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"
32 #include "libavcodec/bytestream.h"
33 #include "libavcodec/vorbis_parser.h"
36 #include "flac_picture.h"
39 #include "vorbiscomment.h"
40 #include "replaygain.h"
42 static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
44 int i, cnum, h, m, s, ms, keylen = strlen(key);
45 AVChapter *chapter = NULL;
47 if (keylen < 9 || sscanf(key, "CHAPTER%03d", &cnum) != 1)
51 if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
54 avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 },
55 ms + 1000 * (s + 60 * (m + 60 * h)),
56 AV_NOPTS_VALUE, NULL);
58 } else if (!strcmp(key + keylen - 4, "NAME")) {
59 for (i = 0; i < as->nb_chapters; i++)
60 if (as->chapters[i]->id == cnum) {
61 chapter = as->chapters[i];
67 av_dict_set(&chapter->metadata, "title", val, AV_DICT_DONT_STRDUP_VAL);
75 int ff_vorbis_stream_comment(AVFormatContext *as, AVStream *st,
76 const uint8_t *buf, int size)
78 int updates = ff_vorbis_comment(as, &st->metadata, buf, size, 1);
81 st->event_flags |= AVSTREAM_EVENT_FLAG_METADATA_UPDATED;
87 int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
88 const uint8_t *buf, int size,
91 const uint8_t *p = buf;
92 const uint8_t *end = buf + size;
97 /* must have vendor_length and user_comment_list_length */
99 return AVERROR_INVALIDDATA;
101 s = bytestream_get_le32(&p);
103 if (end - p - 4 < s || s < 0)
104 return AVERROR_INVALIDDATA;
108 n = bytestream_get_le32(&p);
110 while (end - p >= 4 && n > 0) {
114 s = bytestream_get_le32(&p);
116 if (end - p < s || s < 0)
123 v = memchr(t, '=', s);
134 tt = av_malloc(tl + 1);
135 ct = av_malloc(vl + 1);
139 return AVERROR(ENOMEM);
142 for (j = 0; j < tl; j++)
143 tt[j] = av_toupper(t[j]);
149 /* The format in which the pictures are stored is the FLAC format.
150 * Xiph says: "The binary FLAC picture structure is base64 encoded
151 * and placed within a VorbisComment with the tag name
152 * 'METADATA_BLOCK_PICTURE'. This is the preferred and
153 * recommended way of embedding cover art within VorbisComments."
155 if (!strcmp(tt, "METADATA_BLOCK_PICTURE") && parse_picture) {
156 int ret, len = AV_BASE64_DECODE_SIZE(vl);
157 char *pict = av_malloc(len);
160 av_log(as, AV_LOG_WARNING, "out-of-memory error. Skipping cover art block.\n");
165 ret = av_base64_decode(pict, ct, len);
169 ret = ff_flac_parse_picture(as, pict, ret);
172 av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
175 } else if (!ogm_chapter(as, tt, ct)) {
177 if (av_dict_get(*m, tt, NULL, 0)) {
178 av_dict_set(m, tt, ";", AV_DICT_APPEND);
180 av_dict_set(m, tt, ct,
181 AV_DICT_DONT_STRDUP_KEY |
189 av_log(as, AV_LOG_INFO,
190 "%"PTRDIFF_SPECIFIER" bytes of comment header remain\n", end - p);
192 av_log(as, AV_LOG_INFO,
193 "truncated comment header, %i comments not found\n", n);
195 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
201 * Parse the vorbis header
203 * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
204 * [vorbis_version] = read 32 bits as unsigned integer | Not used
205 * [audio_channels] = read 8 bit integer as unsigned | Used
206 * [audio_sample_rate] = read 32 bits as unsigned integer | Used
207 * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
208 * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
209 * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
210 * [blocksize_0] = read 4 bits as unsigned integer | Not Used
211 * [blocksize_1] = read 4 bits as unsigned integer | Not Used
212 * [framing_flag] = read one bit | Not Used
215 struct oggvorbis_private {
217 unsigned char *packet[3];
