c19b54552c9a5b3688ce601aa74de20fd9cbf6ae
[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
40 static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
41 {
42 int i, cnum, h, m, s, ms, keylen = strlen(key);
43 AVChapter *chapter = NULL;
44
45 if (keylen < 9 || sscanf(key, "CHAPTER%03d", &cnum) != 1)
46 return 0;
47
48 if (keylen <= 10) {
49 if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
50 return 0;
51
52 avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 },
53 ms + 1000 * (s + 60 * (m + 60 * h)),
54 AV_NOPTS_VALUE, NULL);
55 av_free(val);
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];
60 break;
61 }
62 if (!chapter)
63 return 0;
64
65 av_dict_set(&chapter->metadata, "title", val, AV_DICT_DONT_STRDUP_VAL);
66 } else
67 return 0;
68
69 av_free(key);
70 return 1;
71 }
72
73 int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
74 const uint8_t *buf, int size)
75 {
76 const uint8_t *p = buf;
77 const uint8_t *end = buf + size;
78 unsigned n, j;
79 int s;
80
81 /* must have vendor_length and user_comment_list_length */
82 if (size < 8)
83 return AVERROR_INVALIDDATA;
84
85 s = bytestream_get_le32(&p);
86
87 if (end - p - 4 < s || s < 0)
88 return AVERROR_INVALIDDATA;
89
90 p += s;
91
92 n = bytestream_get_le32(&p);
93
94 while (end - p >= 4 && n > 0) {
95 const char *t, *v;
96 int tl, vl;
97
98 s = bytestream_get_le32(&p);
99
100 if (end - p < s || s < 0)
101 break;
102
103 t = p;
104 p += s;
105 n--;
106
107 v = memchr(t, '=', s);
108 if (!v)
109 continue;
110
111 tl = v - t;
112 vl = s - tl - 1;
113 v++;
114
115 if (tl && vl) {
116 char *tt, *ct;
117
118 tt = av_malloc(tl + 1);
119 ct = av_malloc(vl + 1);
120 if (!tt || !ct) {
121 av_freep(&tt);
122 av_freep(&ct);
123 return AVERROR(ENOMEM);
124 }
125
126 for (j = 0; j < tl; j++)
127 tt[j] = av_toupper(t[j]);
128 tt[tl] = 0;
129
130 memcpy(ct, v, vl);
131 ct[vl] = 0;
132
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."
138 */
139 if (!strcmp(tt, "METADATA_BLOCK_PICTURE")) {
140 int ret;
141 char *pict = av_malloc(vl);
142
143 if (!pict) {
144 av_freep(&tt);
145 av_freep(&ct);
146 return AVERROR(ENOMEM);
147 }
148 if ((ret = av_base64_decode(pict, ct, vl)) > 0)
149 ret = ff_flac_parse_picture(as, pict, ret);
150 av_freep(&tt);
151 av_freep(&ct);
152 av_freep(&pict);
153 if (ret < 0) {
154 av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
155 continue;
156 }
157 } else if (!ogm_chapter(as, tt, ct))
158 av_dict_set(m, tt, ct,
159 AV_DICT_DONT_STRDUP_KEY |
160 AV_DICT_DONT_STRDUP_VAL);
161 }
162 }
163
164 if (p != end)
165 av_log(as, AV_LOG_INFO,
166 "%ti bytes of comment header remain\n", end - p);
167 if (n > 0)
168 av_log(as, AV_LOG_INFO,
169 "truncated comment header, %i comments not found\n", n);
170
171 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
172
173 return 0;
174 }
175
176 /*
177 * Parse the vorbis header
178 *
179 * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
180 * [vorbis_version] = read 32 bits as unsigned integer | Not used
181 * [audio_channels] = read 8 bit integer as unsigned | Used
182 * [audio_sample_rate] = read 32 bits as unsigned integer | Used
183 * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
184 * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
185 * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
