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/dict.h"
31 #include "libavcodec/bytestream.h"
32 #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, const uint8_t *key, const uint8_t *val)
44 int i, cnum, h, m, s, ms, keylen = strlen(key);
45 AVChapter *chapter = NULL;
47 if (keylen < 9 || av_strncasecmp(key, "CHAPTER", 7) || sscanf(key+7, "%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);
57 } else if (!av_strcasecmp(key + keylen - 4, "NAME")) {
58 for (i = 0; i < as->nb_chapters; i++)
59 if (as->chapters[i]->id == cnum) {
60 chapter = as->chapters[i];
66 av_dict_set(&chapter->metadata, "title", val, 0);
73 int ff_vorbis_stream_comment(AVFormatContext *as, AVStream *st,
74 const uint8_t *buf, int size)
76 int updates = ff_vorbis_comment(as, &st->metadata, buf, size, 1);
79 st->event_flags |= AVSTREAM_EVENT_FLAG_METADATA_UPDATED;
86 * This function temporarily modifies the (const qualified) input buffer
87 * and reverts its changes before return. The input buffer needs to have
88 * at least one byte of padding.
90 static int vorbis_parse_single_comment(AVFormatContext *as, AVDictionary **m,
91 const uint8_t *buf, uint32_t size,
92 int *updates, int parse_picture)
94 char *t = (char*)buf, *v = memchr(t, '=', size);
113 /* The format in which the pictures are stored is the FLAC format.
114 * Xiph says: "The binary FLAC picture structure is base64 encoded
115 * and placed within a VorbisComment with the tag name
116 * 'METADATA_BLOCK_PICTURE'. This is the preferred and
117 * recommended way of embedding cover art within VorbisComments."
119 if (!av_strcasecmp(t, "METADATA_BLOCK_PICTURE") && parse_picture) {
120 int ret, len = AV_BASE64_DECODE_SIZE(vl);
121 uint8_t *pict = av_malloc(len + AV_INPUT_BUFFER_PADDING_SIZE);
124 av_log(as, AV_LOG_WARNING, "out-of-memory error. Skipping cover art block.\n");
127 ret = av_base64_decode(pict, v, len);
129 ret = ff_flac_parse_picture(as, &pict, ret, 0);
132 av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
135 } else if (!ogm_chapter(as, t, v)) {
137 if (av_dict_get(*m, t, NULL, 0))
138 av_dict_set(m, t, ";", AV_DICT_APPEND);
139 av_dict_set(m, t, v, AV_DICT_APPEND);
148 int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
149 const uint8_t *buf, int size,
152 const uint8_t *p = buf;
153 const uint8_t *end = buf + size;
158 /* must have vendor_length and user_comment_list_length */
160 return AVERROR_INVALIDDATA;
162 s = bytestream_get_le32(&p);
164 if (end - p - 4 < s || s < 0)
165 return AVERROR_INVALIDDATA;
169 n = bytestream_get_le32(&p);
171 while (end - p >= 4 && n > 0) {
172 s = bytestream_get_le32(&p);
174 if (end - p < s || s < 0)
177 ret = vorbis_parse_single_comment(as, m, p, s, &updates, parse_picture);
185 av_log(as, AV_LOG_INFO,
186 "%"PTRDIFF_SPECIFIER" bytes of comment header remain\n", end - p);
188 av_log(as, AV_LOG_INFO,
189 "truncated comment header, %i comments not found\n", n);
191 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
197 * Parse the vorbis header
199 * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
200 * [vorbis_version] = read 32 bits as unsigned integer | Not used
201 * [audio_channels] = read 8 bit integer as unsigned | Used
202 * [audio_sample_rate] = read 32 bits as unsigned integer | Used
203 * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
204 * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
205 * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
206 * [blocksize_0] = read 4 bits as unsigned integer | Not Used
207 * [blocksize_1] = read 4 bits as unsigned integer | Not Used
208 * [framing_flag] = read one bit | Not Used
211 struct oggvorbis_private {
213 unsigned char *packet[3];
214 AVVorbisParseContext *vp;
219 static int fixup_vorbis_headers(AVFormatContext *as,
220 struct oggvorbis_private *priv,
223 int i, offset, len, err;
227 len = priv->len[0] + priv->len[1] + priv->len[2];
228 buf_len = len + len / 255 + 64;
231 return AVERROR_INVALIDDATA;
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_mallocz(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;
315 av_log(s, AV_LOG_VERBOSE, "Ignoring packet with unknown type %d\n", pkt_type);
320 return AVERROR_INVALIDDATA;
322 if (priv->packet[pkt_type >> 1])
323 return AVERROR_INVALIDDATA;
324 if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
325 return priv->vp ? 0 : AVERROR_INVALIDDATA;
327 priv->len[pkt_type >> 1] = os->psize;
328 priv->packet[pkt_type >> 1] = av_memdup(os->buf + os->pstart, os->psize);
329 if (!priv->packet[pkt_type >> 1])
330 return AVERROR(ENOMEM);
331 if (os->buf[os->pstart] == 1) {
332 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
333 unsigned blocksize, bs0, bs1;
338 return AVERROR_INVALIDDATA;
