avformat/output-example: Declare link dependency on libswscale in the Makefile
[libav.git] / libavformat / oggparsevorbis.c
CommitLineData
9146ca37
MR
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>
245976da
DB
26#include "libavutil/avstring.h"
27#include "libavutil/bswap.h"
d2d67e42 28#include "libavutil/dict.h"
9106a698 29#include "libavcodec/get_bits.h"
245976da 30#include "libavcodec/bytestream.h"
f63412fc 31#include "libavcodec/vorbis_parser.h"
9146ca37 32#include "avformat.h"
19711af5 33#include "internal.h"
a0ddef24 34#include "oggdec.h"
fb66c31d 35#include "vorbiscomment.h"
9146ca37 36
8730fad5
DC
37static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
38{
39 int i, cnum, h, m, s, ms, keylen = strlen(key);
40 AVChapter *chapter = NULL;
41
42 if (keylen < 9 || sscanf(key, "CHAPTER%02d", &cnum) != 1)
43 return 0;
44
45 if (keylen == 9) {
46 if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
47 return 0;
48
1fa395e4 49 avpriv_new_chapter(as, cnum, (AVRational){1,1000},
8730fad5
DC
50 ms + 1000*(s + 60*(m + 60*h)),
51 AV_NOPTS_VALUE, NULL);
52 av_free(val);
53 } else if (!strcmp(key+9, "NAME")) {
54 for(i = 0; i < as->nb_chapters; i++)
55 if (as->chapters[i]->id == cnum) {
56 chapter = as->chapters[i];
57 break;
58 }
59 if (!chapter)
60 return 0;
61
d2d67e42
AK
62 av_dict_set(&chapter->metadata, "title", val,
63 AV_DICT_DONT_STRDUP_VAL);
8730fad5
DC
64 } else
65 return 0;
66
67 av_free(key);
68 return 1;
69}
70
9686df2b 71int
d2d67e42 72ff_vorbis_comment(AVFormatContext * as, AVDictionary **m, const uint8_t *buf, int size)
9146ca37 73{
47a0513b
BC
74 const uint8_t *p = buf;
75 const uint8_t *end = buf + size;
98422c44
RD
76 unsigned n, j;
77 int s;
9146ca37 78
f5475e1b 79 if (size < 8) /* must have vendor_length and user_comment_list_length */
9146ca37
MR
80 return -1;
81
0a770ae7 82 s = bytestream_get_le32(&p);
9146ca37 83
98422c44 84 if (end - p - 4 < s || s < 0)
9146ca37
MR
85 return -1;
86
87 p += s;
9146ca37 88
0a770ae7 89 n = bytestream_get_le32(&p);
9146ca37 90
98422c44 91 while (end - p >= 4 && n > 0) {
47a0513b 92 const char *t, *v;
9146ca37
MR
93 int tl, vl;
94
0a770ae7 95 s = bytestream_get_le32(&p);
9146ca37 96
98422c44 97 if (end - p < s || s < 0)
9146ca37
MR
98 break;
99
100 t = p;
101 p += s;
9146ca37
MR
102 n--;
103
4bd684bc 104 v = memchr(t, '=', s);
9146ca37
MR
105 if (!v)
106 continue;
107
108 tl = v - t;
109 vl = s - tl - 1;
110 v++;
111
4bd684bc 112 if (tl && vl) {
e3b44649
JR
113 char *tt, *ct;
114
115 tt = av_malloc(tl + 1);
116 ct = av_malloc(vl + 1);
117 if (!tt || !ct) {
118 av_freep(&tt);
119 av_freep(&ct);
120 av_log(as, AV_LOG_WARNING, "out-of-memory error. skipping VorbisComment tag.\n");
121 continue;
122 }
9146ca37
MR
123
124 for (j = 0; j < tl; j++)
efa7f420 125 tt[j] = av_toupper(t[j]);
9146ca37
MR
126 tt[tl] = 0;
127
4bd684bc 128 memcpy(ct, v, vl);
9146ca37
MR
129 ct[vl] = 0;
130
8730fad5 131 if (!ogm_chapter(as, tt, ct))
d2d67e42
AK
132 av_dict_set(m, tt, ct,
133 AV_DICT_DONT_STRDUP_KEY |
134 AV_DICT_DONT_STRDUP_VAL);
9146ca37
MR
135 }
136 }
137
972c5f9e 138 if (p != end)
7d507ceb 139 av_log(as, AV_LOG_INFO, "%ti bytes of comment header remain\n", end-p);
9146ca37 140 if (n > 0)
4bd684bc
MR
141 av_log(as, AV_LOG_INFO,
142 "truncated comment header, %i comments not found\n", n);
9146ca37 143
ad7768f4 144 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
03700d39 145
9146ca37
MR
146 return 0;
147}
148
149
150/** Parse the vorbis header
151 * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
152 * [vorbis_version] = read 32 bits as unsigned integer | Not used
153 * [audio_channels] = read 8 bit integer as unsigned | Used
