2 * Copyright (C) 2005 Michael Ahlberg, Måns Rullgård
4 * Permission is hereby granted, free of charge, to any person
5 * obtaining a copy of this software and associated documentation
6 * files (the "Software"), to deal in the Software without
7 * restriction, including without limitation the rights to use, copy,
8 * modify, merge, publish, distribute, sublicense, and/or sell copies
9 * of the Software, and to permit persons to whom the Software is
10 * furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice shall be
13 * included in all copies or substantial portions of the Software.
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
17 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
19 * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
20 * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
27 #include "libavutil/avstring.h"
28 #include "libavutil/base64.h"
29 #include "libavutil/bswap.h"
30 #include "libavutil/dict.h"
31 #include "libavcodec/bytestream.h"
32 #include "libavcodec/get_bits.h"
33 #include "libavcodec/vorbis_parser.h"
35 #include "flac_picture.h"
38 #include "vorbiscomment.h"
40 static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
42 int i, cnum, h, m, s, ms, keylen = strlen(key);
43 AVChapter *chapter = NULL;
45 if (keylen < 9 || sscanf(key, "CHAPTER%03d", &cnum) != 1)
49 if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
52 avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 },
53 ms + 1000 * (s + 60 * (m + 60 * h)),
54 AV_NOPTS_VALUE, NULL);
56 } else if (!strcmp(key + keylen - 4, "NAME")) {
57 for (i = 0; i < as->nb_chapters; i++)
58 if (as->chapters[i]->id == cnum) {
59 chapter = as->chapters[i];
65 av_dict_set(&chapter->metadata, "title", val, AV_DICT_DONT_STRDUP_VAL);
73 int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
74 const uint8_t *buf, int size)
76 const uint8_t *p = buf;
77 const uint8_t *end = buf + size;
81 /* must have vendor_length and user_comment_list_length */
83 return AVERROR_INVALIDDATA;
85 s = bytestream_get_le32(&p);
87 if (end - p - 4 < s || s < 0)
88 return AVERROR_INVALIDDATA;
92 n = bytestream_get_le32(&p);
94 while (end - p >= 4 && n > 0) {
98 s = bytestream_get_le32(&p);
100 if (end - p < s || s < 0)
107 v = memchr(t, '=', s);
118 tt = av_malloc(tl + 1);
119 ct = av_malloc(vl + 1);
123 return AVERROR(ENOMEM);
126 for (j = 0; j < tl; j++)
127 tt[j] = av_toupper(t[j]);
133 /* The format in which the pictures are stored is the FLAC format.
134 * Xiph says: "The binary FLAC picture structure is base64 encoded
135 * and placed within a VorbisComment with the tag name
136 * 'METADATA_BLOCK_PICTURE'. This is the preferred and
137 * recommended way of embedding cover art within VorbisComments."
139 if (!strcmp(tt, "METADATA_BLOCK_PICTURE")) {
141 char *pict = av_malloc(vl);
144 av_log(as, AV_LOG_WARNING, "out-of-memory error. Skipping cover art block.\n");
149 if ((ret = av_base64_decode(pict, ct, vl)) > 0)
150 ret = ff_flac_parse_picture(as, pict, ret);
155 av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
158 } else if (!ogm_chapter(as, tt, ct)) {
159 if (m && av_dict_get(*m, tt, NULL, 0)) {
160 av_dict_set(m, tt, ";", AV_DICT_APPEND);
162 av_dict_set(m, tt, ct,
163 AV_DICT_DONT_STRDUP_KEY |
171 av_log(as, AV_LOG_INFO,
172 "%ti bytes of comment header remain\n", end - p);
174 av_log(as, AV_LOG_INFO,
175 "truncated comment header, %i comments not found\n", n);
177 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
183 * Parse the vorbis header
185 * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
186 * [vorbis_version] = read 32 bits as unsigned integer | Not used
