Converted globalhotkeys to XCB.
[vlc.git] / modules / packetizer / mpeg4video.c
1 /*****************************************************************************
2  * mpeg4video.c: mpeg 4 video packetizer
3  *****************************************************************************
4  * Copyright (C) 2001-2006 the VideoLAN team
5  * $Id$
6  *
7  * Authors: Gildas Bazin <gbazin@videolan.org>
8  *          Laurent Aimar <fenrir@via.ecp.fr>
9  *          Eric Petit <titer@videolan.org>
10  *
11  * This program is free software; you can redistribute it and/or modify
12  * it under the terms of the GNU General Public License as published by
13  * the Free Software Foundation; either version 2 of the License, or
14  * (at your option) any later version.
15  *
16  * This program is distributed in the hope that it will be useful,
17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19  * GNU General Public License for more details.
20  *
21  * You should have received a copy of the GNU General Public License
22  * along with this program; if not, write to the Free Software
23  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
24  *****************************************************************************/
25
26 /*****************************************************************************
27  * Preamble
28  *****************************************************************************/
29
30 #ifdef HAVE_CONFIG_H
31 # include "config.h"
32 #endif
33
34 #include <vlc_common.h>
35 #include <vlc_plugin.h>
36 #include <vlc_sout.h>
37 #include <vlc_codec.h>
38 #include <vlc_block.h>
39
40 #include "vlc_bits.h"
41 #include "vlc_block_helper.h"
42
43 /*****************************************************************************
44  * Module descriptor
45  *****************************************************************************/
46 static int  Open ( vlc_object_t * );
47 static void Close( vlc_object_t * );
48
49 vlc_module_begin ()
50     set_category( CAT_SOUT )
51     set_subcategory( SUBCAT_SOUT_PACKETIZER )
52     set_description( N_("MPEG4 video packetizer") )
53     set_capability( "packetizer", 50 )
54     set_callbacks( Open, Close )
55 vlc_module_end ()
56
57 /****************************************************************************
58  * Local prototypes
59  ****************************************************************************/
60 static block_t *Packetize( decoder_t *, block_t ** );
61
62 struct decoder_sys_t
63 {
64     /*
65      * Input properties
66      */
67     block_bytestream_t bytestream;
68     int i_state;
69     size_t i_offset;
70     uint8_t p_startcode[3];
71
72     /*
73      * Common properties
74      */
75     mtime_t i_interpolated_pts;
76     mtime_t i_interpolated_dts;
77     mtime_t i_last_ref_pts;
78     mtime_t i_last_time_ref;
79     mtime_t i_time_ref;
80     mtime_t i_last_time;
81     mtime_t i_last_timeincr;
82
83     unsigned int i_flags;
84
85     int         i_fps_num;
86     int         i_fps_den;
87     int         i_last_incr;
88     int         i_last_incr_diff;
89
90     bool  b_frame;
91
92     /* Current frame being built */
93     block_t    *p_frame;
94     block_t    **pp_last;
95 };
96
97 enum {
98     STATE_NOSYNC,
99     STATE_NEXT_SYNC
100 };
101
102 static block_t *ParseMPEGBlock( decoder_t *, block_t * );
103 static int ParseVOL( decoder_t *, es_format_t *, uint8_t *, int );
104 static int ParseVOP( decoder_t *, block_t * );
105 static int vlc_log2( unsigned int );
106
107 #define VIDEO_OBJECT_MASK                       0x01f
108 #define VIDEO_OBJECT_LAYER_MASK                 0x00f
109
110 #define VIDEO_OBJECT_START_CODE                 0x100
111 #define VIDEO_OBJECT_LAYER_START_CODE           0x120
112 #define VISUAL_OBJECT_SEQUENCE_START_CODE       0x1b0
113 #define VISUAL_OBJECT_SEQUENCE_END_CODE         0x1b1
114 #define USER_DATA_START_CODE                    0x1b2
