FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
xmv.c
Go to the documentation of this file.
1 /*
2  * Microsoft XMV demuxer
3  * Copyright (c) 2011 Sven Hesse <drmccoy@drmccoy.de>
4  * Copyright (c) 2011 Matthew Hoops <clone2727@gmail.com>
5  *
6  * This file is part of FFmpeg.
7  *
8  * FFmpeg is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * FFmpeg is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with FFmpeg; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
23 /**
24  * @file
25  * Microsoft XMV demuxer
26  */
27 
28 #include <stdint.h>
29 
30 #include "libavutil/intreadwrite.h"
31 
32 #include "avformat.h"
33 #include "internal.h"
34 #include "riff.h"
35 #include "libavutil/avassert.h"
36 
37 /** The min size of an XMV header. */
38 #define XMV_MIN_HEADER_SIZE 36
39 
40 /** Audio flag: ADPCM'd 5.1 stream, front left / right channels */
41 #define XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT 1
42 /** Audio flag: ADPCM'd 5.1 stream, front center / low frequency channels */
43 #define XMV_AUDIO_ADPCM51_FRONTCENTERLOW 2
44 /** Audio flag: ADPCM'd 5.1 stream, rear left / right channels */
45 #define XMV_AUDIO_ADPCM51_REARLEFTRIGHT 4
46 
47 /** Audio flag: Any of the ADPCM'd 5.1 stream flags. */
48 #define XMV_AUDIO_ADPCM51 (XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT | \
49  XMV_AUDIO_ADPCM51_FRONTCENTERLOW | \
50  XMV_AUDIO_ADPCM51_REARLEFTRIGHT)
51 
52 /** A video packet with an XMV file. */
53 typedef struct XMVVideoPacket {
54  int stream_index; ///< The decoder stream index for this video packet.
55 
56  uint32_t data_size; ///< The size of the remaining video data.
57  uint64_t data_offset; ///< The offset of the video data within the file.
58 
59  uint32_t current_frame; ///< The current frame within this video packet.
60  uint32_t frame_count; ///< The amount of frames within this video packet.
61 
62  int has_extradata; ///< Does the video packet contain extra data?
63  uint8_t extradata[4]; ///< The extra data
64 
65  int64_t last_pts; ///< PTS of the last video frame.
66  int64_t pts; ///< PTS of the most current video frame.
68 
69 /** An audio packet with an XMV file. */
70 typedef struct XMVAudioPacket {
71  int stream_index; ///< The decoder stream index for this audio packet.
72 
73  /* Stream format properties. */
74  uint16_t compression; ///< The type of compression.
75  uint16_t channels; ///< Number of channels.
76  uint32_t sample_rate; ///< Sampling rate.
77  uint16_t bits_per_sample; ///< Bits per compressed sample.
78  uint32_t bit_rate; ///< Bits of compressed data per second.
79  uint16_t flags; ///< Flags
80  unsigned block_align; ///< Bytes per compressed block.
81  uint16_t block_samples; ///< Decompressed samples per compressed block.
82 
83  enum AVCodecID codec_id; ///< The codec ID of the compression scheme.
84 
85  uint32_t data_size; ///< The size of the remaining audio data.
86  uint64_t data_offset; ///< The offset of the audio data within the file.
87 
88  uint32_t frame_size; ///< Number of bytes to put into an audio frame.
89 
90  uint64_t block_count; ///< Running counter of decompressed audio block.
92 
93 /** Context for demuxing an XMV file. */
94 typedef struct XMVDemuxContext {
95  uint16_t audio_track_count; ///< Number of audio track in this file.
96 
97  uint32_t this_packet_size; ///< Size of the current packet.
98  uint32_t next_packet_size; ///< Size of the next packet.
99 
100  uint64_t this_packet_offset; ///< Offset of the current packet.
101  uint64_t next_packet_offset; ///< Offset of the next packet.
102 
103  uint16_t current_stream; ///< The index of the stream currently handling.
104  uint16_t stream_count; ///< The number of streams in this file.
