FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
tta.c
Go to the documentation of this file.
1 /*
2  * TTA (The Lossless True Audio) decoder
3  * Copyright (c) 2006 Alex Beregszaszi
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 /**
23  * @file
24  * TTA (The Lossless True Audio) decoder
25  * @see http://www.true-audio.com/
26  * @see http://tta.corecodec.org/
27  * @author Alex Beregszaszi
28  */
29 
30 #define BITSTREAM_READER_LE
31 //#define DEBUG
32 #include <limits.h>
33 #include "avcodec.h"
34 #include "get_bits.h"
35 #include "internal.h"
36 #include "libavutil/crc.h"
37 #include "libavutil/intreadwrite.h"
38 #include "libavutil/opt.h"
39 
40 #define FORMAT_SIMPLE 1
41 #define FORMAT_ENCRYPTED 2
42 
43 #define MAX_ORDER 16
44 typedef struct TTAFilter {
49 } TTAFilter;
50 
51 typedef struct TTARice {
52  uint32_t k0, k1, sum0, sum1;
53 } TTARice;
54 
55 typedef struct TTAChannel {
59 } TTAChannel;
60 
61 typedef struct TTAContext {
62  AVClass *class;
65  const AVCRC *crc_table;
66 
68  unsigned data_length;
70 
72 
76 } TTAContext;
77 
78 static const uint32_t shift_1[] = {
79  0x00000001, 0x00000002, 0x00000004, 0x00000008,
80  0x00000010, 0x00000020, 0x00000040, 0x00000080,
81  0x00000100, 0x00000200, 0x00000400, 0x00000800,
82  0x00001000, 0x00002000, 0x00004000, 0x00008000,
83  0x00010000, 0x00020000, 0x00040000, 0x00080000,
84  0x00100000, 0x00200000, 0x00400000, 0x00800000,
85  0x01000000, 0x02000000, 0x04000000, 0x08000000,
86  0x10000000, 0x20000000, 0x40000000, 0x80000000,
87  0x80000000, 0x80000000, 0x80000000, 0x80000000,
88  0x80000000, 0x80000000, 0x80000000, 0x80000000
89 };
90 
91 static const uint32_t * const shift_16 = shift_1 + 4;
92 
93 static const int32_t ttafilter_configs[4] = {
94  10,
95  9,
96  10,
97  12
98 };
99 
101  memset(c, 0, sizeof(TTAFilter));
102  if (s->format == FORMAT_ENCRYPTED) {
103  int i;
104  for (i = 0; i < 8; i++)
105  c->qm[i] = sign_extend(s->crc_pass[i], 8);
106  }
107  c->shift = shift;
108  c->round = shift_1[shift-1];
109 // c->round = 1 << (shift - 1);
110 }
111 
112 static inline void ttafilter_process(TTAFilter *c, int32_t *in)
113 {
114  register int32_t *dl = c->dl, *qm = c->qm, *dx = c->dx, sum = c->round;
115 
116  if (c->error < 0) {
117  qm[0] -= dx[0]; qm[1] -= dx[1]; qm[2] -= dx[2]; qm[3] -= dx[3];
118  qm[4] -= dx[4]; qm[5] -= dx[5]; qm[6] -= dx[6]; qm[7] -= dx[7];
119  } else if (c->error > 0) {
120  qm[0] += dx[0]; qm[1] += dx[1]; qm[2] += dx[2]; qm[3] += dx[3];
121  qm[4] += dx[4]; qm[5] += dx[5]; qm[6] += dx[6]; qm[7] += dx[7];
122  }
123 
124  sum += dl[0] * qm[0] + dl[1] * qm[1] + dl[2] * qm[2] + dl[3] * qm[3] +
125  dl[4] * qm[4] + dl[5] * qm[5] + dl[6] * qm[6] + dl[7] * qm[7];
126 
127  dx[0] = dx[1]; dx[1] = dx[2]; dx[2] = dx[3]; dx[3] = dx[4];
128  dl[0] = dl[1]; dl[1] = dl[2]; dl[2] = dl[3]; dl[3] = dl[4];
129 
130  dx[4] = ((dl[4] >> 30) | 1);
131  dx[5] = ((dl[5] >> 30) | 2) & ~1;
132  dx[6] = ((dl[6] >> 30) | 2) & ~1;
133  dx[7] = ((dl[7] >> 30) | 4) & ~3;
134 
135  c->error = *in;
136  *in += (sum >> c->shift);
137 
138  dl[4] = -dl[5]; dl[5] = -dl[6];
139  dl[6] = *in - dl[7]; dl[7] = *in;
140  dl[5] += dl[6]; dl[4] += dl[5];
141 }
142 
143 static void rice_init(TTARice *c, uint32_t k0, uint32_t k1)
144 {
145  c->k0 = k0;
146  c->k1 = k1;
147  c->sum0 = shift_16[k0];
