| 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 "get_bits.h" |
| 29 | #include "parser.h" |
| 30 | #include "xiph.h" |
| 31 | #include "vorbis_parser.h" |
| 32 | |
| 33 | static int parse_id_header(AVCodecContext *avctx, VorbisParseContext *s, |
| 34 | const uint8_t *buf, int buf_size) |
| 35 | { |
| 36 | /* Id header should be 30 bytes */ |
| 37 | if (buf_size < 30) { |
| 38 | av_log(avctx, AV_LOG_ERROR, "Id header is too short\n"); |
| 39 | return AVERROR_INVALIDDATA; |
| 40 | } |
| 41 | |
| 42 | /* make sure this is the Id header */ |
| 43 | if (buf[0] != 1) { |
| 44 | av_log(avctx, AV_LOG_ERROR, "Wrong packet type in Id header\n"); |
| 45 | return AVERROR_INVALIDDATA; |
| 46 | } |
| 47 | |
| 48 | /* check for header signature */ |
| 49 | if (memcmp(&buf[1], "vorbis", 6)) { |
| 50 | av_log(avctx, AV_LOG_ERROR, "Invalid packet signature in Id header\n"); |
| 51 | return AVERROR_INVALIDDATA; |
| 52 | } |
| 53 | |
| 54 | if (!(buf[29] & 0x1)) { |
| 55 | av_log(avctx, AV_LOG_ERROR, "Invalid framing bit in Id header\n"); |
| 56 | return AVERROR_INVALIDDATA; |
| 57 | } |
| 58 | |
| 59 | s->blocksize[0] = 1 << (buf[28] & 0xF); |
| 60 | s->blocksize[1] = 1 << (buf[28] >> 4); |
| 61 | |
| 62 | return 0; |
| 63 | } |
| 64 | |
| 65 | static int parse_setup_header(AVCodecContext *avctx, VorbisParseContext *s, |
| 66 | const uint8_t *buf, int buf_size) |
| 67 | { |
| 68 | GetBitContext gb, gb0; |
| 69 | uint8_t *rev_buf; |
| 70 | int i, ret = 0; |
| 71 | int got_framing_bit, mode_count, got_mode_header, last_mode_count = 0; |
| 72 | |
| 73 | /* avoid overread */ |
| 74 | if (buf_size < 7) { |
| 75 | av_log(avctx, AV_LOG_ERROR, "Setup header is too short\n"); |
| 76 | return AVERROR_INVALIDDATA; |
| 77 | } |
| 78 | |
| 79 | /* make sure this is the Setup header */ |
| 80 | if (buf[0] != 5) { |
| 81 | av_log(avctx, AV_LOG_ERROR, "Wrong packet type in Setup header\n"); |
| 82 | return AVERROR_INVALIDDATA; |
| 83 | } |
| 84 | |
| 85 | /* check for header signature */ |
| 86 | if (memcmp(&buf[1], "vorbis", 6)) { |
| 87 | av_log(avctx, AV_LOG_ERROR, "Invalid packet signature in Setup header\n"); |
| 88 | return AVERROR_INVALIDDATA; |
| 89 | } |
| 90 | |
| 91 | /* reverse bytes so we can easily read backwards with get_bits() */ |
| 92 | if (!(rev_buf = av_malloc(buf_size))) { |
| 93 | av_log(avctx, AV_LOG_ERROR, "Out of memory\n"); |
| 94 | return AVERROR(ENOMEM); |
| 95 | } |
| 96 | for (i = 0; i < buf_size; i++) |
| 97 | rev_buf[i] = buf[buf_size - 1 - i]; |
| 98 | init_get_bits(&gb, rev_buf, buf_size * 8); |
| 99 | |
| 100 | got_framing_bit = 0; |
| 101 | while (get_bits_left(&gb) > 97) { |
| 102 | if (get_bits1(&gb)) { |
| 103 | got_framing_bit = get_bits_count(&gb); |
| 104 | break; |
| 105 | } |
| 106 | } |
| 107 | if (!got_framing_bit) { |
| 108 | av_log(avctx, AV_LOG_ERROR, "Invalid Setup header\n"); |
| 109 | ret = AVERROR_INVALIDDATA; |
| 110 | goto bad_header; |
| 111 | } |
| 112 | |
| 113 | /* Now we search backwards to find possible valid mode counts. This is not |
| 114 | * fool-proof because we could have false positive matches and read too |
| 115 | * far, but there isn't really any way to be sure without parsing through |
| 116 | * all the many variable-sized fields before the modes. This approach seems |