186 * [blocksize_0] = read 4 bits as unsigned integer | Not Used
187 * [blocksize_1] = read 4 bits as unsigned integer | Not Used
188 * [framing_flag] = read one bit | Not Used
189 */
190
191 struct oggvorbis_private {
192 unsigned int len[3];
193 unsigned char *packet[3];
194 VorbisParseContext vp;
195 int64_t final_pts;
196 int final_duration;
197 };
198
199 static int fixup_vorbis_headers(AVFormatContext *as,
200 struct oggvorbis_private *priv,
201 uint8_t **buf)
202 {
203 int i, offset, len, err;
204 unsigned char *ptr;
205
206 len = priv->len[0] + priv->len[1] + priv->len[2];
207 ptr = *buf = av_mallocz(len + len / 255 + 64);
208 if (!ptr)
209 return AVERROR(ENOMEM);
210
211 ptr[0] = 2;
212 offset = 1;
213 offset += av_xiphlacing(&ptr[offset], priv->len[0]);
214 offset += av_xiphlacing(&ptr[offset], priv->len[1]);
215 for (i = 0; i < 3; i++) {
216 memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
217 offset += priv->len[i];
218 av_freep(&priv->packet[i]);
219 }
220 if ((err = av_reallocp(buf, offset + FF_INPUT_BUFFER_PADDING_SIZE)) < 0)
221 return err;
222 return offset;
223 }
224
225 static void vorbis_cleanup(AVFormatContext *s, int idx)
226 {
227 struct ogg *ogg = s->priv_data;
228 struct ogg_stream *os = ogg->streams + idx;
229 struct oggvorbis_private *priv = os->private;
230 int i;
231 if (os->private)
232 for (i = 0; i < 3; i++)
233 av_freep(&priv->packet[i]);
234 }
235
236 static int vorbis_header(AVFormatContext *s, int idx)
237 {
238 struct ogg *ogg = s->priv_data;
239 AVStream *st = s->streams[idx];
240 struct ogg_stream *os = ogg->streams + idx;
241 struct oggvorbis_private *priv;
242 int pkt_type = os->buf[os->pstart];
243
244 if (!os->private) {
245 os->private = av_mallocz(sizeof(struct oggvorbis_private));
246 if (!os->private)
247 return AVERROR(ENOMEM);
248 }
249
250 if (!(pkt_type & 1))
251 return 0;
252
253 if (os->psize < 1 || pkt_type > 5)
254 return AVERROR_INVALIDDATA;
255
256 priv = os->private;
257
258 if (priv->packet[pkt_type >> 1])
259 return AVERROR_INVALIDDATA;
260 if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
261 return AVERROR_INVALIDDATA;
262
263 priv->len[pkt_type >> 1] = os->psize;
264 priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
265 if (!priv->packet[pkt_type >> 1])
266 return AVERROR(ENOMEM);
267 memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
268 if (os->buf[os->pstart] == 1) {
269 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
270 unsigned blocksize, bs0, bs1;
271 int srate;
272
273 if (os->psize != 30)
274 return AVERROR_INVALIDDATA;
275
276 if (bytestream_get_le32(&p) != 0) /* vorbis_version */
277 return AVERROR_INVALIDDATA;
278
279 st->codec->channels = bytestream_get_byte(&p);
280 srate = bytestream_get_le32(&p);
281 p += 4; // skip maximum bitrate
282 st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
283 p += 4; // skip minimum bitrate
284
285 blocksize = bytestream_get_byte(&p);
286 bs0 = blocksize & 15;
287 bs1 = blocksize >> 4;
288
289 if (bs0 > bs1)
290 return AVERROR_INVALIDDATA;
291 if (bs0 < 6 || bs1 > 13)
292 return AVERROR_INVALIDDATA;
293
294 if (bytestream_get_byte(&p) != 1) /* framing_flag */
295 return AVERROR_INVALIDDATA;
296
297 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
298 st->codec->codec_id = AV_CODEC_ID_VORBIS;
299
300 if (srate > 0) {
301 st->codec->sample_rate = srate;