148  c->sum1 = shift_16[k1];
149 }
150 
152 {
153  int ret = 0;
154 
155  // count ones
156  while (get_bits_left(gb) > 0 && get_bits1(gb))
157  ret++;
158  return ret;
159 }
160 
161 static const int64_t tta_channel_layouts[7] = {
165  0,
167  AV_CH_LAYOUT_5POINT1_BACK|AV_CH_BACK_CENTER,
169 };
170 
171 static int tta_check_crc(TTAContext *s, const uint8_t *buf, int buf_size)
172 {
173  uint32_t crc, CRC;
174 
175  CRC = AV_RL32(buf + buf_size);
176  crc = av_crc(s->crc_table, 0xFFFFFFFFU, buf, buf_size);
177  if (CRC != (crc ^ 0xFFFFFFFFU)) {
178  av_log(s->avctx, AV_LOG_ERROR, "CRC error\n");
179  return AVERROR_INVALIDDATA;
180  }
181 
182  return 0;
183 }
184 
185 static uint64_t tta_check_crc64(uint8_t *pass)
186 {
187  uint64_t crc = UINT64_MAX, poly = 0x42F0E1EBA9EA3693U;
188  uint8_t *end = pass + strlen(pass);
189  int i;
190 
191  while (pass < end) {
192  crc ^= (uint64_t)*pass++ << 56;
193  for (i = 0; i < 8; i++)
194  crc = (crc << 1) ^ (poly & (((int64_t) crc) >> 63));
195  }
196 
197  return crc ^ UINT64_MAX;
198 }
199 
201 {
202  TTAContext *s = avctx->priv_data;
203  int total_frames;
204 
205  s->avctx = avctx;
206 
207  // 30bytes includes a seektable with one frame
208  if (avctx->extradata_size < 30)
209  return AVERROR_INVALIDDATA;
210 
211  init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size * 8);
212  if (show_bits_long(&s->gb, 32) == AV_RL32("TTA1"))
213  {
214  if (avctx->err_recognition & AV_EF_CRCCHECK) {
216  tta_check_crc(s, avctx->extradata, 18);
217  }
218 
219  /* signature */
220  skip_bits_long(&s->gb, 32);
221 
222  s->format = get_bits(&s->gb, 16);
223  if (s->format > 2) {
224  av_log(avctx, AV_LOG_ERROR, "Invalid format\n");
225  return AVERROR_INVALIDDATA;
226  }
227  if (s->format == FORMAT_ENCRYPTED) {
228  if (!s->pass) {
229  av_log(avctx, AV_LOG_ERROR, "Missing password for encrypted stream. Please use the -password option\n");
230  return AVERROR(EINVAL);
231  }
233  }
234  avctx->channels = s->channels = get_bits(&s->gb, 16);
235  if (s->channels > 1 && s->channels < 9)
237  avctx->bits_per_raw_sample = get_bits(&s->gb, 16);
238  s->bps = (avctx->bits_per_raw_sample + 7) / 8;
239  avctx->sample_rate = get_bits_long(&s->gb, 32);
240  s->data_length = get_bits_long(&s->gb, 32);
241  skip_bits_long(&s->gb, 32); // CRC32 of header
242 
243  if (s->channels == 0) {
244  av_log(avctx, AV_LOG_ERROR, "Invalid number of channels\n");
245  return AVERROR_INVALIDDATA;
246  } else if (avctx->sample_rate == 0) {
247  av_log(avctx, AV_LOG_ERROR, "Invalid samplerate\n");
248  return AVERROR_INVALIDDATA;
249  }
250 
251  switch(s->bps) {
252  case 1: avctx->sample_fmt = AV_SAMPLE_FMT_U8; break;
253  case 2:
254  avctx->sample_fmt = AV_SAMPLE_FMT_S16;
255  break;
256  case 3:
257  avctx->sample_fmt = AV_SAMPLE_FMT_S32;
258  break;
259  //case 4: avctx->sample_fmt = AV_SAMPLE_FMT_S32; break;
260  default:
261  av_log(avctx, AV_LOG_ERROR, "Invalid/unsupported sample format.\n");
262  return AVERROR_INVALIDDATA;
263  }
264 
265  // prevent overflow
266  if (avctx->sample_rate > 0x7FFFFFu) {
267  av_log(avctx, AV_LOG_ERROR, "sample_rate too large\n");
268  return AVERROR(EINVAL);
269  }
270  s->frame_length = 256 * avctx->sample_rate / 245;
271 
273  total_frames = s->data_length / s->frame_length +
274  (s->last_frame_length ? 1 : 0);
275 
