aac0f0e98074d49b1d0e10e4496cb9f3524a0ef2
[libav.git] / libavformat / oggparsevorbis.c
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/bswap.h"
30 #include "libavutil/dict.h"
31 #include "libavcodec/bytestream.h"
32 #include "libavcodec/get_bits.h"
33 #include "libavcodec/vorbis_parser.h"
34 #include "avformat.h"
35 #include "flac_picture.h"
36 #include "internal.h"
37 #include "oggdec.h"
38 #include "vorbiscomment.h"
39 #include "replaygain.h"
40
41 static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
42 {
43 int i, cnum, h, m, s, ms, keylen = strlen(key);
44 AVChapter *chapter = NULL;
45
46 if (keylen < 9 || sscanf(key, "CHAPTER%03d", &cnum) != 1)
47 return 0;
48
49 if (keylen <= 10) {
50 if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
51 return 0;
52
53 avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 },
54 ms + 1000 * (s + 60 * (m + 60 * h)),
55 AV_NOPTS_VALUE, NULL);
56 av_free(val);
57 } else if (!strcmp(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, AV_DICT_DONT_STRDUP_VAL);
67 } else
68 return 0;
69
70 av_free(key);
71 return 1;
72 }
73
74 int ff_vorbis_stream_comment(AVFormatContext *as, AVStream *st,
75 const uint8_t *buf, int size)
76 {
77 int updates = ff_vorbis_comment(as, &st->metadata, buf, size, 1);
78
79 if (updates > 0) {
80 st->event_flags |= AVSTREAM_EVENT_FLAG_METADATA_UPDATED;
81 }
82
83 return updates;
84 }
85
86 int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
87 const uint8_t *buf, int size,
88 int parse_picture)
89 {
90 const uint8_t *p = buf;
91 const uint8_t *end = buf + size;
92 int updates = 0;
93 unsigned n, j;
94 int s;
95
96 /* must have vendor_length and user_comment_list_length */
97 if (size < 8)
98 return AVERROR_INVALIDDATA;
99
100 s = bytestream_get_le32(&p);
101
102 if (end - p - 4 < s || s < 0)
103 return AVERROR_INVALIDDATA;
104
105 p += s;
106
107 n = bytestream_get_le32(&p);
108
109 while (end - p >= 4 && n > 0) {
110 const char *t, *v;
111 int tl, vl;
112
113 s = bytestream_get_le32(&p);
114
115 if (end - p < s || s < 0)
116 break;
117
118 t = p;
119 p += s;
120 n--;
121
122 v = memchr(t, '=', s);
123 if (!v)
124 continue;
125
126 tl = v - t;
127 vl = s - tl - 1;
128 v++;
129
130 if (tl && vl) {
131 char *tt, *ct;
132
133 tt = av_malloc(tl + 1);
134 ct = av_malloc(vl + 1);
135 if (!tt || !ct) {
136 av_freep(&tt);
137 av_freep(&ct);
138 return AVERROR(ENOMEM);
139 }
140
141 for (j = 0; j < tl; j++)
142 tt[j] = av_toupper(t[j]);
143 tt[tl] = 0;
144
145 memcpy(ct, v, vl);
146 ct[vl] = 0;
147
148 /* The format in which the pictures are stored is the FLAC format.
149 * Xiph says: "The binary FLAC picture structure is base64 encoded
150 * and placed within a VorbisComment with the tag name
151 * 'METADATA_BLOCK_PICTURE'. This is the preferred and
152 * recommended way of embedding cover art within VorbisComments."