115 #define GROUP_OF_VOP_START_CODE                 0x1b3
116 #define VIDEO_SESSION_ERROR_CODE                0x1b4
117 #define VISUAL_OBJECT_START_CODE                0x1b5
118 #define VOP_START_CODE                          0x1b6
119 #define FACE_OBJECT_START_CODE                  0x1ba
120 #define FACE_OBJECT_PLANE_START_CODE            0x1bb
121 #define MESH_OBJECT_START_CODE                  0x1bc
122 #define MESH_OBJECT_PLANE_START_CODE            0x1bd
123 #define STILL_TEXTURE_OBJECT_START_CODE         0x1be
124 #define TEXTURE_SPATIAL_LAYER_START_CODE        0x1bf
125 #define TEXTURE_SNR_LAYER_START_CODE            0x1c0
126
127 /*****************************************************************************
128  * Open: probe the packetizer and return score
129  *****************************************************************************/
130 static int Open( vlc_object_t *p_this )
131 {
132     decoder_t     *p_dec = (decoder_t*)p_this;
133     decoder_sys_t *p_sys;
134
135     switch( p_dec->fmt_in.i_codec )
136     {
137         case VLC_FOURCC( 'm', '4', 's', '2'):
138         case VLC_FOURCC( 'M', '4', 'S', '2'):
139         case VLC_FOURCC( 'm', 'p', '4', 's'):
140         case VLC_FOURCC( 'M', 'P', '4', 'S'):
141         case VLC_FOURCC( 'm', 'p', '4', 'v'):
142         case VLC_FOURCC( 'M', 'P', '4', 'V'):
143         case VLC_FOURCC( 'D', 'I', 'V', 'X'):
144         case VLC_FOURCC( 'd', 'i', 'v', 'x'):
145         case VLC_FOURCC( 'X', 'V', 'I', 'D'):
146         case VLC_FOURCC( 'X', 'v', 'i', 'D'):
147         case VLC_FOURCC( 'x', 'v', 'i', 'd'):
148         case VLC_FOURCC( 'D', 'X', '5', '0'):
149         case VLC_FOURCC( 'd', 'x', '5', '0'):
150         case VLC_FOURCC( 0x04, 0,   0,   0):
151         case VLC_FOURCC( '3', 'I', 'V', '2'):
152         case VLC_FOURCC( 'm', '4', 'c', 'c'):
153         case VLC_FOURCC( 'M', '4', 'C', 'C'):
154             break;
155
156         default:
157             return VLC_EGENERIC;
158     }
159
160     /* Allocate the memory needed to store the decoder's structure */
161     if( ( p_dec->p_sys = p_sys = malloc( sizeof(decoder_sys_t) ) ) == NULL )
162         return VLC_ENOMEM;
163     memset( p_sys, 0, sizeof(decoder_sys_t) );
164
165     /* Misc init */
166     p_sys->i_state = STATE_NOSYNC;
167     p_sys->bytestream = block_BytestreamInit();
168     p_sys->p_startcode[0] = 0;
169     p_sys->p_startcode[1] = 0;
170     p_sys->p_startcode[2] = 1;
171     p_sys->i_offset = 0;
172     p_sys->p_frame = NULL;
173     p_sys->pp_last = &p_sys->p_frame;
174
175     /* Setup properties */
176     es_format_Copy( &p_dec->fmt_out, &p_dec->fmt_in );
177     p_dec->fmt_out.i_codec = VLC_FOURCC( 'm', 'p', '4', 'v' );
178
179     if( p_dec->fmt_in.i_extra )
180     {
181         /* We have a vol */
182         p_dec->fmt_out.i_extra = p_dec->fmt_in.i_extra;
183         p_dec->fmt_out.p_extra = malloc( p_dec->fmt_in.i_extra );
184         memcpy( p_dec->fmt_out.p_extra, p_dec->fmt_in.p_extra,
185                 p_dec->fmt_in.i_extra );
186
187         msg_Dbg( p_dec, "opening with vol size: %d", p_dec->fmt_in.i_extra );
188         ParseVOL( p_dec, &p_dec->fmt_out,
189                   p_dec->fmt_out.p_extra, p_dec->fmt_out.i_extra );
190     }
191     else
192     {
193         /* No vol, we'll have to look for one later on */
194         p_dec->fmt_out.i_extra = 0;
195         p_dec->fmt_out.p_extra = 0;
196     }
197
198     /* Set callback */
199     p_dec->pf_packetize = Packetize;
200
201     return VLC_SUCCESS;
202 }
203
204 /*****************************************************************************
205  * Close: clean up the packetizer
206  *****************************************************************************/
207 static void Close( vlc_object_t *p_this )
208 {
209     decoder_t *p_dec = (decoder_t*)p_this;
210
211     block_BytestreamRelease( &p_dec->p_sys->bytestream );
