1 /*
2 * Copyright (C) 2005 Michael Ahlberg, Måns Rullgård
3 *
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:
11 *
12 * The above copyright notice and this permission notice shall be
13 * included in all copies or substantial portions of the Software.
14 *
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.
23 */
24
25 #include <stdlib.h>
26
27 #include "libavutil/avstring.h"
28 #include "libavutil/base64.h"
29 #include "libavutil/dict.h"
30
31 #include "libavcodec/bytestream.h"
32 #include "libavcodec/vorbis_parser.h"
33
34 #include "avformat.h"
35 #include "demux.h"
36 #include "flac_picture.h"
37 #include "internal.h"
38 #include "oggdec.h"
39 #include "vorbiscomment.h"
40 #include "replaygain.h"
41
ogm_chapter(AVFormatContext * as,const uint8_t * key,const uint8_t * val)42 static int ogm_chapter(AVFormatContext *as, const uint8_t *key, const uint8_t *val)
43 {
44 int i, cnum, h, m, s, ms, keylen = strlen(key);
45 AVChapter *chapter = NULL;
46
47 if (keylen < 9 || av_strncasecmp(key, "CHAPTER", 7) || sscanf(key+7, "%03d", &cnum) != 1)
48 return 0;
49
50 if (keylen <= 10) {
51 if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
52 return 0;
53
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];
61 break;
62 }
63 if (!chapter)
64 return 0;
65
66 av_dict_set(&chapter->metadata, "title", val, 0);
67 } else
68 return 0;
69
70 return 1;
71 }
72
ff_vorbis_stream_comment(AVFormatContext * as,AVStream * st,const uint8_t * buf,int size)73 int ff_vorbis_stream_comment(AVFormatContext *as, AVStream *st,
74 const uint8_t *buf, int size)
75 {
76 int updates = ff_vorbis_comment(as, &st->metadata, buf, size, 1);
77
78 if (updates > 0) {
79 st->event_flags |= AVSTREAM_EVENT_FLAG_METADATA_UPDATED;
80 }
81
82 return updates;
83 }
84
85 /**
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.
89 */
vorbis_parse_single_comment(AVFormatContext * as,AVDictionary ** m,const uint8_t * buf,uint32_t size,int * updates,int parse_picture)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)
93 {
94 char *t = (char*)buf, *v = memchr(t, '=', size);
95 int tl, vl;
96 char backup;
97
98 if (!v)
99 return 0;
100
101 tl = v - t;
102 vl = size - tl - 1;
103 v++;
104
105 if (!tl || !vl)
106 return 0;
107
108 t[tl] = 0;
109
110 backup = v[vl];
111 v[vl] = 0;
112
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."
118 */
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);
122
123 if (!pict) {
124 av_log(as, AV_LOG_WARNING, "out-of-memory error. Skipping cover art block.\n");
125 goto end;
126 }
127 ret = av_base64_decode(pict, v, len);
128 if (ret > 0)
129 ret = ff_flac_parse_picture(as, &pict, ret, 0);
130 av_freep(&pict);
131 if (ret < 0) {
132 av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
133 goto end;
134 }
135 } else if (!ogm_chapter(as, t, v)) {
136 (*updates)++;
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);
140 }
141 end:
142 t[tl] = '=';
143 v[vl] = backup;
144
145 return 0;
146 }
147
ff_vorbis_comment(AVFormatContext * as,AVDictionary ** m,const uint8_t * buf,int size,int parse_picture)148 int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
149 const uint8_t *buf, int size,
150 int parse_picture)
151 {
152 const uint8_t *p = buf;
153 const uint8_t *end = buf + size;
154 int updates = 0;
155 unsigned n;
156 int s, ret;
157
158 /* must have vendor_length and user_comment_list_length */
159 if (size < 8)
160 return AVERROR_INVALIDDATA;
161
162 s = bytestream_get_le32(&p);
163
164 if (end - p - 4 < s || s < 0)
165 return AVERROR_INVALIDDATA;
166
167 p += s;
168
169 n = bytestream_get_le32(&p);
170
171 while (end - p >= 4 && n > 0) {
172 s = bytestream_get_le32(&p);
173
174 if (end - p < s || s < 0)
175 break;
176
177 ret = vorbis_parse_single_comment(as, m, p, s, &updates, parse_picture);
178 if (ret < 0)
179 return ret;
180 p += s;
181 n--;
182 }
183
184 if (p != end)
185 av_log(as, AV_LOG_INFO,
186 "%"PTRDIFF_SPECIFIER" bytes of comment header remain\n", end - p);
187 if (n > 0)
188 av_log(as, AV_LOG_INFO,
189 "truncated comment header, %i comments not found\n", n);
190
191 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
192