153 */
154 if (!strcmp(tt, "METADATA_BLOCK_PICTURE") && parse_picture) {
155 int ret;
156 char *pict = av_malloc(vl);
157
158 if (!pict) {
159 av_freep(&tt);
160 av_freep(&ct);
161 return AVERROR(ENOMEM);
162 }
163 if ((ret = av_base64_decode(pict, ct, vl)) > 0)
164 ret = ff_flac_parse_picture(as, pict, ret);
165 av_freep(&tt);
166 av_freep(&ct);
167 av_freep(&pict);
168 if (ret < 0) {
169 av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
170 continue;
171 }
172 } else if (!ogm_chapter(as, tt, ct)) {
173 updates++;
174 av_dict_set(m, tt, ct,
175 AV_DICT_DONT_STRDUP_KEY |
176 AV_DICT_DONT_STRDUP_VAL);
177 }
178 }
179 }
180
181 if (p != end)
182 av_log(as, AV_LOG_INFO,
183 "%ti bytes of comment header remain\n", end - p);
184 if (n > 0)
185 av_log(as, AV_LOG_INFO,
186 "truncated comment header, %i comments not found\n", n);
187
188 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
189
190 return updates;
191 }
192
193 /*
194 * Parse the vorbis header
195 *
196 * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
197 * [vorbis_version] = read 32 bits as unsigned integer | Not used
198 * [audio_channels] = read 8 bit integer as unsigned | Used
199 * [audio_sample_rate] = read 32 bits as unsigned integer | Used
200 * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
201 * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
202 * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
203 * [blocksize_0] = read 4 bits as unsigned integer | Not Used
204 * [blocksize_1] = read 4 bits as unsigned integer | Not Used
205 * [framing_flag] = read one bit | Not Used
206 */
207
208 struct oggvorbis_private {
209 unsigned int len[3];
210 unsigned char *packet[3];
211 AVVorbisParseContext *vp;
212 int64_t final_pts;
213 int final_duration;
214 };
215
216 static int fixup_vorbis_headers(AVFormatContext *as,
217 struct oggvorbis_private *priv,
218 uint8_t **buf)
219 {
220 int i, offset, len, err;
221 unsigned char *ptr;
222
223 len = priv->len[0] + priv->len[1] + priv->len[2];
224 ptr = *buf = av_mallocz(len + len / 255 + 64);
225 if (!ptr)
226 return AVERROR(ENOMEM);
227
228 ptr[0] = 2;
229 offset = 1;
230 offset += av_xiphlacing(&ptr[offset], priv->len[0]);
231 offset += av_xiphlacing(&ptr[offset], priv->len[1]);
232 for (i = 0; i < 3; i++) {
233 memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
234 offset += priv->len[i];
235 av_freep(&priv->packet[i]);
236 }
237 if ((err = av_reallocp(buf, offset + AV_INPUT_BUFFER_PADDING_SIZE)) < 0)
238 return err;
239 return offset;
240 }
241
242 static void vorbis_cleanup(AVFormatContext *s, int idx)
243 {
244 struct ogg *ogg = s->priv_data;
245 struct ogg_stream *os = ogg->streams + idx;
246 struct oggvorbis_private *priv = os->private;
247 int i;
248 if (os->private) {
249 av_vorbis_parse_free(&priv->vp);
250 for (i = 0; i < 3; i++)
251 av_freep(&priv->packet[i]);
252 }
253 }
254
255 static int vorbis_header(AVFormatContext *s, int idx)
256 {
257 struct ogg *ogg = s->priv_data;
258 AVStream *st = s->streams[idx];
259 struct ogg_stream *os = ogg->streams + idx;
260 struct oggvorbis_private *priv;
261 int pkt_type = os->buf[os->pstart];
262
263 if (!os->private) {
264 os->private = av_mallocz(sizeof(struct oggvorbis_private));
265 if (!os->private)
266 return AVERROR(ENOMEM);
267 }
268
269 if (!(pkt_type & 1))
270 return 0;
271
272 if (os->psize < 1 || pkt_type > 5)
273 return AVERROR_INVALIDDATA;
274
275 priv = os->private;
276
277 if (priv->packet[pkt_type >> 1])
278 return AVERROR_INVALIDDATA;
279 if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
280 return AVERROR_INVALIDDATA;
281
282 priv->len[pkt_type >> 1] = os->psize;
283 priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
284 if (!priv->packet[pkt_type >> 1])
285 return AVERROR(ENOMEM);
286 memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
287 if (os->buf[os->pstart] == 1) {
288 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
289 unsigned blocksize, bs0, bs1;
290 int srate;
291
292 if (os->psize != 30)
293 return AVERROR_INVALIDDATA;
294
295 if (bytestream_get_le32(&p) != 0) /* vorbis_version */
296 return AVERROR_INVALIDDATA;
297