218 AVVorbisParseContext *vp;
223 static int fixup_vorbis_headers(AVFormatContext *as,
224 struct oggvorbis_private *priv,
227 int i, offset, len, err;
231 len = priv->len[0] + priv->len[1] + priv->len[2];
232 buf_len = len + len / 255 + 64;
233 ptr = *buf = av_realloc(NULL, buf_len);
235 return AVERROR(ENOMEM);
236 memset(*buf, '\0', buf_len);
240 offset += av_xiphlacing(&ptr[offset], priv->len[0]);
241 offset += av_xiphlacing(&ptr[offset], priv->len[1]);
242 for (i = 0; i < 3; i++) {
243 memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
244 offset += priv->len[i];
245 av_freep(&priv->packet[i]);
247 if ((err = av_reallocp(buf, offset + AV_INPUT_BUFFER_PADDING_SIZE)) < 0)
252 static void vorbis_cleanup(AVFormatContext *s, int idx)
254 struct ogg *ogg = s->priv_data;
255 struct ogg_stream *os = ogg->streams + idx;
256 struct oggvorbis_private *priv = os->private;
259 av_vorbis_parse_free(&priv->vp);
260 for (i = 0; i < 3; i++)
261 av_freep(&priv->packet[i]);
265 static int vorbis_update_metadata(AVFormatContext *s, int idx)
267 struct ogg *ogg = s->priv_data;
268 struct ogg_stream *os = ogg->streams + idx;
269 AVStream *st = s->streams[idx];
275 /* New metadata packet; release old data. */
276 av_dict_free(&st->metadata);
277 ret = ff_vorbis_stream_comment(s, st, os->buf + os->pstart + 7,
282 /* Update the metadata if possible. */
283 av_freep(&os->new_metadata);
285 os->new_metadata = av_packet_pack_dictionary(st->metadata, &os->new_metadata_size);
286 /* Send an empty dictionary to indicate that metadata has been cleared. */
288 os->new_metadata = av_malloc(1);
289 os->new_metadata_size = 0;
295 static int vorbis_header(AVFormatContext *s, int idx)
297 struct ogg *ogg = s->priv_data;
298 AVStream *st = s->streams[idx];
299 struct ogg_stream *os = ogg->streams + idx;
300 struct oggvorbis_private *priv;
301 int pkt_type = os->buf[os->pstart];
304 os->private = av_mallocz(sizeof(struct oggvorbis_private));
306 return AVERROR(ENOMEM);
312 return priv->vp ? 0 : AVERROR_INVALIDDATA;
314 if (os->psize < 1 || pkt_type > 5)
315 return AVERROR_INVALIDDATA;
317 if (priv->packet[pkt_type >> 1])
318 return AVERROR_INVALIDDATA;
319 if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
320 return AVERROR_INVALIDDATA;
322 priv->len[pkt_type >> 1] = os->psize;
323 priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
324 if (!priv->packet[pkt_type >> 1])
325 return AVERROR(ENOMEM);
326 memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
327 if (os->buf[os->pstart] == 1) {
328 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
329 unsigned blocksize, bs0, bs1;
334 return AVERROR_INVALIDDATA;
336 if (bytestream_get_le32(&p) != 0) /* vorbis_version */
337 return AVERROR_INVALIDDATA;
339 channels = bytestream_get_byte(&p);
340 if (st->codecpar->channels && channels != st->codecpar->channels) {
341 av_log(s, AV_LOG_ERROR, "Channel change is not supported\n");
342 return AVERROR_PATCHWELCOME;
344 st->codecpar->channels = channels;
345 srate = bytestream_get_le32(&p);
346 p += 4; // skip maximum bitrate
347 st->codecpar->bit_rate = bytestream_get_le32(&p); // nominal bitrate
348 p += 4; // skip minimum bitrate
350 blocksize = bytestream_get_byte(&p);
351 bs0 = blocksize & 15;
352 bs1 = blocksize >> 4;
355 return AVERROR_INVALIDDATA;
356 if (bs0 < 6 || bs1 > 13)
357 return AVERROR_INVALIDDATA;
359 if (bytestream_get_byte(&p) != 1) /* framing_flag */
360 return AVERROR_INVALIDDATA;
362 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
363 st->codecpar->codec_id = AV_CODEC_ID_VORBIS;