187 * [audio_channels] = read 8 bit integer as unsigned | Used
188 * [audio_sample_rate] = read 32 bits as unsigned integer | Used
189 * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
190 * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
191 * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
192 * [blocksize_0] = read 4 bits as unsigned integer | Not Used
193 * [blocksize_1] = read 4 bits as unsigned integer | Not Used
194 * [framing_flag] = read one bit | Not Used
197 struct oggvorbis_private {
199 unsigned char *packet[3];
200 VorbisParseContext vp;
205 static int fixup_vorbis_headers(AVFormatContext *as,
206 struct oggvorbis_private *priv,
209 int i, offset, len, err;
213 len = priv->len[0] + priv->len[1] + priv->len[2];
214 buf_len = len + len / 255 + 64;
215 ptr = *buf = av_realloc(NULL, buf_len);
217 return AVERROR(ENOMEM);
218 memset(*buf, '\0', buf_len);
222 offset += av_xiphlacing(&ptr[offset], priv->len[0]);
223 offset += av_xiphlacing(&ptr[offset], priv->len[1]);
224 for (i = 0; i < 3; i++) {
225 memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
226 offset += priv->len[i];
227 av_freep(&priv->packet[i]);
229 if ((err = av_reallocp(buf, offset + FF_INPUT_BUFFER_PADDING_SIZE)) < 0)
234 static void vorbis_cleanup(AVFormatContext *s, int idx)
236 struct ogg *ogg = s->priv_data;
237 struct ogg_stream *os = ogg->streams + idx;
238 struct oggvorbis_private *priv = os->private;
241 for (i = 0; i < 3; i++)
242 av_freep(&priv->packet[i]);
245 static int vorbis_update_metadata(AVFormatContext *s, int idx)
247 struct ogg *ogg = s->priv_data;
248 struct ogg_stream *os = ogg->streams + idx;
249 AVStream *st = s->streams[idx];
255 /* New metadata packet; release old data. */
256 av_dict_free(&st->metadata);
257 ret = ff_vorbis_comment(s, &st->metadata, os->buf + os->pstart + 7,
262 /* Update the metadata if possible. */
263 av_freep(&os->new_metadata);
265 os->new_metadata = av_packet_pack_dictionary(st->metadata, &os->new_metadata_size);
266 /* Send an empty dictionary to indicate that metadata has been cleared. */
268 os->new_metadata = av_malloc(1);
269 os->new_metadata_size = 0;
275 static int vorbis_header(AVFormatContext *s, int idx)
277 struct ogg *ogg = s->priv_data;
278 AVStream *st = s->streams[idx];
279 struct ogg_stream *os = ogg->streams + idx;
280 struct oggvorbis_private *priv;
281 int pkt_type = os->buf[os->pstart];
284 os->private = av_mallocz(sizeof(struct oggvorbis_private));
286 return AVERROR(ENOMEM);
292 if (os->psize < 1 || pkt_type > 5)
293 return AVERROR_INVALIDDATA;
297 if (priv->packet[pkt_type >> 1])
298 return AVERROR_INVALIDDATA;
299 if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
300 return AVERROR_INVALIDDATA;
302 priv->len[pkt_type >> 1] = os->psize;
303 priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
304 if (!priv->packet[pkt_type >> 1])
305 return AVERROR(ENOMEM);
306 memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
307 if (os->buf[os->pstart] == 1) {
308 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
309 unsigned blocksize, bs0, bs1;
314 return AVERROR_INVALIDDATA;
316 if (bytestream_get_le32(&p) != 0) /* vorbis_version */
317 return AVERROR_INVALIDDATA;
319 channels = bytestream_get_byte(&p);
320 if (st->codec->channels && channels != st->codec->channels) {
321 av_log(s, AV_LOG_ERROR, "Channel change is not supported\n");
322 return AVERROR_PATCHWELCOME;
324 st->codec->channels = channels;
325 srate = bytestream_get_le32(&p);
326 p += 4; // skip maximum bitrate
327 st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
328 p += 4; // skip minimum bitrate
330 blocksize = bytestream_get_byte(&p);