212     if( p_dec->p_sys->p_frame ) block_ChainRelease( p_dec->p_sys->p_frame );
213     free( p_dec->p_sys );
214 }
215
216 /****************************************************************************
217  * Packetize: the whole thing
218  ****************************************************************************/
219 static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
220 {
221     decoder_sys_t *p_sys = p_dec->p_sys;
222     block_t       *p_pic;
223     mtime_t       i_pts, i_dts;
224
225     if( pp_block == NULL || *pp_block == NULL ) return NULL;
226
227     if( (*pp_block)->i_flags&(BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
228     {
229         if( (*pp_block)->i_flags&BLOCK_FLAG_CORRUPTED )
230         {
231             p_sys->i_state = STATE_NOSYNC;
232             block_BytestreamEmpty( &p_sys->bytestream );
233
234             if( p_sys->p_frame )
235                 block_ChainRelease( p_sys->p_frame );
236             p_sys->p_frame = NULL;
237             p_sys->pp_last = &p_sys->p_frame;
238         }
239         p_sys->i_interpolated_pts =
240         p_sys->i_interpolated_dts =
241         p_sys->i_last_ref_pts =
242         p_sys->i_last_time_ref =
243         p_sys->i_time_ref =
244         p_sys->i_last_time =
245         p_sys->i_last_timeincr = 0;
246
247         block_Release( *pp_block );
248         return NULL;
249     }
250
251     block_BytestreamPush( &p_sys->bytestream, *pp_block );
252
253     while( 1 )
254     {
255         switch( p_sys->i_state )
256         {
257
258         case STATE_NOSYNC:
259             if( block_FindStartcodeFromOffset( &p_sys->bytestream,
260                     &p_sys->i_offset, p_sys->p_startcode, 3 ) == VLC_SUCCESS )
261             {
262                 p_sys->i_state = STATE_NEXT_SYNC;
263             }
264
265             if( p_sys->i_offset )
266             {
267                 block_SkipBytes( &p_sys->bytestream, p_sys->i_offset );
268                 p_sys->i_offset = 0;
269                 block_BytestreamFlush( &p_sys->bytestream );
270             }
271
272             if( p_sys->i_state != STATE_NEXT_SYNC )
273             {
274                 /* Need more data */
275                 return NULL;
276             }
277
278             p_sys->i_offset = 1; /* To find next startcode */
279
280         case STATE_NEXT_SYNC:
281             /* TODO: If p_block == NULL, flush the buffer without checking the
282              * next sync word */
283
284             /* Find the next startcode */
285             if( block_FindStartcodeFromOffset( &p_sys->bytestream,
286                     &p_sys->i_offset, p_sys->p_startcode, 3 ) != VLC_SUCCESS )
287             {
288                 /* Need more data */
289                 return NULL;
290             }
291
292             /* Get the new fragment and set the pts/dts */
293             p_pic = block_New( p_dec, p_sys->i_offset );
294             block_BytestreamFlush( &p_sys->bytestream );
295             p_pic->i_pts = i_pts = p_sys->bytestream.p_block->i_pts;
296             p_pic->i_dts = i_dts = p_sys->bytestream.p_block->i_dts;
297
298             block_GetBytes( &p_sys->bytestream, p_pic->p_buffer,
299                             p_pic->i_buffer );
300
301             p_sys->i_offset = 0;
302
303             /* Get picture if any */
304             if( !( p_pic = ParseMPEGBlock( p_dec, p_pic ) ) )
305             {
306                 p_sys->i_state = STATE_NOSYNC;
307                 break;
308             }
309
310             /* don't reuse the same timestamps several times */
311             if( i_pts == p_sys->bytestream.p_block->i_pts &&
312                 i_dts == p_sys->bytestream.p_block->i_dts )
313             {
314                 p_sys->bytestream.p_block->i_pts = 0;
315                 p_sys->bytestream.p_block->i_dts = 0;
316             }
317
318             /* We've just started the stream, wait for the first PTS.
