorbis" : "theora";
00396             int header_type = st->codec->codec_id == CODEC_ID_VORBIS ? 3 : 0x81;
00397             int framing_bit = st->codec->codec_id == CODEC_ID_VORBIS ? 1 : 0;
00398 
00399             if (ff_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
00400                                       st->codec->codec_id == CODEC_ID_VORBIS ? 30 : 42,
00401                                       oggstream->header, oggstream->header_len) < 0) {
00402                 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
00403                 av_freep(&st->priv_data);
00404                 return -1;
00405             }
00406 
00407             p = ogg_write_vorbiscomment(7, st->codec->flags & CODEC_FLAG_BITEXACT,
00408                                         &oggstream->header_len[1], &s->metadata,
00409                                         framing_bit);
00410             if (!p)
00411                 return AVERROR(ENOMEM);
00412 
00413             oggstream->header[1] = p;
00414             bytestream_put_byte(&p, header_type);
00415             bytestream_put_buffer(&p, cstr, 6);
00416 
00417             if (st->codec->codec_id == CODEC_ID_THEORA) {
00420                 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
00421                 oggstream->vrev = oggstream->header[0][9];
00422                 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
00423                        oggstream->kfgshift, oggstream->vrev);
00424             }
00425         }
00426     }
00427 
00428     for (j = 0; j < s->nb_streams; j++) {
00429         OGGStreamContext *oggstream = s->streams[j]->priv_data;
00430         ogg_buffer_data(s, s->streams[j], oggstream->header[0],
00431                         oggstream->header_len[0], 0);
00432         oggstream->page.flags |= 2; // bos
00433         ogg_buffer_page(s, oggstream);
00434     }
00435     for (j = 0; j < s->nb_streams; j++) {
00436         AVStream *st = s->streams[j];
00437         OGGStreamContext *oggstream = st->priv_data;
00438         for (i = 1; i < 3; i++) {
00439             if (oggstream && oggstream->header_len[i])
00440                 ogg_buffer_data(s, st, oggstream->header[i],
00441                                 oggstream->header_len[i], 0);
00442         }
00443         ogg_buffer_page(s, oggstream);
00444     }
00445     return 0;
00446 }
00447 
00448 static void ogg_write_pages(AVFormatContext *s, int flush)
00449 {
00450     OGGContext *ogg = s->priv_data;
00451     OGGPageList *next, *p;
00452 
00453     if (!ogg->page_list)
00454         return;
00455 
00456     for (p = ogg->page_list; p; ) {
00457         OGGStreamContext *oggstream =
00458             s->streams[p->page.stream_index]->priv_data;
00459         if (oggstream->page_count < 2 && !flush)
00460             break;
00461         ogg_write_page(s, &p->page,
00462                        flush && oggstream->page_count == 1 ? 4 : 0); // eos
00463         next = p->next;
00464         av_freep(&p);
00465         p = next;
00466     }
00467     ogg->page_list = p;
00468 }
00469 
00470 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
00471 {
00472     AVStream *st = s->streams[pkt->stream_index];
00473     OGGStreamContext *oggstream = st->priv_data;
00474     int ret;
00475     int64_t granule;
00476 
00477     if (st->codec->codec_id == CODEC_ID_THEORA) {
00478         int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
00479         int pframe_count;
00480         if (pkt->flags & AV_PKT_FLAG_KEY)
00481             oggstream->last_kf_pts = pts;
00482         pframe_count = pts - oggstream->last_kf_pts;
00483         // prevent frame count from overflow if key frame flag is not set
00484         if (pframe_count >= (1<<oggstream->kfgshift)) {
00485             oggstream->last_kf_pts += pframe_count;
00486             pframe_count = 0;
00487         }
00488         granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
00489     } else
00490         granule = pkt->pts + pkt->duration;
00491 
00492     ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule);
00493     if (ret < 0)
00494         return ret;
00495 
00496     ogg_write_pages(s, 0);
00497 
00498     oggstream->last_granule = granule;
00499 
00500     return 0;
00501 }
00502 
00503 static int ogg_write_trailer(AVFormatContext *s)
00504 {
00505     int i;
00506 
00507     /* flush current page */
00508     for (i = 0; i < s->nb_streams; i++)
00509         ogg_buffer_page(s, s->streams[i]->priv_data);
00510 
00511     ogg_write_pages(s, 1);
00512 
00513     for (i = 0; i < s->nb_streams; i++) {
00514         AVStream *st = s->streams[i];
00515         OGGStreamContext *oggstream = st->priv_data;
00516         if (st->codec->codec_id == CODEC_ID_FLAC ||
00517             st->codec->codec_id == CODEC_ID_SPEEX) {
00518             av_free(oggstream->header[0]);
00519             av_free(oggstream->header[1]);
00520         }
00521         av_freep(&st->priv_data);
00522     }
00523     return 0;
00524 }
00525 
00526 AVOutputFormat ff_ogg_muxer = {
00527     "ogg",
00528     NULL_IF_CONFIG_SMALL("Ogg"),
00529     "application/ogg",
00530     "ogg,ogv,spx",
00531     sizeof(OGGContext),
00532     CODEC_ID_FLAC,
00533     CODEC_ID_THEORA,
00534     ogg_write_header,
00535     ogg_write_packet,
00536     ogg_write_trailer,
00537     .priv_class = &ogg_muxer_class,
00538 };