366 st->codecpar->sample_rate = srate;
367 avpriv_set_pts_info(st, 64, 1, srate);
369 } else if (os->buf[os->pstart] == 3) {
370 if (vorbis_update_metadata(s, idx) >= 0 && priv->len[1] > 10) {
373 int ret = ff_replaygain_export(st, st->metadata);
377 // drop all metadata we parsed and which is not required by libvorbis
378 new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
379 if (new_len >= 16 && new_len < os->psize) {
380 AV_WL32(priv->packet[1] + new_len - 5, 0);
381 priv->packet[1][new_len - 1] = 1;
382 priv->len[1] = new_len;
386 int ret = fixup_vorbis_headers(s, priv, &st->codecpar->extradata);
388 st->codecpar->extradata_size = 0;
391 st->codecpar->extradata_size = ret;
393 priv->vp = av_vorbis_parse_init(st->codecpar->extradata, st->codecpar->extradata_size);
395 av_freep(&st->codecpar->extradata);
396 st->codecpar->extradata_size = 0;
397 return AVERROR_UNKNOWN;
404 static int vorbis_packet(AVFormatContext *s, int idx)
406 struct ogg *ogg = s->priv_data;
407 struct ogg_stream *os = ogg->streams + idx;
408 struct oggvorbis_private *priv = os->private;
409 int duration, flags = 0;
412 return AVERROR_INVALIDDATA;
414 /* first packet handling
415 * here we parse the duration of each packet in the first page and compare
416 * the total duration to the page granule to find the encoder delay and
417 * set the first timestamp */
418 if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & OGG_FLAG_EOS) && (int64_t)os->granule>=0) {
420 uint8_t *last_pkt = os->buf + os->pstart;
421 uint8_t *next_pkt = last_pkt;
423 av_vorbis_parse_reset(priv->vp);
426 d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
428 os->pflags |= AV_PKT_FLAG_CORRUPT;
430 } else if (flags & VORBIS_FLAG_COMMENT) {
431 vorbis_update_metadata(s, idx);
435 last_pkt = next_pkt = next_pkt + os->psize;
436 for (; seg < os->nsegs; seg++) {
437 if (os->segments[seg] < 255) {
438 int d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
440 duration = os->granule;
442 } else if (flags & VORBIS_FLAG_COMMENT) {
443 vorbis_update_metadata(s, idx);
447 last_pkt = next_pkt + os->segments[seg];
449 next_pkt += os->segments[seg];
452 os->lastdts = os->granule - duration;
454 if (!os->granule && duration) //hack to deal with broken files (Ticket3710)
455 os->lastpts = os->lastdts = AV_NOPTS_VALUE;
457 if (s->streams[idx]->start_time == AV_NOPTS_VALUE) {
458 s->streams[idx]->start_time = FFMAX(os->lastpts, 0);
459 if (s->streams[idx]->duration != AV_NOPTS_VALUE)
460 s->streams[idx]->duration -= s->streams[idx]->start_time;
462 priv->final_pts = AV_NOPTS_VALUE;
463 av_vorbis_parse_reset(priv->vp);
466 /* parse packet duration */
468 duration = av_vorbis_parse_frame_flags(priv->vp, os->buf + os->pstart, 1, &flags);
470 os->pflags |= AV_PKT_FLAG_CORRUPT;
472 } else if (flags & VORBIS_FLAG_COMMENT) {
473 vorbis_update_metadata(s, idx);
476 os->pduration = duration;
479 /* final packet handling
480 * here we save the pts of the first packet in the final page, sum up all
481 * packet durations in the final page except for the last one, and compare
482 * to the page granule to find the duration of the final packet */
483 if (os->flags & OGG_FLAG_EOS) {
484 if (os->lastpts != AV_NOPTS_VALUE) {
485 priv->final_pts = os->lastpts;
486 priv->final_duration = 0;
488 if (os->segp == os->nsegs)
489 os->pduration = os->granule - priv->final_pts - priv->final_duration;
490 priv->final_duration += os->pduration;
496 const struct ogg_codec ff_vorbis_codec = {
497 .magic = "\001vorbis",
499 .header = vorbis_header,
500 .packet = vorbis_packet,
501 .cleanup = vorbis_cleanup,