319              * We discard here so we can still get the sequence header. */
320             if( p_sys->i_interpolated_pts <= 0 &&
321                 p_sys->i_interpolated_dts <= 0 )
322             {
323                 msg_Dbg( p_dec, "need a starting pts/dts" );
324                 p_sys->i_state = STATE_NOSYNC;
325                 block_Release( p_pic );
326                 break;
327             }
328
329             /* When starting the stream we can have the first frame with
330              * a null DTS (i_interpolated_pts is initialized to 0) */
331             if( !p_pic->i_dts ) p_pic->i_dts = p_pic->i_pts;
332
333             /* So p_block doesn't get re-added several times */
334             *pp_block = block_BytestreamPop( &p_sys->bytestream );
335
336             p_sys->i_state = STATE_NOSYNC;
337
338             return p_pic;
339         }
340     }
341 }
342
343 /*****************************************************************************
344  * ParseMPEGBlock: Re-assemble fragments into a block containing a picture
345  *****************************************************************************/
346 static block_t *ParseMPEGBlock( decoder_t *p_dec, block_t *p_frag )
347 {
348     decoder_sys_t *p_sys = p_dec->p_sys;
349     block_t *p_pic = NULL;
350
351     if( p_frag->p_buffer[3] == 0xB0 || p_frag->p_buffer[3] == 0xB1 || p_frag->p_buffer[3] == 0xB2 )
352     {   /* VOS and USERDATA */
353 #if 0
354         /* Remove VOS start/end code from the original stream */
355         block_Release( p_frag );
356 #else
357         /* Append the block for now since ts/ps muxers rely on VOL
358          * being present in the stream */
359         block_ChainLastAppend( &p_sys->pp_last, p_frag );
360 #endif
361         return NULL;
362     }
363     if( p_frag->p_buffer[3] >= 0x20 && p_frag->p_buffer[3] <= 0x2f )
364     {
365         /* Copy the complete VOL */
366         if( (size_t)p_dec->fmt_out.i_extra != p_frag->i_buffer )
367         {
368             p_dec->fmt_out.p_extra =
369                 realloc( p_dec->fmt_out.p_extra, p_frag->i_buffer );
370             p_dec->fmt_out.i_extra = p_frag->i_buffer;
371         }
372         memcpy( p_dec->fmt_out.p_extra, p_frag->p_buffer, p_frag->i_buffer );
373         ParseVOL( p_dec, &p_dec->fmt_out,
374                   p_dec->fmt_out.p_extra, p_dec->fmt_out.i_extra );
375
376 #if 0
377         /* Remove from the original stream */
378         block_Release( p_frag );
379 #else
380         /* Append the block for now since ts/ps muxers rely on VOL
381          * being present in the stream */
382         block_ChainLastAppend( &p_sys->pp_last, p_frag );
383 #endif
384         return NULL;
385     }
386     else
387     {
388         if( !p_dec->fmt_out.i_extra )
389         {
390             msg_Warn( p_dec, "waiting for VOL" );
391             block_Release( p_frag );
392             return NULL;
393         }
394
395         /* Append the block */
396         block_ChainLastAppend( &p_sys->pp_last, p_frag );
397     }
398
399     if( p_frag->p_buffer[3] == 0xb6 &&
400         ParseVOP( p_dec, p_frag ) == VLC_SUCCESS )
401     {
402         /* We are dealing with a VOP */
403         p_pic = block_ChainGather( p_sys->p_frame );
404         p_pic->i_flags = p_sys->i_flags;
405         p_pic->i_pts = p_sys->i_interpolated_pts;
406         p_pic->i_dts = p_sys->i_interpolated_dts;
407