193 return updates;
194 }
195
196 /*
197 * Parse the vorbis header
198 *
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
209 */
210
211 struct oggvorbis_private {
212 unsigned int len[3];
213 unsigned char *packet[3];
214 AVVorbisParseContext *vp;
215 int64_t final_pts;
216 int final_duration;
217 };
218
fixup_vorbis_headers(AVFormatContext * as,struct oggvorbis_private * priv,uint8_t ** buf)219 static int fixup_vorbis_headers(AVFormatContext *as,
220 struct oggvorbis_private *priv,
221 uint8_t **buf)
222 {
223 int i, offset, len, err;
224 int buf_len;
225 unsigned char *ptr;
226
227 len = priv->len[0] + priv->len[1] + priv->len[2];
228 buf_len = len + len / 255 + 64;
229
230 if (*buf)
231 return AVERROR_INVALIDDATA;
232
233 ptr = *buf = av_realloc(NULL, buf_len);
234 if (!ptr)
235 return AVERROR(ENOMEM);
236 memset(*buf, '\0', buf_len);
237
238 ptr[0] = 2;
239 offset = 1;
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]);
246 }
247 if ((err = av_reallocp(buf, offset + AV_INPUT_BUFFER_PADDING_SIZE)) < 0)
248 return err;
249 return offset;
250 }
251
vorbis_cleanup(AVFormatContext * s,int idx)252 static void vorbis_cleanup(AVFormatContext *s, int idx)
253 {
254 struct ogg *ogg = s->priv_data;
255 struct ogg_stream *os = ogg->streams + idx;
256 struct oggvorbis_private *priv = os->private;
257 int i;
258 if (os->private) {
259 av_vorbis_parse_free(&priv->vp);
260 for (i = 0; i < 3; i++)
261 av_freep(&priv->packet[i]);
262 }
263 }
264
vorbis_update_metadata(AVFormatContext * s,int idx)265 static int vorbis_update_metadata(AVFormatContext *s, int idx)
266 {
267 struct ogg *ogg = s->priv_data;
268 struct ogg_stream *os = ogg->streams + idx;
269 AVStream *st = s->streams[idx];
270 int ret;
271
272 if (os->psize <= 8)
273 return 0;
274
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,
278 os->psize - 8);
279 if (ret < 0)
280 return ret;
281
282 /* Update the metadata if possible. */
283 av_freep(&os->new_metadata);
284 if (st->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. */
287 } else {
288 os->new_metadata = av_mallocz(1);
289 os->new_metadata_size = 0;
290 }
291
292 return ret;
293 }
294
vorbis_header(AVFormatContext * s,int idx)295 static int vorbis_header(AVFormatContext *s, int idx)
296 {
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];
302
303 if (!os->private) {
304 os->private = av_mallocz(sizeof(struct oggvorbis_private));
305 if (!os->private)
306 return AVERROR(ENOMEM);
307 }
308
309 priv = os->private;
310
311 if (!(pkt_type & 1))
312 return priv->vp ? 0 : AVERROR_INVALIDDATA;
313
314 if (os->psize < 1 || pkt_type > 5)
315 return AVERROR_INVALIDDATA;
316
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 priv->vp ? 0 : AVERROR_INVALIDDATA;
321
322 priv->len[pkt_type >> 1] = os->psize;
323 priv->packet[pkt_type >> 1] = av_memdup(os->buf + os->pstart, os->psize);
324 if (!priv->packet[pkt_type >> 1])
325 return AVERROR(ENOMEM);
326 if (os->buf[os->pstart] == 1) {
327 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
328 unsigned blocksize, bs0, bs1;
329 int srate;
330 int channels;
331
332 if (os->psize != 30)
333 return AVERROR_INVALIDDATA;
334
335 if (bytestream_get_le32(&p) != 0) /* vorbis_version */
336 return AVERROR_INVALIDDATA;
337
338 channels = bytestream_get_byte(&p);
339 if (st->codecpar->ch_layout.nb_channels &&
340 channels != st->codecpar->ch_layout.nb_channels) {
341 av_log(s, AV_LOG_ERROR, "Channel change is not supported\n");
342 return AVERROR_PATCHWELCOME;
343 }
344 st->codecpar->ch_layout.nb_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
349
350 blocksize = bytestream_get_byte(&p);
351 bs0 = blocksize & 15;
352 bs1 = blocksize >> 4;
353
354 if (bs0 > bs1)
355 return AVERROR_INVALIDDATA;
356 if (bs0 < 6 || bs1 > 13)
357 return AVERROR_INVALIDDATA;
358
359 if (bytestream_get_byte(&p) != 1) /* framing_flag */
360 return AVERROR_INVALIDDATA;
361
362 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
363 st->codecpar->codec_id = AV_CODEC_ID_VORBIS;
364
365 if (srate > 0) {
366 st->codecpar->sample_rate = srate;
367 avpriv_set_pts_info(st, 64, 1, srate);
368 }
369 } else if (os->buf[os->pstart] == 3) {