276  av_log(avctx, AV_LOG_DEBUG, "format: %d chans: %d bps: %d rate: %d block: %d\n",
277  s->format, avctx->channels, avctx->bits_per_coded_sample, avctx->sample_rate,
278  avctx->block_align);
279  av_log(avctx, AV_LOG_DEBUG, "data_length: %d frame_length: %d last: %d total: %d\n",
280  s->data_length, s->frame_length, s->last_frame_length, total_frames);
281 
282  // FIXME: seek table
283  if (avctx->extradata_size <= 26 || total_frames > INT_MAX / 4 ||
284  avctx->extradata_size - 26 < total_frames * 4)
285  av_log(avctx, AV_LOG_WARNING, "Seek table missing or too small\n");
286  else if (avctx->err_recognition & AV_EF_CRCCHECK) {
287  if (tta_check_crc(s, avctx->extradata + 22, total_frames * 4))
288  return AVERROR_INVALIDDATA;
289  }
290  skip_bits_long(&s->gb, 32 * total_frames);
291  skip_bits_long(&s->gb, 32); // CRC32 of seektable
292 
293  if(s->frame_length >= UINT_MAX / (s->channels * sizeof(int32_t))){
294  av_log(avctx, AV_LOG_ERROR, "frame_length too large\n");
295  return AVERROR_INVALIDDATA;
296  }
297 
298  if (s->bps < 3) {
300  if (!s->decode_buffer)
301  return AVERROR(ENOMEM);
302  } else
303  s->decode_buffer = NULL;
304  s->ch_ctx = av_malloc(avctx->channels * sizeof(*s->ch_ctx));
305  if (!s->ch_ctx) {
306  av_freep(&s->decode_buffer);
307  return AVERROR(ENOMEM);
308  }
309  } else {
310  av_log(avctx, AV_LOG_ERROR, "Wrong extradata present\n");
311  return AVERROR_INVALIDDATA;
312  }
313 
314  return 0;
315 }
316 
317 static int tta_decode_frame(AVCodecContext *avctx, void *data,
318  int *got_frame_ptr, AVPacket *avpkt)
319 {
320  AVFrame *frame = data;
321  const uint8_t *buf = avpkt->data;
322  int buf_size = avpkt->size;
323  TTAContext *s = avctx->priv_data;
324  int i, ret;
325  int cur_chan = 0, framelen = s->frame_length;
326  int32_t *p;
327 
328  if (avctx->err_recognition & AV_EF_CRCCHECK) {
329  if (buf_size < 4 || tta_check_crc(s, buf, buf_size - 4))
330  return AVERROR_INVALIDDATA;
331  }
332 
333  init_get_bits(&s->gb, buf, buf_size*8);
334 
335  /* get output buffer */
336  frame->nb_samples = framelen;
337  if ((ret = ff_get_buffer(avctx, frame)) < 0) {
338  av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
339  return ret;
340  }
341 
342  // decode directly to output buffer for 24-bit sample format
343  if (s->bps == 3)
344  s->decode_buffer = (int32_t *)frame->data[0];
345 
346  // init per channel states
347  for (i = 0; i < s->channels; i++) {
348  s->ch_ctx[i].predictor = 0;
350  rice_init(&s->ch_ctx[i].rice, 10, 10);
351  }
352 
353  i = 0;
354  for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++) {
355  int32_t *predictor = &s->ch_ctx[cur_chan].predictor;
356  TTAFilter *filter = &s->ch_ctx[cur_chan].filter;
357  TTARice *rice = &s->ch_ctx[cur_chan].rice;
358  uint32_t unary, depth, k;
359  int32_t value;
360 
361  unary = tta_get_unary(&s->gb);
362 
363  if (unary == 0) {
364  depth = 0;
365  k = rice->k0;
366  } else {
367  depth = 1;
368  k = rice->k1;
369  unary--;
370  }
371 
372  if (get_bits_left(&s->gb) < k) {
373  ret = AVERROR_INVALIDDATA;
374  goto error;
375  }
376 
377  if (k) {
378  if (k > MIN_CACHE_BITS) {
379  ret = AVERROR_INVALIDDATA;
380  goto error;
381  }
382  value = (unary << k) + get_bits(&s->gb, k);
383  } else
384  value = unary;
385 
386  // FIXME: copy paste from original
387  switch (depth) {