115329f1 154 * [audio_sample_rate] = read 32 bits as unsigned integer | Used
9146ca37
MR
155 * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
156 * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
157 * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
158 * [blocksize_0] = read 4 bits as unsigned integer | Not Used
159 * [blocksize_1] = read 4 bits as unsigned integer | Not Used
160 * [framing_flag] = read one bit | Not Used
161 * */
162
77be08ee 163struct oggvorbis_private {
ad2b531d
MR
164 unsigned int len[3];
165 unsigned char *packet[3];
f63412fc
JR
166 VorbisParseContext vp;
167 int64_t final_pts;
168 int final_duration;
77be08ee 169};
ad2b531d
MR
170
171
172static unsigned int
77be08ee 173fixup_vorbis_headers(AVFormatContext * as, struct oggvorbis_private *priv,
19f4ceca 174 uint8_t **buf)
ad2b531d 175{
5626f994 176 int i, offset, len, err;
ad2b531d
MR
177 unsigned char *ptr;
178
179 len = priv->len[0] + priv->len[1] + priv->len[2];
180 ptr = *buf = av_mallocz(len + len/255 + 64);
181
182 ptr[0] = 2;
183 offset = 1;
184 offset += av_xiphlacing(&ptr[offset], priv->len[0]);
185 offset += av_xiphlacing(&ptr[offset], priv->len[1]);
4bd684bc 186 for (i = 0; i < 3; i++) {
ad2b531d
MR
187 memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
188 offset += priv->len[i];
2ac41150 189 av_freep(&priv->packet[i]);
ad2b531d 190 }
5626f994
AK
191 if ((err = av_reallocp(buf, offset + FF_INPUT_BUFFER_PADDING_SIZE)) < 0)
192 return err;
ad2b531d
MR
193 return offset;
194}
195
fba8e5b6 196static void vorbis_cleanup(AVFormatContext *s, int idx)
d894f747
LB
197{
198 struct ogg *ogg = s->priv_data;
199 struct ogg_stream *os = ogg->streams + idx;
200 struct oggvorbis_private *priv = os->private;
201 int i;
202 if (os->private)
203 for (i = 0; i < 3; i++)
204 av_freep(&priv->packet[i]);
205}
ad2b531d 206
9146ca37
MR
207static int
208vorbis_header (AVFormatContext * s, int idx)
209{
77be08ee
MR
210 struct ogg *ogg = s->priv_data;
211 struct ogg_stream *os = ogg->streams + idx;
9146ca37 212 AVStream *st = s->streams[idx];
77be08ee 213 struct oggvorbis_private *priv;
8f8320d7 214 int pkt_type = os->buf[os->pstart];
9146ca37 215
8f8320d7 216 if (!os->private) {
77be08ee 217 os->private = av_mallocz(sizeof(struct oggvorbis_private));
4bd684bc 218 if (!os->private)
ad2b531d
MR
219 return 0;
220 }
9146ca37 221
3562684d
LB
222 if (!(pkt_type & 1))
223 return 0;
224
8f8320d7 225 if (os->psize < 1 || pkt_type > 5)
f5475e1b
MR
226 return -1;
227
ad2b531d 228 priv = os->private;
73c44cb2
AC
229
230 if (priv->packet[pkt_type>>1])
231 return -1;
232 if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
233 return -1;
234
8f8320d7
DC
235 priv->len[pkt_type >> 1] = os->psize;
236 priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
237 memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
9146ca37 238 if (os->buf[os->pstart] == 1) {
47a0513b 239 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
736e63ed 240 unsigned blocksize, bs0, bs1;
ce20edb7 241 int srate;
f5475e1b
MR
242
243 if (os->psize != 30)
244 return -1;
245
736e63ed
MR
246 if (bytestream_get_le32(&p) != 0) /* vorbis_version */
247 return -1;
248
739587bf 249 st->codec->channels = bytestream_get_byte(&p);
ce20edb7 250 srate = bytestream_get_le32(&p);
739587bf
MR
251 p += 4; // skip maximum bitrate
252 st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
736e63ed
MR
253 p += 4; // skip minimum bitrate
254
255 blocksize = bytestream_get_byte(&p);
256 bs0 = blocksize & 15;
257 bs1 = blocksize >> 4;
258
259 if (bs0 > bs1)
260 return -1;
261 if (bs0 < 6 || bs1 > 13)
262 return -1;
263
264 if (bytestream_get_byte(&p) != 1) /* framing_flag */
265 return -1;
9146ca37 266