408         /* Reset context */
409         p_sys->p_frame = NULL;
410         p_sys->pp_last = &p_sys->p_frame;
411     }
412
413     return p_pic;
414 }
415
416 /* ParseVOL:
417  *  TODO:
418  *      - support aspect ratio
419  */
420 static int ParseVOL( decoder_t *p_dec, es_format_t *fmt,
421                      uint8_t *p_vol, int i_vol )
422 {
423     decoder_sys_t *p_sys = p_dec->p_sys;
424     int i_vo_type, i_vo_ver_id, i_ar, i_shape;
425     bs_t s;
426
427     for( ;; )
428     {
429         if( p_vol[0] == 0x00 && p_vol[1] == 0x00 && p_vol[2] == 0x01 &&
430             p_vol[3] >= 0x20 && p_vol[3] <= 0x2f ) break;
431
432         p_vol++; i_vol--;
433         if( i_vol <= 4 ) return VLC_EGENERIC;
434     }
435
436     bs_init( &s, &p_vol[4], i_vol - 4 );
437
438     bs_skip( &s, 1 );   /* random access */
439     i_vo_type = bs_read( &s, 8 );
440     if( bs_read1( &s ) )
441     {
442         i_vo_ver_id = bs_read( &s, 4 );
443         bs_skip( &s, 3 );
444     }
445     else
446     {
447         i_vo_ver_id = 1;
448     }
449     i_ar = bs_read( &s, 4 );
450     if( i_ar == 0xf )
451     {
452         int i_ar_width, i_ar_height;
453
454         i_ar_width = bs_read( &s, 8 );
455         i_ar_height= bs_read( &s, 8 );
456     }
457     if( bs_read1( &s ) )
458     {
459         int i_chroma_format;
460         int i_low_delay;
461
462         /* vol control parameter */
463         i_chroma_format = bs_read( &s, 2 );
464         i_low_delay = bs_read1( &s );
465
466         if( bs_read1( &s ) )
467         {
468             bs_skip( &s, 16 );
469             bs_skip( &s, 16 );
470             bs_skip( &s, 16 );
471             bs_skip( &s, 3 );
472             bs_skip( &s, 11 );
473             bs_skip( &s, 1 );
474             bs_skip( &s, 16 );
475         }
476     }
477     /* shape 0->RECT, 1->BIN, 2->BIN_ONLY, 3->GRAY */
478     i_shape = bs_read( &s, 2 );
479     if( i_shape == 3 && i_vo_ver_id != 1 )
480     {
481         bs_skip( &s, 4 );
482     }
483
484     if( !bs_read1( &s ) ) return VLC_EGENERIC; /* Marker */
485
486     p_sys->i_fps_num = bs_read( &s, 16 ); /* Time increment resolution*/
487     if( !p_sys->i_fps_num ) p_sys->i_fps_num = 1;
488
489     if( !bs_read1( &s ) ) return VLC_EGENERIC; /* Marker */
490
491     if( bs_read1( &s ) )
492     {
493         int i_time_increment_bits = vlc_log2( p_sys->i_fps_num - 1 ) + 1;
494
495         if( i_time_increment_bits < 1 ) i_time_increment_bits = 1;
496
497         p_sys->i_fps_den = bs_read( &s, i_time_increment_bits );
498     }
499     if( i_shape == 0 )
500     {
501         bs_skip( &s, 1 );
502         fmt->video.i_width = bs_read( &s, 13 );
503         bs_skip( &s, 1 );
504         fmt->video.i_height= bs_read( &s, 13 );
505         bs_skip( &s, 1 );
506     }
507
508     return VLC_SUCCESS;
509 }
510
511 static int ParseVOP( decoder_t *p_dec, block_t *p_vop )
512 {
513     decoder_sys_t *p_sys = p_dec->p_sys;
514     int64_t i_time_increment, i_time_ref;
515     int i_modulo_time_base = 0, i_time_increment_bits;
516     bs_t s;
517
518     bs_init( &s, &p_vop->p_buffer[4], p_vop->i_buffer - 4 );
519
520     switch( bs_read( &s, 2 ) )
521     {
522     case 0:
523         p_sys->i_flags = BLOCK_FLAG_TYPE_I;
524         break;
525     case 1:
526         p_sys->i_flags = BLOCK_FLAG_TYPE_P;
527         break;
528     case 2:
529         p_sys->i_flags = BLOCK_FLAG_TYPE_B;
