| 1 | /* |
| 2 | * xWMA demuxer |
| 3 | * Copyright (c) 2011 Max Horn |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 7 | * FFmpeg is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU Lesser General Public |
| 9 | * License as published by the Free Software Foundation; either |
| 10 | * version 2.1 of the License, or (at your option) any later version. |
| 11 | * |
| 12 | * FFmpeg is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * Lesser General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU Lesser General Public |
| 18 | * License along with FFmpeg; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | */ |
| 21 | |
| 22 | #include <inttypes.h> |
| 23 | #include <stdint.h> |
| 24 | |
| 25 | #include "avformat.h" |
| 26 | #include "internal.h" |
| 27 | #include "riff.h" |
| 28 | |
| 29 | /* |
| 30 | * Demuxer for xWMA, a Microsoft audio container used by XAudio 2. |
| 31 | */ |
| 32 | |
| 33 | typedef struct { |
| 34 | int64_t data_end; |
| 35 | } XWMAContext; |
| 36 | |
| 37 | static int xwma_probe(AVProbeData *p) |
| 38 | { |
| 39 | if (!memcmp(p->buf, "RIFF", 4) && !memcmp(p->buf + 8, "XWMA", 4)) |
| 40 | return AVPROBE_SCORE_MAX; |
| 41 | return 0; |
| 42 | } |
| 43 | |
| 44 | static int xwma_read_header(AVFormatContext *s) |
| 45 | { |
| 46 | int64_t size; |
| 47 | int ret; |
| 48 | uint32_t dpds_table_size = 0; |
| 49 | uint32_t *dpds_table = NULL; |
| 50 | unsigned int tag; |
| 51 | AVIOContext *pb = s->pb; |
| 52 | AVStream *st; |
| 53 | XWMAContext *xwma = s->priv_data; |
| 54 | int i; |
| 55 | |
| 56 | /* The following code is mostly copied from wav.c, with some |
| 57 | * minor alterations. |
| 58 | */ |
| 59 | |
| 60 | /* check RIFF header */ |
| 61 | tag = avio_rl32(pb); |
| 62 | if (tag != MKTAG('R', 'I', 'F', 'F')) |
| 63 | return -1; |
| 64 | avio_rl32(pb); /* file size */ |
| 65 | tag = avio_rl32(pb); |
| 66 | if (tag != MKTAG('X', 'W', 'M', 'A')) |
| 67 | return -1; |
| 68 | |
| 69 | /* parse fmt header */ |
| 70 | tag = avio_rl32(pb); |
| 71 | if (tag != MKTAG('f', 'm', 't', ' ')) |
| 72 | return -1; |
| 73 | size = avio_rl32(pb); |
| 74 | st = avformat_new_stream(s, NULL); |
| 75 | if (!st) |
| 76 | return AVERROR(ENOMEM); |
| 77 | |
| 78 | ret = ff_get_wav_header(pb, st->codec, size); |
| 79 | if (ret < 0) |
| 80 | return ret; |
| 81 | st->need_parsing = AVSTREAM_PARSE_NONE; |
| 82 | |
| 83 | /* All xWMA files I have seen contained WMAv2 data. If there are files |
| 84 | * using WMA Pro or some other codec, then we need to figure out the right |
| 85 | * extradata for that. Thus, ask the user for feedback, but try to go on |
| 86 | * anyway. |
| 87 | */ |
| 88 | if (st->codec->codec_id != AV_CODEC_ID_WMAV2) { |
| 89 | avpriv_request_sample(s, "Unexpected codec (tag 0x04%x; id %d)", |
| 90 | st->codec->codec_tag, st->codec->codec_id); |
| 91 | } else { |
| 92 | /* In all xWMA files I have seen, there is no extradata. But the WMA |
| 93 | * codecs require extradata, so we provide our own fake extradata. |
| 94 | * |
| 95 | * First, check that there really was no extradata in the header. If |
| 96 | * there was, then try to use it, after asking the user to provide a |
| 97 | * sample of this unusual file. |
| 98 | */ |
| 99 | if (st->codec->extradata_size != 0) { |
| 100 | /* Surprise, surprise: We *did* get some extradata. No idea |
| 101 | * if it will work, but just go on and try it, after asking |