340 if (bytestream_get_le32(&p) != 0) /* vorbis_version */
341 return AVERROR_INVALIDDATA;
343 channels = bytestream_get_byte(&p);
344 if (st->codecpar->ch_layout.nb_channels &&
345 channels != st->codecpar->ch_layout.nb_channels) {
346 av_log(s, AV_LOG_ERROR, "Channel change is not supported\n");
347 return AVERROR_PATCHWELCOME;
349 st->codecpar->ch_layout.nb_channels = channels;
350 srate = bytestream_get_le32(&p);
351 p += 4; // skip maximum bitrate
352 st->codecpar->bit_rate = bytestream_get_le32(&p); // nominal bitrate
353 p += 4; // skip minimum bitrate
355 blocksize = bytestream_get_byte(&p);
356 bs0 = blocksize & 15;
357 bs1 = blocksize >> 4;
360 return AVERROR_INVALIDDATA;
361 if (bs0 < 6 || bs1 > 13)
362 return AVERROR_INVALIDDATA;
364 if (bytestream_get_byte(&p) != 1) /* framing_flag */
365 return AVERROR_INVALIDDATA;
367 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
368 st->codecpar->codec_id = AV_CODEC_ID_VORBIS;
371 st->codecpar->sample_rate = srate;
372 avpriv_set_pts_info(st, 64, 1, srate);
374 } else if (os->buf[os->pstart] == 3) {
375 if (vorbis_update_metadata(s, idx) >= 0 && priv->len[1] > 10) {
378 int ret = ff_replaygain_export(st, st->metadata);
382 // drop all metadata we parsed and which is not required by libvorbis
383 new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
384 if (new_len >= 16 && new_len < os->psize) {
385 AV_WL32(priv->packet[1] + new_len - 5, 0);
386 priv->packet[1][new_len - 1] = 1;
387 priv->len[1] = new_len;
394 return AVERROR_INVALIDDATA;
396 ret = fixup_vorbis_headers(s, priv, &st->codecpar->extradata);
398 st->codecpar->extradata_size = 0;
401 st->codecpar->extradata_size = ret;
403 priv->vp = av_vorbis_parse_init(st->codecpar->extradata, st->codecpar->extradata_size);
405 av_freep(&st->codecpar->extradata);
406 st->codecpar->extradata_size = 0;
407 return AVERROR_UNKNOWN;
414 static int vorbis_packet(AVFormatContext *s, int idx)
416 struct ogg *ogg = s->priv_data;
417 struct ogg_stream *os = ogg->streams + idx;
418 struct oggvorbis_private *priv = os->private;
419 int duration, flags = 0;
422 return AVERROR_INVALIDDATA;
424 /* first packet handling
425 * here we parse the duration of each packet in the first page and compare
426 * the total duration to the page granule to find the encoder delay and
427 * set the first timestamp */
428 if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & OGG_FLAG_EOS) && (int64_t)os->granule>=0) {
430 uint8_t *last_pkt = os->buf + os->pstart;
431 uint8_t *next_pkt = last_pkt;
433 av_vorbis_parse_reset(priv->vp);
436 d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
438 os->pflags |= AV_PKT_FLAG_CORRUPT;
440 } else if (flags & VORBIS_FLAG_COMMENT) {
441 vorbis_update_metadata(s, idx);
445 last_pkt = next_pkt = next_pkt + os->psize;
446 for (; seg < os->nsegs; seg++) {
447 if (os->segments[seg] < 255) {
448 int d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
450 duration = os->granule;
452 } else if (flags & VORBIS_FLAG_COMMENT) {
453 vorbis_update_metadata(s, idx);
457 last_pkt = next_pkt + os->segments[seg];
459 next_pkt += os->segments[seg];
462 os->lastdts = os->granule - duration;
464 if (!os->granule && duration) //hack to deal with broken files (Ticket3710)
465 os->lastpts = os->lastdts = AV_NOPTS_VALUE;
467 if (s->streams[idx]->start_time == AV_NOPTS_VALUE) {
468 s->streams[idx]->start_time = FFMAX(os->lastpts, 0);
469 if (s->streams[idx]->duration != AV_NOPTS_VALUE)
470 s->streams[idx]->duration -= s->streams[idx]->start_time;
472 priv->final_pts = AV_NOPTS_VALUE;
473 av_vorbis_parse_reset(priv->vp);
476 /* parse packet duration */
478 duration = av_vorbis_parse_frame_flags(priv->vp, os->buf + os->pstart, 1, &flags);
480 os->pflags |= AV_PKT_FLAG_CORRUPT;
482 } else if (flags & VORBIS_FLAG_COMMENT) {
483 vorbis_update_metadata(s, idx);
486 os->pduration = duration;
489 /* final packet handling
490 * here we save the pts of the first packet in the final page, sum up all
491 * packet durations in the final page except for the last one, and compare
492 * to the page granule to find the duration of the final packet */
493 if (os->flags & OGG_FLAG_EOS) {
494 if (os->lastpts != AV_NOPTS_VALUE) {
495 priv->final_pts = os->lastpts;
496 priv->final_duration = 0;
498 if (os->segp == os->nsegs) {
499 int64_t skip = priv->final_pts + priv->final_duration + os->pduration - os->granule;
501 os->end_trimming = skip;
502 os->pduration = os->granule - priv->final_pts - priv->final_duration;
504 priv->final_duration += os->pduration;
510 const struct ogg_codec ff_vorbis_codec = {
511 .magic = "\001vorbis",
513 .header = vorbis_header,
514 .packet = vorbis_packet,
515 .cleanup = vorbis_cleanup,