302 avpriv_set_pts_info(st, 64, 1, srate);
303 }
304 } else if (os->buf[os->pstart] == 3) {
305 if (os->psize > 8 &&
306 ff_vorbis_comment(s, &st->metadata, os->buf + os->pstart + 7,
307 os->psize - 8) >= 0) {
308 // drop all metadata we parsed and which is not required by libvorbis
309 unsigned new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
310 if (new_len >= 16 && new_len < os->psize) {
311 AV_WL32(priv->packet[1] + new_len - 5, 0);
312 priv->packet[1][new_len - 1] = 1;
313 priv->len[1] = new_len;
314 }
315 }
316 } else {
317 int ret = fixup_vorbis_headers(s, priv, &st->codec->extradata);
318 if (ret < 0) {
319 st->codec->extradata_size = 0;
320 return ret;
321 }
322 st->codec->extradata_size = ret;
323 if ((ret = avpriv_vorbis_parse_extradata(st->codec, &priv->vp))) {
324 av_freep(&st->codec->extradata);
325 st->codec->extradata_size = 0;
326 return ret;
327 }
328 }
329
330 return 1;
331 }
332
333 static int vorbis_packet(AVFormatContext *s, int idx)
334 {
335 struct ogg *ogg = s->priv_data;
336 struct ogg_stream *os = ogg->streams + idx;
337 struct oggvorbis_private *priv = os->private;
338 int duration;
339
340 /* first packet handling
341 * here we parse the duration of each packet in the first page and compare
342 * the total duration to the page granule to find the encoder delay and
343 * set the first timestamp */
344 if (!os->lastpts) {
345 int seg;
346 uint8_t *last_pkt = os->buf + os->pstart;
347 uint8_t *next_pkt = last_pkt;
348 int first_duration = 0;
349
350 avpriv_vorbis_parse_reset(&priv->vp);
351 duration = 0;
352 for (seg = 0; seg < os->nsegs; seg++) {
353 if (os->segments[seg] < 255) {
354 int d = avpriv_vorbis_parse_frame(&priv->vp, last_pkt, 1);
355 if (d < 0) {
356 duration = os->granule;
357 break;
358 }
359 if (!duration)
360 first_duration = d;
361 duration += d;
362 last_pkt = next_pkt + os->segments[seg];
363 }
364 next_pkt += os->segments[seg];
365 }
366 os->lastpts =
367 os->lastdts = os->granule - duration;
368 s->streams[idx]->start_time = os->lastpts + first_duration;
369 if (s->streams[idx]->duration)
370 s->streams[idx]->duration -= s->streams[idx]->start_time;
371 s->streams[idx]->cur_dts = AV_NOPTS_VALUE;
372 priv->final_pts = AV_NOPTS_VALUE;
373 avpriv_vorbis_parse_reset(&priv->vp);
374 }
375
376 /* parse packet duration */
377 if (os->psize > 0) {
378 duration = avpriv_vorbis_parse_frame(&priv->vp, os->buf + os->pstart, 1);
379 if (duration <= 0) {
380 os->pflags |= AV_PKT_FLAG_CORRUPT;
381 return 0;
382 }
383 os->pduration = duration;
384 }
385
386 /* final packet handling
387 * here we save the pts of the first packet in the final page, sum up all
388 * packet durations in the final page except for the last one, and compare
389 * to the page granule to find the duration of the final packet */
390 if (os->flags & OGG_FLAG_EOS) {
391 if (os->lastpts != AV_NOPTS_VALUE) {
392 priv->final_pts = os->lastpts;
393 priv->final_duration = 0;
394 }
395 if (os->segp == os->nsegs)
396 os->pduration = os->granule - priv->final_pts - priv->final_duration;
397 priv->final_duration += os->pduration;
398 }
399
400 return 0;
401 }
402
403 const struct ogg_codec ff_vorbis_codec = {
404 .magic = "\001vorbis",
405 .magicsize = 7,
406 .header = vorbis_header,
407 .packet = vorbis_packet,
408 .cleanup = vorbis_cleanup,
409 .nb_header = 3,
410 };