331 bs0 = blocksize & 15;
332 bs1 = blocksize >> 4;
335 return AVERROR_INVALIDDATA;
336 if (bs0 < 6 || bs1 > 13)
337 return AVERROR_INVALIDDATA;
339 if (bytestream_get_byte(&p) != 1) /* framing_flag */
340 return AVERROR_INVALIDDATA;
342 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
343 st->codec->codec_id = AV_CODEC_ID_VORBIS;
346 st->codec->sample_rate = srate;
347 avpriv_set_pts_info(st, 64, 1, srate);
349 } else if (os->buf[os->pstart] == 3) {
350 if (vorbis_update_metadata(s, idx) >= 0) {
351 // drop all metadata we parsed and which is not required by libvorbis
352 unsigned new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
353 if (new_len >= 16 && new_len < os->psize) {
354 AV_WL32(priv->packet[1] + new_len - 5, 0);
355 priv->packet[1][new_len - 1] = 1;
356 priv->len[1] = new_len;
360 int ret = fixup_vorbis_headers(s, priv, &st->codec->extradata);
362 st->codec->extradata_size = 0;
365 st->codec->extradata_size = ret;
366 if ((ret = avpriv_vorbis_parse_extradata(st->codec, &priv->vp))) {
367 av_freep(&st->codec->extradata);
368 st->codec->extradata_size = 0;
376 static int vorbis_packet(AVFormatContext *s, int idx)
378 struct ogg *ogg = s->priv_data;
379 struct ogg_stream *os = ogg->streams + idx;
380 struct oggvorbis_private *priv = os->private;
381 int duration, flags = 0;
383 /* first packet handling
384 * here we parse the duration of each packet in the first page and compare
385 * the total duration to the page granule to find the encoder delay and
386 * set the first timestamp */
387 if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & OGG_FLAG_EOS)) {
389 uint8_t *last_pkt = os->buf + os->pstart;
390 uint8_t *next_pkt = last_pkt;
392 avpriv_vorbis_parse_reset(&priv->vp);
395 d = avpriv_vorbis_parse_frame_flags(&priv->vp, last_pkt, 1, &flags);
397 os->pflags |= AV_PKT_FLAG_CORRUPT;
399 } else if (flags & VORBIS_FLAG_COMMENT) {
400 vorbis_update_metadata(s, idx);
404 last_pkt = next_pkt = next_pkt + os->psize;
405 for (; seg < os->nsegs; seg++) {
406 if (os->segments[seg] < 255) {
407 int d = avpriv_vorbis_parse_frame_flags(&priv->vp, last_pkt, 1, &flags);
409 duration = os->granule;
411 } else if (flags & VORBIS_FLAG_COMMENT) {
412 vorbis_update_metadata(s, idx);
416 last_pkt = next_pkt + os->segments[seg];
418 next_pkt += os->segments[seg];
421 os->lastdts = os->granule - duration;
422 if (s->streams[idx]->start_time == AV_NOPTS_VALUE) {
423 s->streams[idx]->start_time = FFMAX(os->lastpts, 0);
424 if (s->streams[idx]->duration)
425 s->streams[idx]->duration -= s->streams[idx]->start_time;
427 priv->final_pts = AV_NOPTS_VALUE;
428 avpriv_vorbis_parse_reset(&priv->vp);
431 /* parse packet duration */
433 duration = avpriv_vorbis_parse_frame_flags(&priv->vp, os->buf + os->pstart, 1, &flags);
435 os->pflags |= AV_PKT_FLAG_CORRUPT;
437 } else if (flags & VORBIS_FLAG_COMMENT) {
438 vorbis_update_metadata(s, idx);
441 os->pduration = duration;
444 /* final packet handling
445 * here we save the pts of the first packet in the final page, sum up all
446 * packet durations in the final page except for the last one, and compare
447 * to the page granule to find the duration of the final packet */
448 if (os->flags & OGG_FLAG_EOS) {
449 if (os->lastpts != AV_NOPTS_VALUE) {
450 priv->final_pts = os->lastpts;
451 priv->final_duration = 0;
453 if (os->segp == os->nsegs)
454 os->pduration = os->granule - priv->final_pts - priv->final_duration;
455 priv->final_duration += os->pduration;
461 const struct ogg_codec ff_vorbis_codec = {
462 .magic = "\001vorbis",
464 .header = vorbis_header,
465 .packet = vorbis_packet,
466 .cleanup = vorbis_cleanup,