| 117 | * to work well in testing, and it is similar to how it is handled in |
| 118 | * liboggz. */ |
| 119 | mode_count = 0; |
| 120 | got_mode_header = 0; |
| 121 | while (get_bits_left(&gb) >= 97) { |
| 122 | if (get_bits(&gb, 8) > 63 || get_bits(&gb, 16) || get_bits(&gb, 16)) |
| 123 | break; |
| 124 | skip_bits(&gb, 1); |
| 125 | mode_count++; |
| 126 | if (mode_count > 64) |
| 127 | break; |
| 128 | gb0 = gb; |
| 129 | if (get_bits(&gb0, 6) + 1 == mode_count) { |
| 130 | got_mode_header = 1; |
| 131 | last_mode_count = mode_count; |
| 132 | } |
| 133 | } |
| 134 | if (!got_mode_header) { |
| 135 | av_log(avctx, AV_LOG_ERROR, "Invalid Setup header\n"); |
| 136 | ret = AVERROR_INVALIDDATA; |
| 137 | goto bad_header; |
| 138 | } |
| 139 | /* All samples I've seen use <= 2 modes, so ask for a sample if we find |
| 140 | * more than that, as it is most likely a false positive. If we get any |
| 141 | * we may need to approach this the long way and parse the whole Setup |
| 142 | * header, but I hope very much that it never comes to that. */ |
| 143 | if (last_mode_count > 2) { |
| 144 | avpriv_request_sample(avctx, |
| 145 | "%d modes (either a false positive or a " |
| 146 | "sample from an unknown encoder)", |
| 147 | last_mode_count); |
| 148 | } |
| 149 | /* We're limiting the mode count to 63 so that we know that the previous |
| 150 | * block flag will be in the first packet byte. */ |
| 151 | if (last_mode_count > 63) { |
| 152 | av_log(avctx, AV_LOG_ERROR, "Unsupported mode count: %d\n", |
| 153 | last_mode_count); |
| 154 | ret = AVERROR_INVALIDDATA; |
| 155 | goto bad_header; |
| 156 | } |
| 157 | s->mode_count = mode_count = last_mode_count; |
| 158 | /* Determine the number of bits required to code the mode and turn that |
| 159 | * into a bitmask to directly access the mode from the first frame byte. */ |
| 160 | s->mode_mask = ((1 << (av_log2(mode_count - 1) + 1)) - 1) << 1; |
| 161 | /* The previous window flag is the next bit after the mode */ |
| 162 | s->prev_mask = (s->mode_mask | 0x1) + 1; |
| 163 | |
| 164 | init_get_bits(&gb, rev_buf, buf_size * 8); |
| 165 | skip_bits_long(&gb, got_framing_bit); |
| 166 | for (i = mode_count - 1; i >= 0; i--) { |
| 167 | skip_bits_long(&gb, 40); |
| 168 | s->mode_blocksize[i] = get_bits1(&gb); |
| 169 | } |
| 170 | |
| 171 | bad_header: |
| 172 | av_free(rev_buf); |
| 173 | return ret; |
| 174 | } |
| 175 | |
| 176 | int avpriv_vorbis_parse_extradata(AVCodecContext *avctx, VorbisParseContext *s) |
| 177 | { |
| 178 | uint8_t *header_start[3]; |
| 179 | int header_len[3]; |
| 180 | int ret; |
| 181 | |
| 182 | s->avctx = avctx; |
| 183 | s->extradata_parsed = 1; |
| 184 | |
| 185 | if ((ret = avpriv_split_xiph_headers(avctx->extradata, |
| 186 | avctx->extradata_size, 30, |
| 187 | header_start, header_len)) < 0) { |
| 188 | av_log(avctx, AV_LOG_ERROR, "Extradata corrupt.\n"); |
| 189 | return ret; |
| 190 | } |
| 191 | |
| 192 | if ((ret = parse_id_header(avctx, s, header_start[0], header_len[0])) < 0) |
| 193 | return ret; |
| 194 | |
| 195 | if ((ret = parse_setup_header(avctx, s, header_start[2], header_len[2])) < 0) |
| 196 | return ret; |
| 197 | |
| 198 | s->valid_extradata = 1; |