| 102 | * the user for a sample. |
| 103 | */ |
| 104 | avpriv_request_sample(s, "Unexpected extradata (%d bytes)", |
| 105 | st->codec->extradata_size); |
| 106 | } else { |
| 107 | st->codec->extradata_size = 6; |
| 108 | st->codec->extradata = av_mallocz(6 + FF_INPUT_BUFFER_PADDING_SIZE); |
| 109 | if (!st->codec->extradata) |
| 110 | return AVERROR(ENOMEM); |
| 111 | |
| 112 | /* setup extradata with our experimentally obtained value */ |
| 113 | st->codec->extradata[4] = 31; |
| 114 | } |
| 115 | } |
| 116 | |
| 117 | if (!st->codec->channels) { |
| 118 | av_log(s, AV_LOG_WARNING, "Invalid channel count: %d\n", |
| 119 | st->codec->channels); |
| 120 | return AVERROR_INVALIDDATA; |
| 121 | } |
| 122 | if (!st->codec->bits_per_coded_sample) { |
| 123 | av_log(s, AV_LOG_WARNING, "Invalid bits_per_coded_sample: %d\n", |
| 124 | st->codec->bits_per_coded_sample); |
| 125 | return AVERROR_INVALIDDATA; |
| 126 | } |
| 127 | |
| 128 | /* set the sample rate */ |
| 129 | avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate); |
| 130 | |
| 131 | /* parse the remaining RIFF chunks */ |
| 132 | for (;;) { |
| 133 | if (pb->eof_reached) { |
| 134 | ret = AVERROR_EOF; |
| 135 | goto end; |
| 136 | } |
| 137 | /* read next chunk tag */ |
| 138 | tag = avio_rl32(pb); |
| 139 | size = avio_rl32(pb); |
| 140 | if (tag == MKTAG('d', 'a', 't', 'a')) { |
| 141 | /* We assume that the data chunk comes last. */ |
| 142 | break; |
| 143 | } else if (tag == MKTAG('d','p','d','s')) { |
| 144 | /* Quoting the MSDN xWMA docs on the dpds chunk: "Contains the |
| 145 | * decoded packet cumulative data size array, each element is the |
| 146 | * number of bytes accumulated after the corresponding xWMA packet |
| 147 | * is decoded in order." |
| 148 | * |
| 149 | * Each packet has size equal to st->codec->block_align, which in |
| 150 | * all cases I saw so far was always 2230. Thus, we can use the |
| 151 | * dpds data to compute a seeking index. |
| 152 | */ |
| 153 | |
| 154 | /* Error out if there is more than one dpds chunk. */ |
| 155 | if (dpds_table) { |
| 156 | av_log(s, AV_LOG_ERROR, "two dpds chunks present\n"); |
| 157 | ret = AVERROR_INVALIDDATA; |
| 158 | goto end; |
| 159 | } |
| 160 | |
| 161 | /* Compute the number of entries in the dpds chunk. */ |
| 162 | if (size & 3) { /* Size should be divisible by four */ |
| 163 | av_log(s, AV_LOG_WARNING, |
| 164 | "dpds chunk size %"PRId64" not divisible by 4\n", size); |
| 165 | } |
| 166 | dpds_table_size = size / 4; |
| 167 | if (dpds_table_size == 0 || dpds_table_size >= INT_MAX / 4) { |
| 168 | av_log(s, AV_LOG_ERROR, |
| 169 | "dpds chunk size %"PRId64" invalid\n", size); |
| 170 | return AVERROR_INVALIDDATA; |
| 171 | } |
| 172 | |
| 173 | /* Allocate some temporary storage to keep the dpds data around. |
| 174 | * for processing later on. |
| 175 | */ |
| 176 | dpds_table = av_malloc(dpds_table_size * sizeof(uint32_t)); |
| 177 | if (!dpds_table) { |
| 178 | return AVERROR(ENOMEM); |
| 179 | } |
| 180 | |
| 181 | for (i = 0; i < dpds_table_size; ++i) { |
| 182 | dpds_table[i] = avio_rl32(pb); |
| 183 | size -= 4; |
| 184 | } |
| 185 | } |
| 186 | avio_skip(pb, size); |
| 187 | } |
| 188 | |
| 189 | /* Determine overall data length */ |
| 190 | if (size < 0) { |
| 191 | ret = AVERROR_INVALIDDATA; |
| 192 | goto end; |
| 193 | } |
| 194 | if (!size) { |
| 195 | xwma->data_end = INT64_MAX; |
| 196 | } else |