388  case 1:
389  rice->sum1 += value - (rice->sum1 >> 4);
390  if (rice->k1 > 0 && rice->sum1 < shift_16[rice->k1])
391  rice->k1--;
392  else if(rice->sum1 > shift_16[rice->k1 + 1])
393  rice->k1++;
394  value += shift_1[rice->k0];
395  default:
396  rice->sum0 += value - (rice->sum0 >> 4);
397  if (rice->k0 > 0 && rice->sum0 < shift_16[rice->k0])
398  rice->k0--;
399  else if(rice->sum0 > shift_16[rice->k0 + 1])
400  rice->k0++;
401  }
402 
403  // extract coded value
404  *p = 1 + ((value >> 1) ^ ((value & 1) - 1));
405 
406  // run hybrid filter
407  ttafilter_process(filter, p);
408 
409  // fixed order prediction
410 #define PRED(x, k) (int32_t)((((uint64_t)x << k) - x) >> k)
411  switch (s->bps) {
412  case 1: *p += PRED(*predictor, 4); break;
413  case 2:
414  case 3: *p += PRED(*predictor, 5); break;
415  case 4: *p += *predictor; break;
416  }
417  *predictor = *p;
418 
419  // flip channels
420  if (cur_chan < (s->channels-1))
421  cur_chan++;
422  else {
423  // decorrelate in case of multiple channels
424  if (s->channels > 1) {
425  int32_t *r = p - 1;
426  for (*p += *r / 2; r > p - s->channels; r--)
427  *r = *(r + 1) - *r;
428  }
429  cur_chan = 0;
430  i++;
431  // check for last frame
432  if (i == s->last_frame_length && get_bits_left(&s->gb) / 8 == 4) {
433  frame->nb_samples = framelen = s->last_frame_length;
434  break;
435  }
436  }
437  }
438 
439  align_get_bits(&s->gb);
440  if (get_bits_left(&s->gb) < 32) {
441  ret = AVERROR_INVALIDDATA;
442  goto error;
443  }
444  skip_bits_long(&s->gb, 32); // frame crc
445 
446  // convert to output buffer
447  switch (s->bps) {
448  case 1: {
449  uint8_t *samples = (uint8_t *)frame->data[0];
450  for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
451  *samples++ = *p + 0x80;
452  break;
453  }
454  case 2: {
455  int16_t *samples = (int16_t *)frame->data[0];
456  for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
457  *samples++ = *p;
458  break;
459  }
460  case 3: {
461  // shift samples for 24-bit sample format
462  int32_t *samples = (int32_t *)frame->data[0];
463  for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
464  *samples++ <<= 8;
465  // reset decode buffer
466  s->decode_buffer = NULL;
467  break;
468  }
469  }
470 
471  *got_frame_ptr = 1;
472 
473  return buf_size;
474 error:
475  // reset decode buffer
476  if (s->bps == 3)
477  s->decode_buffer = NULL;
478  return ret;
479 }
480 
482  TTAContext *s = avctx->priv_data;
483 
484  if (s->bps < 3)
486  s->decode_buffer = NULL;
487  av_freep(&s->ch_ctx);
488 
489  return 0;
490 }
491 
492 #define OFFSET(x) offsetof(TTAContext, x)
493 #define DEC (AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM)
494 static const AVOption options[] = {
495  { "password", "Set decoding password", OFFSET(pass), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, DEC },
496  { NULL },
497 };
498 
499 static const AVClass tta_decoder_class = {
500  .class_name = "TTA Decoder",
501  .item_name = av_default_item_name,
502  .option = options,
503  .version = LIBAVUTIL_VERSION_INT,
504 };
505 
507  .name = "tta",
508  .type = AVMEDIA_TYPE_AUDIO,
509  .id = AV_CODEC_ID_TTA,
510  .priv_data_size = sizeof(TTAContext),
514  .capabilities = CODEC_CAP_DR1,
515  .long_name = NULL_IF_CONFIG_SMALL("TTA (True Audio)"),
516  .priv_class = &tta_decoder_class,
517 };