| 199 | s->previous_blocksize = s->blocksize[s->mode_blocksize[0]]; |
| 200 | |
| 201 | return 0; |
| 202 | } |
| 203 | |
| 204 | int avpriv_vorbis_parse_frame_flags(VorbisParseContext *s, const uint8_t *buf, |
| 205 | int buf_size, int *flags) |
| 206 | { |
| 207 | int duration = 0; |
| 208 | |
| 209 | if (s->valid_extradata && buf_size > 0) { |
| 210 | int mode, current_blocksize; |
| 211 | int previous_blocksize = s->previous_blocksize; |
| 212 | |
| 213 | if (buf[0] & 1) { |
| 214 | /* If the user doesn't care about special packets, it's a bad one. */ |
| 215 | if (!flags) |
| 216 | goto bad_packet; |
| 217 | |
| 218 | /* Set the flag for which kind of special packet it is. */ |
| 219 | if (buf[0] == 1) |
| 220 | *flags |= VORBIS_FLAG_HEADER; |
| 221 | else if (buf[0] == 3) |
| 222 | *flags |= VORBIS_FLAG_COMMENT; |
| 223 | else |
| 224 | goto bad_packet; |
| 225 | |
| 226 | /* Special packets have no duration. */ |
| 227 | return 0; |
| 228 | |
| 229 | bad_packet: |
| 230 | av_log(s->avctx, AV_LOG_ERROR, "Invalid packet\n"); |
| 231 | return AVERROR_INVALIDDATA; |
| 232 | } |
| 233 | if (s->mode_count == 1) |
| 234 | mode = 0; |
| 235 | else |
| 236 | mode = (buf[0] & s->mode_mask) >> 1; |
| 237 | if (mode >= s->mode_count) { |
| 238 | av_log(s->avctx, AV_LOG_ERROR, "Invalid mode in packet\n"); |
| 239 | return AVERROR_INVALIDDATA; |
| 240 | } |
| 241 | if(s->mode_blocksize[mode]){ |
| 242 | int flag = !!(buf[0] & s->prev_mask); |
| 243 | previous_blocksize = s->blocksize[flag]; |
| 244 | } |
| 245 | current_blocksize = s->blocksize[s->mode_blocksize[mode]]; |
| 246 | duration = (previous_blocksize + current_blocksize) >> 2; |
| 247 | s->previous_blocksize = current_blocksize; |
| 248 | } |
| 249 | |
| 250 | return duration; |
| 251 | } |
| 252 | |
| 253 | int avpriv_vorbis_parse_frame(VorbisParseContext *s, const uint8_t *buf, |
| 254 | int buf_size) |
| 255 | { |
| 256 | return avpriv_vorbis_parse_frame_flags(s, buf, buf_size, NULL); |
| 257 | } |
| 258 | |
| 259 | void avpriv_vorbis_parse_reset(VorbisParseContext *s) |
| 260 | { |
| 261 | if (s->valid_extradata) |
| 262 | s->previous_blocksize = s->blocksize[0]; |
| 263 | } |
| 264 | |
| 265 | #if CONFIG_VORBIS_PARSER |
| 266 | static int vorbis_parse(AVCodecParserContext *s1, AVCodecContext *avctx, |
| 267 | const uint8_t **poutbuf, int *poutbuf_size, |
| 268 | const uint8_t *buf, int buf_size) |
| 269 | { |
| 270 | VorbisParseContext *s = s1->priv_data; |
| 271 | int duration; |
| 272 | |
| 273 | if (!s->extradata_parsed && avctx->extradata && avctx->extradata_size) |
| 274 | if (avpriv_vorbis_parse_extradata(avctx, s)) |
| 275 | goto end; |
| 276 | |
| 277 | if ((duration = avpriv_vorbis_parse_frame(s, buf, buf_size)) >= 0) |
| 278 | s1->duration = duration; |
| 279 | |
| 280 | end: |
| 281 | /* always return the full packet. this parser isn't doing any splitting or |
| 282 | combining, only packet analysis */ |
| 283 | *poutbuf = buf; |
| 284 | *poutbuf_size = buf_size; |
| 285 | return buf_size; |
| 286 | } |
| 287 | |
| 288 | AVCodecParser ff_vorbis_parser = { |
| 289 | .codec_ids = { AV_CODEC_ID_VORBIS }, |
| 290 | .priv_data_size = sizeof(VorbisParseContext), |
| 291 | .parser_parse = vorbis_parse, |
| 292 | }; |
| 293 | #endif /* CONFIG_VORBIS_PARSER */ |