105 
106  XMVVideoPacket video; ///< The video packet contained in each packet.
107  XMVAudioPacket *audio; ///< The audio packets contained in each packet.
109 
110 static int xmv_probe(AVProbeData *p)
111 {
112  uint32_t file_version;
113 
114  if (p->buf_size < XMV_MIN_HEADER_SIZE)
115  return 0;
116 
117  file_version = AV_RL32(p->buf + 16);
118  if ((file_version == 0) || (file_version > 4))
119  return 0;
120 
121  if (!memcmp(p->buf + 12, "xobX", 4))
122  return AVPROBE_SCORE_MAX;
123 
124  return 0;
125 }
126 
128 {
129  XMVDemuxContext *xmv = s->priv_data;
130  AVIOContext *pb = s->pb;
131  AVStream *vst = NULL;
132 
133  uint32_t file_version;
134  uint32_t this_packet_size;
135  uint16_t audio_track;
136 
137  avio_skip(pb, 4); /* Next packet size */
138 
139  this_packet_size = avio_rl32(pb);
140 
141  avio_skip(pb, 4); /* Max packet size */
142  avio_skip(pb, 4); /* "xobX" */
143 
144  file_version = avio_rl32(pb);
145  if ((file_version != 4) && (file_version != 2))
146  av_log_ask_for_sample(s, "Found uncommon version %d\n", file_version);
147 
148 
149  /* Video track */
150 
151  vst = avformat_new_stream(s, NULL);
152  if (!vst)
153  return AVERROR(ENOMEM);
154 
155  avpriv_set_pts_info(vst, 32, 1, 1000);
156 
159  vst->codec->codec_tag = MKBETAG('W', 'M', 'V', '2');
160  vst->codec->width = avio_rl32(pb);
161  vst->codec->height = avio_rl32(pb);
162 
163  vst->duration = avio_rl32(pb);
164 
165  xmv->video.stream_index = vst->index;
166 
167  /* Audio tracks */
168 
169  xmv->audio_track_count = avio_rl16(pb);
170 
171  avio_skip(pb, 2); /* Unknown (padding?) */
172 
173  xmv->audio = av_malloc(xmv->audio_track_count * sizeof(XMVAudioPacket));
174  if (!xmv->audio)
175  return AVERROR(ENOMEM);
176 
177  for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
178  XMVAudioPacket *packet = &xmv->audio[audio_track];
179  AVStream *ast = NULL;
180 
181  packet->compression = avio_rl16(pb);
182  packet->channels = avio_rl16(pb);
183  packet->sample_rate = avio_rl32(pb);
184  packet->bits_per_sample = avio_rl16(pb);
185  packet->flags = avio_rl16(pb);
186 
187  if (!packet->channels) {
188  av_log(s, AV_LOG_ERROR, "0 channels\n");
189  return AVERROR(EINVAL);
190  }
191 
192  packet->bit_rate = packet->bits_per_sample *
193  packet->sample_rate *
194  packet->channels;
195  packet->block_align = 36 * packet->channels;
196  packet->block_samples = 64;
197  packet->codec_id = ff_wav_codec_get_id(packet->compression,
198  packet->bits_per_sample);
199 
200  packet->stream_index = -1;
201 
202  packet->frame_size = 0;
203  packet->block_count = 0;
204 
205  /* TODO: ADPCM'd 5.1 sound is encoded in three separate streams.
206  * Those need to be interleaved to a proper 5.1 stream. */
207  if (packet->flags & XMV_AUDIO_ADPCM51)
208  av_log(s, AV_LOG_WARNING, "Unsupported 5.1 ADPCM audio stream "
209  "(0x%04X)\n", packet->flags);
210 
211  ast = avformat_new_stream(s, NULL);
212  if (!ast)
213  return AVERROR(ENOMEM);
214 
216  ast->codec->codec_id = packet->codec_id;
217  ast->codec->codec_tag = packet->compression;
218  ast->codec->channels = packet->channels;
219  ast->codec->sample_rate = packet->sample_rate;
221  ast->codec->bit_rate = packet->bit_rate;
222  ast->codec->block_align = 36 * packet->channels;
223 
224  avpriv_set_pts_info(ast, 32, packet->block_samples, packet->sample_rate);
225 
226  packet->stream_index = ast->index;
227 
228  ast->duration = vst->duration;
229  }
230 
231 
232  /* Initialize the packet context */
233 
234  xmv->next_packet_offset = avio_tell(pb);
235  xmv->next_packet_size = this_packet_size - xmv->next_packet_offset;
236  xmv->stream_count = xmv->audio_track_count + 1;
237 
238  return 0;
239 }
240 
241 static void xmv_read_extradata(uint8_t *extradata, AVIOContext *pb)
242 {
243  /* Read the XMV extradata */
244 
245  uint32_t data = avio_rl32(pb);
246 
247  int mspel_bit = !!(data & 0x01);
248  int loop_filter = !!(data & 0x02);
249  int abt_flag = !!(data & 0x04);
250  int j_type_bit = !!(data & 0x08);
251  int top_left_mv_flag = !!(data & 0x10);