| 197 | xwma->data_end = avio_tell(pb) + size; |
| 198 | |
| 199 | |
| 200 | if (dpds_table && dpds_table_size) { |
| 201 | int64_t cur_pos; |
| 202 | const uint32_t bytes_per_sample |
| 203 | = (st->codec->channels * st->codec->bits_per_coded_sample) >> 3; |
| 204 | |
| 205 | /* Estimate the duration from the total number of output bytes. */ |
| 206 | const uint64_t total_decoded_bytes = dpds_table[dpds_table_size - 1]; |
| 207 | |
| 208 | if (!bytes_per_sample) { |
| 209 | av_log(s, AV_LOG_ERROR, |
| 210 | "Invalid bits_per_coded_sample %d for %d channels\n", |
| 211 | st->codec->bits_per_coded_sample, st->codec->channels); |
| 212 | ret = AVERROR_INVALIDDATA; |
| 213 | goto end; |
| 214 | } |
| 215 | |
| 216 | st->duration = total_decoded_bytes / bytes_per_sample; |
| 217 | |
| 218 | /* Use the dpds data to build a seek table. We can only do this after |
| 219 | * we know the offset to the data chunk, as we need that to determine |
| 220 | * the actual offset to each input block. |
| 221 | * Note: If we allowed ourselves to assume that the data chunk always |
| 222 | * follows immediately after the dpds block, we could of course guess |
| 223 | * the data block's start offset already while reading the dpds chunk. |
| 224 | * I decided against that, just in case other chunks ever are |
| 225 | * discovered. |
| 226 | */ |
| 227 | cur_pos = avio_tell(pb); |
| 228 | for (i = 0; i < dpds_table_size; ++i) { |
| 229 | /* From the number of output bytes that would accumulate in the |
| 230 | * output buffer after decoding the first (i+1) packets, we compute |
| 231 | * an offset / timestamp pair. |
| 232 | */ |
| 233 | av_add_index_entry(st, |
| 234 | cur_pos + (i+1) * st->codec->block_align, /* pos */ |
| 235 | dpds_table[i] / bytes_per_sample, /* timestamp */ |
| 236 | st->codec->block_align, /* size */ |
| 237 | 0, /* duration */ |
| 238 | AVINDEX_KEYFRAME); |
| 239 | } |
| 240 | } else if (st->codec->bit_rate) { |
| 241 | /* No dpds chunk was present (or only an empty one), so estimate |
| 242 | * the total duration using the average bits per sample and the |
| 243 | * total data length. |
| 244 | */ |
| 245 | st->duration = (size<<3) * st->codec->sample_rate / st->codec->bit_rate; |
| 246 | } |
| 247 | |
| 248 | end: |
| 249 | av_free(dpds_table); |
| 250 | |
| 251 | return ret; |
| 252 | } |
| 253 | |
| 254 | static int xwma_read_packet(AVFormatContext *s, AVPacket *pkt) |
| 255 | { |
| 256 | int ret, size; |
| 257 | int64_t left; |
| 258 | AVStream *st; |
| 259 | XWMAContext *xwma = s->priv_data; |
| 260 | |
| 261 | st = s->streams[0]; |
| 262 | |
| 263 | left = xwma->data_end - avio_tell(s->pb); |
| 264 | if (left <= 0) { |
| 265 | return AVERROR_EOF; |
| 266 | } |
| 267 | |
| 268 | /* read a single block; the default block size is 2230. */ |
| 269 | size = (st->codec->block_align > 1) ? st->codec->block_align : 2230; |
| 270 | size = FFMIN(size, left); |
| 271 | |
| 272 | ret = av_get_packet(s->pb, pkt, size); |
| 273 | if (ret < 0) |
| 274 | return ret; |
| 275 | |
| 276 | pkt->stream_index = 0; |
| 277 | return ret; |
| 278 | } |
| 279 | |
| 280 | AVInputFormat ff_xwma_demuxer = { |
| 281 | .name = "xwma", |
| 282 | .long_name = NULL_IF_CONFIG_SMALL("Microsoft xWMA"), |
| 283 | .priv_data_size = sizeof(XWMAContext), |
| 284 | .read_probe = xwma_probe, |
| 285 | .read_header = xwma_read_header, |
| 286 | .read_packet = xwma_read_packet, |
| 287 | }; |