1 /*
2 * Copyright (c) 2012 Justin Ruggles
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file
23 * Vorbis audio parser
24 *
25 * Determines the duration for each packet.
26 */
27
28 #include "config_components.h"
29
30 #include "libavutil/log.h"
31
32 #include "get_bits.h"
33 #include "parser.h"
34 #include "xiph.h"
35 #include "vorbis_parser_internal.h"
36
37 static const AVClass vorbis_parser_class = {
38 .class_name = "Vorbis parser",
39 .item_name = av_default_item_name,
40 .version = LIBAVUTIL_VERSION_INT,
41 };
42
parse_id_header(AVVorbisParseContext *s, const uint8_t *buf, int buf_size)43 static int parse_id_header(AVVorbisParseContext *s,
44 const uint8_t *buf, int buf_size)
45 {
46 /* Id header should be 30 bytes */
47 if (buf_size < 30) {
48 av_log(s, AV_LOG_ERROR, "Id header is too short\n");
49 return AVERROR_INVALIDDATA;
50 }
51
52 /* make sure this is the Id header */
53 if (buf[0] != 1) {
54 av_log(s, AV_LOG_ERROR, "Wrong packet type in Id header\n");
55 return AVERROR_INVALIDDATA;
56 }
57
58 /* check for header signature */
59 if (memcmp(&buf[1], "vorbis", 6)) {
60 av_log(s, AV_LOG_ERROR, "Invalid packet signature in Id header\n");
61 return AVERROR_INVALIDDATA;
62 }
63
64 if (!(buf[29] & 0x1)) {
65 av_log(s, AV_LOG_ERROR, "Invalid framing bit in Id header\n");
66 return AVERROR_INVALIDDATA;
67 }
68
69 s->blocksize[0] = 1 << (buf[28] & 0xF);
70 s->blocksize[1] = 1 << (buf[28] >> 4);
71
72 return 0;
73 }
74
parse_setup_header(AVVorbisParseContext *s, const uint8_t *buf, int buf_size)75 static int parse_setup_header(AVVorbisParseContext *s,
76 const uint8_t *buf, int buf_size)
77 {
78 GetBitContext gb, gb0;
79 uint8_t *rev_buf;
80 int i, ret = 0;
81 int got_framing_bit, mode_count, got_mode_header, last_mode_count = 0;
82
83 /* avoid overread */
84 if (buf_size < 7) {
85 av_log(s, AV_LOG_ERROR, "Setup header is too short\n");
86 return AVERROR_INVALIDDATA;
87 }
88
89 /* make sure this is the Setup header */
90 if (buf[0] != 5) {
91 av_log(s, AV_LOG_ERROR, "Wrong packet type in Setup header\n");
92 return AVERROR_INVALIDDATA;
93 }
94
95 /* check for header signature */
96 if (memcmp(&buf[1], "vorbis", 6)) {
97 av_log(s, AV_LOG_ERROR, "Invalid packet signature in Setup header\n");
98 return AVERROR_INVALIDDATA;
99 }
100
101 /* reverse bytes so we can easily read backwards with get_bits() */
102 if (!(rev_buf = av_malloc(buf_size))) {
103 av_log(s, AV_LOG_ERROR, "Out of memory\n");
104 return AVERROR(ENOMEM);
105 }
106 for (i = 0; i < buf_size; i++)
107 rev_buf[i] = buf[buf_size - 1 - i];
108 init_get_bits(&gb, rev_buf, buf_size * 8);
109
110 got_framing_bit = 0;
111 while (get_bits_left(&gb) > 97) {
112 if (get_bits1(&gb)) {
113 got_framing_bit = get_bits_count(&gb);
114 break;
115 }
116 }
117 if (!got_framing_bit) {
118 av_log(s, AV_LOG_ERROR, "Invalid Setup header\n");
119 ret = AVERROR_INVALIDDATA;
120 goto bad_header;
121 }
122
123 /* Now we search backwards to find possible valid mode counts. This is not
124 * fool-proof because we could have false positive matches and read too
125 * far, but there isn't really any way to be sure without parsing through
126 * all the many variable-sized fields before the modes. This approach seems
127 * to work well in testing, and it is similar to how it is handled in
128 * liboggz. */
129 mode_count = 0;
130 got_mode_header = 0;