252  int per_mb_rl_bit = !!(data & 0x20);
253  int slice_count = (data >> 6) & 7;
254 
255  /* Write it back as standard WMV2 extradata */
256 
257  data = 0;
258 
259  data |= mspel_bit << 15;
260  data |= loop_filter << 14;
261  data |= abt_flag << 13;
262  data |= j_type_bit << 12;
263  data |= top_left_mv_flag << 11;
264  data |= per_mb_rl_bit << 10;
265  data |= slice_count << 7;
266 
267  AV_WB32(extradata, data);
268 }
269 
271 {
272  XMVDemuxContext *xmv = s->priv_data;
273  AVIOContext *pb = s->pb;
274 
275  uint8_t data[8];
276  uint16_t audio_track;
277  uint64_t data_offset;
278 
279  /* Next packet size */
280  xmv->next_packet_size = avio_rl32(pb);
281 
282  /* Packet video header */
283 
284  if (avio_read(pb, data, 8) != 8)
285  return AVERROR(EIO);
286 
287  xmv->video.data_size = AV_RL32(data) & 0x007FFFFF;
288 
289  xmv->video.current_frame = 0;
290  xmv->video.frame_count = (AV_RL32(data) >> 23) & 0xFF;
291 
292  xmv->video.has_extradata = (data[3] & 0x80) != 0;
293 
294  /* Adding the audio data sizes and the video data size keeps you 4 bytes
295  * short for every audio track. But as playing around with XMV files with
296  * ADPCM audio showed, taking the extra 4 bytes from the audio data gives
297  * you either completely distorted audio or click (when skipping the
298  * remaining 68 bytes of the ADPCM block). Subtracting 4 bytes for every
299  * audio track from the video data works at least for the audio. Probably
300  * some alignment thing?
301  * The video data has (always?) lots of padding, so it should work out...
302  */
303  xmv->video.data_size -= xmv->audio_track_count * 4;
304 
305  xmv->current_stream = 0;
306  if (!xmv->video.frame_count) {
307  xmv->video.frame_count = 1;
308  xmv->current_stream = xmv->stream_count > 1;
309  }
310 
311  /* Packet audio header */
312 
313  for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
314  XMVAudioPacket *packet = &xmv->audio[audio_track];
315 
316  if (avio_read(pb, data, 4) != 4)
317  return AVERROR(EIO);
318 
319  packet->data_size = AV_RL32(data) & 0x007FFFFF;
320  if ((packet->data_size == 0) && (audio_track != 0))
321  /* This happens when I create an XMV with several identical audio
322  * streams. From the size calculations, duplicating the previous
323  * stream's size works out, but the track data itself is silent.
324  * Maybe this should also redirect the offset to the previous track?
325  */
326  packet->data_size = xmv->audio[audio_track - 1].data_size;
327 
328  /* Carve up the audio data in frame_count slices */
329  packet->frame_size = packet->data_size / xmv->video.frame_count;
330  packet->frame_size -= packet->frame_size % packet->block_align;
331  }
332 
333  /* Packet data offsets */
334 
335  data_offset = avio_tell(pb);
336 
337  xmv->video.data_offset = data_offset;
338  data_offset += xmv->video.data_size;
339 
340  for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
341  xmv->audio[audio_track].data_offset = data_offset;
342  data_offset += xmv->audio[audio_track].data_size;
343  }
344 
345  /* Video frames header */
346 
347  /* Read new video extra data */
348  if (xmv->video.data_size > 0) {
349  if (xmv->video.has_extradata) {
351 
352  xmv->video.data_size -= 4;
353  xmv->video.data_offset += 4;
354 
355  if (xmv->video.stream_index >= 0) {
356  AVStream *vst = s->streams[xmv->video.stream_index];
357 
359 
360  if (vst->codec->extradata_size < 4) {
361  av_free(vst->codec->extradata);
362 
363  vst->codec->extradata =
365  vst->codec->extradata_size = 4;
366  }
367 
368  memcpy(vst->codec->extradata, xmv->video.extradata, 4);
369  }
370  }
371  }
372 
373  return 0;
374 }
375 
377 {
378  XMVDemuxContext *xmv = s->priv_data;
379  AVIOContext *pb = s->pb;
380  int result;
381 
382  if (xmv->this_packet_offset == xmv->next_packet_offset)
383  return AVERROR_EOF;
384 
385  /* Seek to it */
387  if (avio_seek(pb, xmv->this_packet_offset, SEEK_SET) != xmv->this_packet_offset)
388  return AVERROR(EIO);
389 
390  /* Update the size */
392  if (xmv->this_packet_size < (12 + xmv->audio_track_count * 4))