530         p_sys->b_frame = true;
531         break;
532     case 3: /* gni ? */
533         p_sys->i_flags = BLOCK_FLAG_TYPE_PB;
534         break;
535     }
536
537     while( bs_read( &s, 1 ) ) i_modulo_time_base++;
538     if( !bs_read1( &s ) ) return VLC_EGENERIC; /* Marker */
539
540     /* VOP time increment */
541     i_time_increment_bits = vlc_log2(p_dec->p_sys->i_fps_num - 1) + 1;
542     if( i_time_increment_bits < 1 ) i_time_increment_bits = 1;
543     i_time_increment = bs_read( &s, i_time_increment_bits );
544
545     /* Interpolate PTS/DTS */
546     if( !(p_sys->i_flags & BLOCK_FLAG_TYPE_B) )
547     {
548         p_sys->i_last_time_ref = p_sys->i_time_ref;
549         p_sys->i_time_ref +=
550             (i_modulo_time_base * p_dec->p_sys->i_fps_num);
551         i_time_ref = p_sys->i_time_ref;
552     }
553     else
554     {
555         i_time_ref = p_sys->i_last_time_ref +
556             (i_modulo_time_base * p_dec->p_sys->i_fps_num);
557     }
558
559 #if 0
560     msg_Err( p_dec, "interp pts/dts (%lli,%lli), pts/dts (%lli,%lli)",
561              p_sys->i_interpolated_pts, p_sys->i_interpolated_dts,
562              p_vop->i_pts, p_vop->i_dts );
563 #endif
564
565     if( p_dec->p_sys->i_fps_num < 5 && /* Work-around buggy streams */
566         p_dec->fmt_in.video.i_frame_rate > 0 &&
567         p_dec->fmt_in.video.i_frame_rate_base > 0 )
568     {
569         p_sys->i_interpolated_pts += INT64_C(1000000) *
570         p_dec->fmt_in.video.i_frame_rate_base /
571         p_dec->fmt_in.video.i_frame_rate;
572     }
573     else if( p_dec->p_sys->i_fps_num )
574         p_sys->i_interpolated_pts +=
575             ( INT64_C(1000000) * (i_time_ref + i_time_increment -
576               p_sys->i_last_time - p_sys->i_last_timeincr) /
577               p_dec->p_sys->i_fps_num );
578
579     p_sys->i_last_time = i_time_ref;
580     p_sys->i_last_timeincr = i_time_increment;
581
582     /* Correct interpolated dts when we receive a new pts/dts */
583     if( p_vop->i_pts > 0 )
584         p_sys->i_interpolated_pts = p_vop->i_pts;
585     if( p_vop->i_dts > 0 )
586         p_sys->i_interpolated_dts = p_vop->i_dts;
587
588     if( (p_sys->i_flags & BLOCK_FLAG_TYPE_B) || !p_sys->b_frame )
589     {
590         /* Trivial case (DTS == PTS) */
591
592         p_sys->i_interpolated_dts = p_sys->i_interpolated_pts;
593
594         if( p_vop->i_pts > 0 )
595             p_sys->i_interpolated_dts = p_vop->i_pts;
596         if( p_vop->i_dts > 0 )
597             p_sys->i_interpolated_dts = p_vop->i_dts;
598
599         p_sys->i_interpolated_pts = p_sys->i_interpolated_dts;
600     }
601     else
602     {
603         if( p_sys->i_last_ref_pts > 0 )
604             p_sys->i_interpolated_dts = p_sys->i_last_ref_pts;
605
606         p_sys->i_last_ref_pts = p_sys->i_interpolated_pts;
607     }
608
609     return VLC_SUCCESS;
610 }
611
612 /* look at ffmpeg av_log2 ;) */
613 static int vlc_log2( unsigned int v )
614 {
615     int n = 0;
616     static const int vlc_log2_table[16] =
617     {
618         0,0,1,1,2,2,2,2, 3,3,3,3,3,3,3,3
619     };
620
621     if( v&0xffff0000 )
622     {
623         v >>= 16;
624         n += 16;
625     }
626     if( v&0xff00 )
627     {
628         v >>= 8;
629         n += 8;
630     }
631     if( v&0xf0 )
632     {
633         v >>= 4;
634         n += 4;
635     }
636     n += vlc_log2_table[v];
637
638     return n;
639 }