72415b2a 267 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
36ef5369 268 st->codec->codec_id = AV_CODEC_ID_VORBIS;
9146ca37 269
ce20edb7
RD
270 if (srate > 0) {
271 st->codec->sample_rate = srate;
c3f9ebf7 272 avpriv_set_pts_info(st, 64, 1, srate);
ce20edb7 273 }
9146ca37 274 } else if (os->buf[os->pstart] == 3) {
8cb3c557
RD
275 if (os->psize > 8 &&
276 ff_vorbis_comment(s, &st->metadata, os->buf + os->pstart + 7, os->psize - 8) >= 0) {
277 // drop all metadata we parsed and which is not required by libvorbis
278 unsigned new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
279 if (new_len >= 16 && new_len < os->psize) {
280 AV_WL32(priv->packet[1] + new_len - 5, 0);
281 priv->packet[1][new_len - 1] = 1;
282 priv->len[1] = new_len;
283 }
284 }
ad2b531d 285 } else {
d872fb0f
MS
286 int ret = fixup_vorbis_headers(s, priv, &st->codec->extradata);
287 if (ret < 0) {
288 st->codec->extradata_size = 0;
289 return ret;
290 }
291 st->codec->extradata_size = ret;
f63412fc
JR
292 if ((ret = avpriv_vorbis_parse_extradata(st->codec, &priv->vp))) {
293 av_freep(&st->codec->extradata);
294 st->codec->extradata_size = 0;
295 return ret;
296 }
9146ca37
MR
297 }
298
8f8320d7 299 return 1;
9146ca37
MR
300}
301
f63412fc
JR
302static int vorbis_packet(AVFormatContext *s, int idx)
303{
304 struct ogg *ogg = s->priv_data;
305 struct ogg_stream *os = ogg->streams + idx;
306 struct oggvorbis_private *priv = os->private;
307 int duration;
308
309 /* first packet handling
310 here we parse the duration of each packet in the first page and compare
311 the total duration to the page granule to find the encoder delay and
312 set the first timestamp */
313 if (!os->lastpts) {
314 int seg;
315 uint8_t *last_pkt = os->buf + os->pstart;
316 uint8_t *next_pkt = last_pkt;
317 int first_duration = 0;
318
319 avpriv_vorbis_parse_reset(&priv->vp);
320 duration = 0;
321 for (seg = 0; seg < os->nsegs; seg++) {
322 if (os->segments[seg] < 255) {
323 int d = avpriv_vorbis_parse_frame(&priv->vp, last_pkt, 1);
324 if (d < 0) {
325 duration = os->granule;
326 break;
327 }
328 if (!duration)
329 first_duration = d;
330 duration += d;
331 last_pkt = next_pkt + os->segments[seg];
332 }
333 next_pkt += os->segments[seg];
334 }
335 os->lastpts = os->lastdts = os->granule - duration;
336 s->streams[idx]->start_time = os->lastpts + first_duration;
337 if (s->streams[idx]->duration)
338 s->streams[idx]->duration -= s->streams[idx]->start_time;
339 s->streams[idx]->cur_dts = AV_NOPTS_VALUE;
340 priv->final_pts = AV_NOPTS_VALUE;
341 avpriv_vorbis_parse_reset(&priv->vp);
342 }
343
344 /* parse packet duration */
345 if (os->psize > 0) {
346 duration = avpriv_vorbis_parse_frame(&priv->vp, os->buf + os->pstart, 1);
347 if (duration <= 0) {
348 os->pflags |= AV_PKT_FLAG_CORRUPT;
349 return 0;
350 }
351 os->pduration = duration;
352 }
353
354 /* final packet handling
355 here we save the pts of the first packet in the final page, sum up all
356 packet durations in the final page except for the last one, and compare
357 to the page granule to find the duration of the final packet */
358 if (os->flags & OGG_FLAG_EOS) {
359 if (os->lastpts != AV_NOPTS_VALUE) {
360 priv->final_pts = os->lastpts;
361 priv->final_duration = 0;
362 }
363 if (os->segp == os->nsegs)
364 os->pduration = os->granule - priv->final_pts - priv->final_duration;
365 priv->final_duration += os->pduration;
366 }
367
368 return 0;
369}
370
77be08ee 371const struct ogg_codec ff_vorbis_codec = {
9146ca37
MR
372 .magic = "\001vorbis",
373 .magicsize = 7,
f63412fc
JR
374 .header = vorbis_header,
375 .packet = vorbis_packet,
d894f747 376 .cleanup= vorbis_cleanup,
7751e469 377 .nb_header = 3,
9146ca37 378};