393  return AVERROR(EIO);
394 
395  /* Process the header */
396  result = xmv_process_packet_header(s);
397  if (result)
398  return result;
399 
400  /* Update the offset */
402 
403  return 0;
404 }
405 
407  AVPacket *pkt, uint32_t stream)
408 {
409  XMVDemuxContext *xmv = s->priv_data;
410  AVIOContext *pb = s->pb;
411  XMVAudioPacket *audio = &xmv->audio[stream];
412 
413  uint32_t data_size;
414  uint32_t block_count;
415  int result;
416 
417  /* Seek to it */
418  if (avio_seek(pb, audio->data_offset, SEEK_SET) != audio->data_offset)
419  return AVERROR(EIO);
420 
421  if ((xmv->video.current_frame + 1) < xmv->video.frame_count)
422  /* Not the last frame, get at most frame_size bytes. */
423  data_size = FFMIN(audio->frame_size, audio->data_size);
424  else
425  /* Last frame, get the rest. */
426  data_size = audio->data_size;
427 
428  /* Read the packet */
429  result = av_get_packet(pb, pkt, data_size);
430  if (result <= 0)
431  return result;
432 
433  pkt->stream_index = audio->stream_index;
434 
435  /* Calculate the PTS */
436 
437  block_count = data_size / audio->block_align;
438 
439  pkt->duration = block_count;
440  pkt->pts = audio->block_count;
441  pkt->dts = AV_NOPTS_VALUE;
442 
443  audio->block_count += block_count;
444 
445  /* Advance offset */
446  audio->data_size -= data_size;
447  audio->data_offset += data_size;
448 
449  return 0;
450 }
451 
453  AVPacket *pkt)
454 {
455  XMVDemuxContext *xmv = s->priv_data;
456  AVIOContext *pb = s->pb;
457  XMVVideoPacket *video = &xmv->video;
458 
459  int result;
460  uint32_t frame_header;
461  uint32_t frame_size, frame_timestamp;
462  uint8_t *data, *end;
463 
464  /* Seek to it */
465  if (avio_seek(pb, video->data_offset, SEEK_SET) != video->data_offset)
466  return AVERROR(EIO);
467 
468  /* Read the frame header */
469  frame_header = avio_rl32(pb);
470 
471  frame_size = (frame_header & 0x1FFFF) * 4 + 4;
472  frame_timestamp = (frame_header >> 17);
473 
474  if ((frame_size + 4) > video->data_size)
475  return AVERROR(EIO);
476 
477  /* Get the packet data */
478  result = av_get_packet(pb, pkt, frame_size);
479  if (result != frame_size)
480  return result;
481 
482  /* Contrary to normal WMV2 video, the bit stream in XMV's
483  * WMV2 is little-endian.
484  * TODO: This manual swap is of course suboptimal.
485  */
486  for (data = pkt->data, end = pkt->data + frame_size; data < end; data += 4)
487  AV_WB32(data, AV_RL32(data));
488 
489  pkt->stream_index = video->stream_index;
490 
491  /* Calculate the PTS */
492 
493  video->last_pts = frame_timestamp + video->pts;
494 
495  pkt->duration = 0;
496  pkt->pts = video->last_pts;
497  pkt->dts = AV_NOPTS_VALUE;
498 
499  video->pts += frame_timestamp;
500 
501  /* Keyframe? */
502  pkt->flags = (pkt->data[0] & 0x80) ? 0 : AV_PKT_FLAG_KEY;
503 
504  /* Advance offset */
505  video->data_size -= frame_size + 4;
506  video->data_offset += frame_size + 4;
507 
508  return 0;
509 }
510 
512  AVPacket *pkt)
513 {
514  XMVDemuxContext *xmv = s->priv_data;
515  int result;
516 
517  if (xmv->video.current_frame == xmv->video.frame_count) {
518  /* No frames left in this packet, so we fetch a new one */
519 
520  result = xmv_fetch_new_packet(s);
521  if (result)
522  return result;
523  }
524 
525  if (xmv->current_stream == 0) {
526  /* Fetch a video frame */
527 
528  result = xmv_fetch_video_packet(s, pkt);
529  if (result)
530  return result;
531 
532  } else {
533  /* Fetch an audio frame */
534 
535  result = xmv_fetch_audio_packet(s, pkt, xmv->current_stream - 1);
536  if (result)
537  return result;
538  }
539 
540  /* Increase our counters */
541  if (++xmv->current_stream >= xmv->stream_count) {
542  xmv->current_stream = 0;
543  xmv->video.current_frame += 1;
544  }
545 
546  return 0;
547 }
548 
550 {
551  XMVDemuxContext *xmv = s->priv_data;
552 
553  av_freep(&xmv->audio);
554 
555  return 0;
556 }
557 
559  .name = "xmv",
560  .long_name = NULL_IF_CONFIG_SMALL("Microsoft XMV"),
561  .priv_data_size = sizeof(XMVDemuxContext),
566 };