298 st->codec->channels = bytestream_get_byte(&p);
299 srate = bytestream_get_le32(&p);
300 p += 4; // skip maximum bitrate
301 st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
302 p += 4; // skip minimum bitrate
303
304 blocksize = bytestream_get_byte(&p);
305 bs0 = blocksize & 15;
306 bs1 = blocksize >> 4;
307
308 if (bs0 > bs1)
309 return AVERROR_INVALIDDATA;
310 if (bs0 < 6 || bs1 > 13)
311 return AVERROR_INVALIDDATA;
312
313 if (bytestream_get_byte(&p) != 1) /* framing_flag */
314 return AVERROR_INVALIDDATA;
315
316 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
317 st->codec->codec_id = AV_CODEC_ID_VORBIS;
318
319 if (srate > 0) {
320 st->codec->sample_rate = srate;
321 avpriv_set_pts_info(st, 64, 1, srate);
322 }
323 } else if (os->buf[os->pstart] == 3) {
324 if (os->psize > 8 &&
325 ff_vorbis_stream_comment(s, st, os->buf + os->pstart + 7,
326 os->psize - 8) >= 0) {
327 unsigned new_len;
328
329 int ret = ff_replaygain_export(st, st->metadata);
330 if (ret < 0)
331 return ret;
332
333 // drop all metadata we parsed and which is not required by libvorbis
334 new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
335 if (new_len >= 16 && new_len < os->psize) {
336 AV_WL32(priv->packet[1] + new_len - 5, 0);
337 priv->packet[1][new_len - 1] = 1;
338 priv->len[1] = new_len;
339 }
340 }
341 } else {
342 int ret = fixup_vorbis_headers(s, priv, &st->codec->extradata);
343 if (ret < 0) {
344 st->codec->extradata_size = 0;
345 return ret;
346 }
347 st->codec->extradata_size = ret;
348
349 priv->vp = av_vorbis_parse_init(st->codec->extradata, st->codec->extradata_size);
350 if (!priv->vp) {
351 av_freep(&st->codec->extradata);
352 st->codec->extradata_size = 0;
353 return ret;
354 }
355 }
356
357 return 1;
358 }
359
360 static int vorbis_packet(AVFormatContext *s, int idx)
361 {
362 struct ogg *ogg = s->priv_data;
363 struct ogg_stream *os = ogg->streams + idx;
364 struct oggvorbis_private *priv = os->private;
365 int duration;
366
367 /* first packet handling
368 * here we parse the duration of each packet in the first page and compare
369 * the total duration to the page granule to find the encoder delay and
370 * set the first timestamp */
371 if (!os->lastpts) {
372 int seg;
373 uint8_t *last_pkt = os->buf + os->pstart;
374 uint8_t *next_pkt = last_pkt;
375 int first_duration = 0;
376
377 av_vorbis_parse_reset(priv->vp);
378 duration = 0;
379 for (seg = 0; seg < os->nsegs; seg++) {
380 if (os->segments[seg] < 255) {
381 int d = av_vorbis_parse_frame(priv->vp, last_pkt, 1);
382 if (d < 0) {
383 duration = os->granule;
384 break;
385 }
386 if (!duration)
387 first_duration = d;
388 duration += d;
389 last_pkt = next_pkt + os->segments[seg];
390 }
391 next_pkt += os->segments[seg];
392 }
393 os->lastpts =
394 os->lastdts = os->granule - duration;
395 s->streams[idx]->start_time = os->lastpts + first_duration;
396 if (s->streams[idx]->duration)
397 s->streams[idx]->duration -= s->streams[idx]->start_time;
398 s->streams[idx]->cur_dts = AV_NOPTS_VALUE;
399 priv->final_pts = AV_NOPTS_VALUE;
400 av_vorbis_parse_reset(priv->vp);
401 }
402
403 /* parse packet duration */
404 if (os->psize > 0) {
405 duration = av_vorbis_parse_frame(priv->vp, os->buf + os->pstart, 1);
406 if (duration <= 0) {
407 os->pflags |= AV_PKT_FLAG_CORRUPT;
408 return 0;
409 }
410 os->pduration = duration;
411 }
412
413 /* final packet handling
414 * here we save the pts of the first packet in the final page, sum up all
415 * packet durations in the final page except for the last one, and compare
416 * to the page granule to find the duration of the final packet */
417 if (os->flags & OGG_FLAG_EOS) {
418 if (os->lastpts != AV_NOPTS_VALUE) {
419 priv->final_pts = os->lastpts;
420 priv->final_duration = 0;
421 }
422 if (os->segp == os->nsegs)
423 os->pduration = os->granule - priv->final_pts - priv->final_duration;
424 priv->final_duration += os->pduration;
425 }
426
427 return 0;
428 }
429
430 const struct ogg_codec ff_vorbis_codec = {
431 .magic = "\001vorbis",
432 .magicsize = 7,
433 .header = vorbis_header,
434 .packet = vorbis_packet,
435 .cleanup = vorbis_cleanup,
436 .nb_header = 3,
437 };