131 while (get_bits_left(&gb) >= 97) {
132 if (get_bits(&gb, 8) > 63 || get_bits(&gb, 16) || get_bits(&gb, 16))
133 break;
134 skip_bits(&gb, 1);
135 mode_count++;
136 if (mode_count > 64)
137 break;
138 gb0 = gb;
139 if (get_bits(&gb0, 6) + 1 == mode_count) {
140 got_mode_header = 1;
141 last_mode_count = mode_count;
142 }
143 }
144 if (!got_mode_header) {
145 av_log(s, AV_LOG_ERROR, "Invalid Setup header\n");
146 ret = AVERROR_INVALIDDATA;
147 goto bad_header;
148 }
149 /* All samples I've seen use <= 2 modes, so ask for a sample if we find
150 * more than that, as it is most likely a false positive. If we get any
151 * we may need to approach this the long way and parse the whole Setup
152 * header, but I hope very much that it never comes to that. */
153 if (last_mode_count > 2) {
154 avpriv_request_sample(s,
155 "%d modes (either a false positive or a "
156 "sample from an unknown encoder)",
157 last_mode_count);
158 }
159 /* We're limiting the mode count to 63 so that we know that the previous
160 * block flag will be in the first packet byte. */
161 if (last_mode_count > 63) {
162 av_log(s, AV_LOG_ERROR, "Unsupported mode count: %d\n",
163 last_mode_count);
164 ret = AVERROR_INVALIDDATA;
165 goto bad_header;
166 }
167 s->mode_count = mode_count = last_mode_count;
168 /* Determine the number of bits required to code the mode and turn that
169 * into a bitmask to directly access the mode from the first frame byte. */
170 s->mode_mask = ((1 << (av_log2(mode_count - 1) + 1)) - 1) << 1;
171 /* The previous window flag is the next bit after the mode */
172 s->prev_mask = (s->mode_mask | 0x1) + 1;
173
174 init_get_bits(&gb, rev_buf, buf_size * 8);
175 skip_bits_long(&gb, got_framing_bit);
176 for (i = mode_count - 1; i >= 0; i--) {
177 skip_bits_long(&gb, 40);
178 s->mode_blocksize[i] = get_bits1(&gb);
179 }
180
181 bad_header:
182 av_free(rev_buf);
183 return ret;
184 }
185
vorbis_parse_init(AVVorbisParseContext *s, const uint8_t *extradata, int extradata_size)186 static int vorbis_parse_init(AVVorbisParseContext *s,
187 const uint8_t *extradata, int extradata_size)
188 {
189 const uint8_t *header_start[3];
190 int header_len[3];
191 int ret;
192
193 s->class = &vorbis_parser_class;
194 s->extradata_parsed = 1;
195
196 if ((ret = avpriv_split_xiph_headers(extradata,
197 extradata_size, 30,
198 header_start, header_len)) < 0) {
199 av_log(s, AV_LOG_ERROR, "Extradata corrupt.\n");
200 return ret;
201 }
202
203 if ((ret = parse_id_header(s, header_start[0], header_len[0])) < 0)
204 return ret;
205
206 if ((ret = parse_setup_header(s, header_start[2], header_len[2])) < 0)
207 return ret;
208
209 s->valid_extradata = 1;
210 s->previous_blocksize = s->blocksize[s->mode_blocksize[0]];
211
212 return 0;
213 }
214
av_vorbis_parse_frame_flags(AVVorbisParseContext *s, const uint8_t *buf, int buf_size, int *flags)215 int av_vorbis_parse_frame_flags(AVVorbisParseContext *s, const uint8_t *buf,
216 int buf_size, int *flags)
217 {
218 int duration = 0;
219
220 if (s->valid_extradata && buf_size > 0) {
221 int mode, current_blocksize;
222 int previous_blocksize = s->previous_blocksize;
223
224 if (buf[0] & 1) {
225 /* If the user doesn't care about special packets, it's a bad one. */
226 if (!flags)
227 goto bad_packet;
228
229 /* Set the flag for which kind of special packet it is. */
230 if (buf[0] == 1)
231 *flags |= VORBIS_FLAG_HEADER;