370 if (vorbis_update_metadata(s, idx) >= 0 && priv->len[1] > 10) {
371 unsigned new_len;
372
373 int ret = ff_replaygain_export(st, st->metadata);
374 if (ret < 0)
375 return ret;
376
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;
383 }
384 }
385 } else {
386 int ret;
387
388 if (priv->vp)
389 return AVERROR_INVALIDDATA;
390
391 ret = fixup_vorbis_headers(s, priv, &st->codecpar->extradata);
392 if (ret < 0) {
393 st->codecpar->extradata_size = 0;
394 return ret;
395 }
396 st->codecpar->extradata_size = ret;
397
398 priv->vp = av_vorbis_parse_init(st->codecpar->extradata, st->codecpar->extradata_size);
399 if (!priv->vp) {
400 av_freep(&st->codecpar->extradata);
401 st->codecpar->extradata_size = 0;
402 return AVERROR_UNKNOWN;
403 }
404 }
405
406 return 1;
407 }
408
vorbis_packet(AVFormatContext * s,int idx)409 static int vorbis_packet(AVFormatContext *s, int idx)
410 {
411 struct ogg *ogg = s->priv_data;
412 struct ogg_stream *os = ogg->streams + idx;
413 struct oggvorbis_private *priv = os->private;
414 int duration, flags = 0;
415
416 if (!priv->vp)
417 return AVERROR_INVALIDDATA;
418
419 /* first packet handling
420 * here we parse the duration of each packet in the first page and compare
421 * the total duration to the page granule to find the encoder delay and
422 * set the first timestamp */
423 if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & OGG_FLAG_EOS) && (int64_t)os->granule>=0) {
424 int seg, d;
425 uint8_t *last_pkt = os->buf + os->pstart;
426 uint8_t *next_pkt = last_pkt;
427
428 av_vorbis_parse_reset(priv->vp);
429 duration = 0;
430 seg = os->segp;
431 d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
432 if (d < 0) {
433 os->pflags |= AV_PKT_FLAG_CORRUPT;
434 return 0;
435 } else if (flags & VORBIS_FLAG_COMMENT) {
436 vorbis_update_metadata(s, idx);
437 flags = 0;
438 }
439 duration += d;
440 last_pkt = next_pkt = next_pkt + os->psize;
441 for (; seg < os->nsegs; seg++) {
442 if (os->segments[seg] < 255) {
443 int d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
444 if (d < 0) {
445 duration = os->granule;
446 break;
447 } else if (flags & VORBIS_FLAG_COMMENT) {
448 vorbis_update_metadata(s, idx);
449 flags = 0;
450 }
451 duration += d;
452 last_pkt = next_pkt + os->segments[seg];
453 }
454 next_pkt += os->segments[seg];
455 }
456 os->lastpts =
457 os->lastdts = os->granule - duration;
458
459 if (!os->granule && duration) //hack to deal with broken files (Ticket3710)
460 os->lastpts = os->lastdts = AV_NOPTS_VALUE;
461
462 if (s->streams[idx]->start_time == AV_NOPTS_VALUE) {
463 s->streams[idx]->start_time = FFMAX(os->lastpts, 0);
464 if (s->streams[idx]->duration != AV_NOPTS_VALUE)
465 s->streams[idx]->duration -= s->streams[idx]->start_time;
466 }
467 priv->final_pts = AV_NOPTS_VALUE;
468 av_vorbis_parse_reset(priv->vp);
469 }
470
471 /* parse packet duration */
472 if (os->psize > 0) {
473 duration = av_vorbis_parse_frame_flags(priv->vp, os->buf + os->pstart, 1, &flags);
474 if (duration < 0) {
475 os->pflags |= AV_PKT_FLAG_CORRUPT;
476 return 0;
477 } else if (flags & VORBIS_FLAG_COMMENT) {
478 vorbis_update_metadata(s, idx);
479 flags = 0;
480 }
481 os->pduration = duration;
482 }
483
484 /* final packet handling
485 * here we save the pts of the first packet in the final page, sum up all
486 * packet durations in the final page except for the last one, and compare
487 * to the page granule to find the duration of the final packet */
488 if (os->flags & OGG_FLAG_EOS) {
489 if (os->lastpts != AV_NOPTS_VALUE) {
490 priv->final_pts = os->lastpts;
491 priv->final_duration = 0;
492 }
493 if (os->segp == os->nsegs) {
494 int64_t skip = priv->final_pts + priv->final_duration + os->pduration - os->granule;
495 if (skip > 0)
496 os->end_trimming = skip;
497 os->pduration = os->granule - priv->final_pts - priv->final_duration;
498 }
499 priv->final_duration += os->pduration;
500 }
501
502 return 0;
503 }
504
505 const struct ogg_codec ff_vorbis_codec = {
506 .magic = "\001vorbis",
507 .magicsize = 7,
508 .header = vorbis_header,
509 .packet = vorbis_packet,
510 .cleanup = vorbis_cleanup,
511 .nb_header = 3,
512 };
513