232 else if (buf[0] == 3)
233 *flags |= VORBIS_FLAG_COMMENT;
234 else if (buf[0] == 5)
235 *flags |= VORBIS_FLAG_SETUP;
236 else
237 goto bad_packet;
238
239 /* Special packets have no duration. */
240 return 0;
241
242 bad_packet:
243 av_log(s, AV_LOG_ERROR, "Invalid packet\n");
244 return AVERROR_INVALIDDATA;
245 }
246 if (s->mode_count == 1)
247 mode = 0;
248 else
249 mode = (buf[0] & s->mode_mask) >> 1;
250 if (mode >= s->mode_count) {
251 av_log(s, AV_LOG_ERROR, "Invalid mode in packet\n");
252 return AVERROR_INVALIDDATA;
253 }
254 if(s->mode_blocksize[mode]){
255 int flag = !!(buf[0] & s->prev_mask);
256 previous_blocksize = s->blocksize[flag];
257 }
258 current_blocksize = s->blocksize[s->mode_blocksize[mode]];
259 duration = (previous_blocksize + current_blocksize) >> 2;
260 s->previous_blocksize = current_blocksize;
261 }
262
263 return duration;
264 }
265
av_vorbis_parse_frame(AVVorbisParseContext *s, const uint8_t *buf, int buf_size)266 int av_vorbis_parse_frame(AVVorbisParseContext *s, const uint8_t *buf,
267 int buf_size)
268 {
269 return av_vorbis_parse_frame_flags(s, buf, buf_size, NULL);
270 }
271
av_vorbis_parse_reset(AVVorbisParseContext *s)272 void av_vorbis_parse_reset(AVVorbisParseContext *s)
273 {
274 if (s->valid_extradata)
275 s->previous_blocksize = s->blocksize[0];
276 }
277
av_vorbis_parse_free(AVVorbisParseContext **s)278 void av_vorbis_parse_free(AVVorbisParseContext **s)
279 {
280 av_freep(s);
281 }
282
av_vorbis_parse_init(const uint8_t *extradata, int extradata_size)283 AVVorbisParseContext *av_vorbis_parse_init(const uint8_t *extradata,
284 int extradata_size)
285 {
286 AVVorbisParseContext *s = av_mallocz(sizeof(*s));
287 int ret;
288
289 if (!s)
290 return NULL;
291
292 ret = vorbis_parse_init(s, extradata, extradata_size);
293 if (ret < 0) {
294 av_vorbis_parse_free(&s);
295 return NULL;
296 }
297
298 return s;
299 }
300
301 #if CONFIG_VORBIS_PARSER
302
303 typedef struct VorbisParseContext {
304 AVVorbisParseContext *vp;
305 } VorbisParseContext;
306
vorbis_parse(AVCodecParserContext *s1, AVCodecContext *avctx, const uint8_t **poutbuf, int *poutbuf_size, const uint8_t *buf, int buf_size)307 static int vorbis_parse(AVCodecParserContext *s1, AVCodecContext *avctx,
308 const uint8_t **poutbuf, int *poutbuf_size,
309 const uint8_t *buf, int buf_size)
310 {
311 VorbisParseContext *s = s1->priv_data;
312 int duration;
313
314 if (!s->vp && avctx->extradata && avctx->extradata_size) {
315 s->vp = av_vorbis_parse_init(avctx->extradata, avctx->extradata_size);
316 }
317 if (!s->vp)
318 goto end;
319
320 if ((duration = av_vorbis_parse_frame(s->vp, buf, buf_size)) >= 0)
321 s1->duration = duration;
322
323 end:
324 /* always return the full packet. this parser isn't doing any splitting or
325 combining, only packet analysis */
326 *poutbuf = buf;
327 *poutbuf_size = buf_size;
328 return buf_size;
329 }
330
vorbis_parser_close(AVCodecParserContext *ctx)331 static void vorbis_parser_close(AVCodecParserContext *ctx)
332 {
333 VorbisParseContext *s = ctx->priv_data;
334 av_vorbis_parse_free(&s->vp);
335 }
336
337 const AVCodecParser ff_vorbis_parser = {
338 .codec_ids = { AV_CODEC_ID_VORBIS },
339 .priv_data_size = sizeof(VorbisParseContext),
340 .parser_parse = vorbis_parse,
341 .parser_close = vorbis_parser_close,
342 };
343 #endif /* CONFIG_VORBIS_PARSER */
344