FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
h264_parser.c
Go to the documentation of this file.
1 /*
2  * H.26L/H.264/AVC/JVT/14496-10/... parser
3  * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
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  * H.264 / AVC / MPEG4 part10 parser.
25  * @author Michael Niedermayer <michaelni@gmx.at>
26  */
27 
28 #define UNCHECKED_BITSTREAM_READER 1
29 
30 #include "libavutil/attributes.h"
31 #include "parser.h"
32 #include "h264data.h"
33 #include "golomb.h"
34 #include "internal.h"
35 #include "mpegutils.h"
36 
37 typedef struct H264ParseContext {
40  int got_first;
42 
43 
45  int buf_size)
46 {
47  H264Context *h = &p->h;
48  int i, j;
49  uint32_t state;
50  ParseContext *pc = &p->pc;
51 
52  int next_avc= h->is_avc ? 0 : buf_size;
53 // mb_addr= pc->mb_addr - 1;
54  state = pc->state;
55  if (state > 13)
56  state = 7;
57 
58  if (h->is_avc && !h->nal_length_size)
59  av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
60 
61  for (i = 0; i < buf_size; i++) {
62  if (i >= next_avc) {
63  int nalsize = 0;
64  i = next_avc;
65  for (j = 0; j < h->nal_length_size; j++)
66  nalsize = (nalsize << 8) | buf[i++];
67  if (nalsize <= 0 || nalsize > buf_size - i) {
68  av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
69  return buf_size;
70  }
71  next_avc = i + nalsize;
72  state = 5;
73  }
74 
75  if (state == 7) {
76  i += h->h264dsp.startcode_find_candidate(buf + i, next_avc - i);
77  if (i < next_avc)
78  state = 2;
79  } else if (state <= 2) {
80  if (buf[i] == 1)
81  state ^= 5; // 2->7, 1->4, 0->5
82  else if (buf[i])
83  state = 7;
84  else
85  state >>= 1; // 2->1, 1->0, 0->0
86  } else if (state <= 5) {
87  int nalu_type = buf[i] & 0x1F;
88  if (nalu_type == NAL_SEI || nalu_type == NAL_SPS ||
89  nalu_type == NAL_PPS || nalu_type == NAL_AUD) {
90  if (pc->frame_start_found) {
91  i++;
92  goto found;
93  }
94  } else if (nalu_type == NAL_SLICE || nalu_type == NAL_DPA ||
95  nalu_type == NAL_IDR_SLICE) {
96  state += 8;
97  continue;
98  }
99  state = 7;
100  } else {
101  h->parse_history[h->parse_history_count++]= buf[i];
102  if (h->parse_history_count>5) {
103  unsigned int mb, last_mb= h->parse_last_mb;
104  GetBitContext gb;
105 
107  h->parse_history_count=0;
108  mb= get_ue_golomb_long(&gb);
109  h->parse_last_mb= mb;
110  if (pc->frame_start_found) {
111  if (mb <= last_mb)
112  goto found;
113  } else
114  pc->frame_start_found = 1;
115  state = 7;
116  }
117  }
118  }
119  pc->state = state;
120  if (h->is_avc)
121  return next_avc;
122  return END_NOT_FOUND;
123 
124 found:
125  pc->state = 7;
126  pc->frame_start_found = 0;
127  if (h->is_avc)
128  return next_avc;
129  return i - (state & 5) - 5 * (state > 7);
130 }
131 
133 {
134  H264ParseContext *p = s->priv_data;
135  H264Context *h = &p->h;
136 
137  h->slice_type_nos = s->pict_type & 3;
138 
140  get_ue_golomb(&h->gb); // redundant_pic_count
141 
142  if (ff_set_ref_count(h) < 0)
143  return AVERROR_INVALIDDATA;
144 
145  if (h->slice_type_nos != AV_PICTURE_TYPE_I) {
146  int list;
147  for (list = 0; list < h->list_count; list++) {
148  if (get_bits1(&h->gb)) {
149  int index;
150  for (index = 0; ; index++) {
151  unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(&h->gb);
152 
153  if (reordering_of_pic_nums_idc < 3)
154  get_ue_golomb(&h->gb);
155  else if (reordering_of_pic_nums_idc > 3) {
157  "illegal reordering_of_pic_nums_idc %d\n",
158  reordering_of_pic_nums_idc);
159  return AVERROR_INVALIDDATA;
160  } else
161  break;
162 
163  if (index >= h->ref_count[list]) {
165  "reference count %d overflow\n", index);
166  return AVERROR_INVALIDDATA;
167  }
168  }
169  }
170  }
171  }
172 
176 
177  if (get_bits1(&h->gb)) { // adaptive_ref_pic_marking_mode_flag
178  int i;
179  for (i = 0; i < MAX_MMCO_COUNT; i++) {
180  MMCOOpcode opcode = get_ue_golomb_31(&h->gb);
181  if (opcode > (unsigned) MMCO_LONG) {
183  "illegal memory management control operation %d\n",
184  opcode);
185  return AVERROR_INVALIDDATA;
186  }
187  if (opcode == MMCO_END)
188  return 0;
189  else if (opcode == MMCO_RESET)
190  return 1;
191 
192  if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
193  get_ue_golomb(&h->gb);
194  if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
195  opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
196  get_ue_golomb_31(&h->gb);
197  }
198  }
199 
200  return 0;
201 }
202 
203 /**
204  * Parse NAL units of found picture and decode some basic information.
205  *
206  * @param s parser context.
207  * @param avctx codec context.
208  * @param buf buffer with field/frame data.
209  * @param buf_size size of the buffer.
210  */
212  AVCodecContext *avctx,
213  const uint8_t * const buf, int buf_size)
214 {
215  H264ParseContext *p = s->priv_data;
216  H264Context *h = &p->h;
217  int buf_index, next_avc;
218  unsigned int pps_id;
219  unsigned int slice_type;
220  int state = -1, got_reset = 0;
221  const uint8_t *ptr;
222  int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
223  int field_poc[2];
224 
225  /* set some sane default values */
227  s->key_frame = 0;
229 
230  h->avctx = avctx;
233 
234  if (!buf_size)
235  return 0;
236 
237  buf_index = 0;
238  next_avc = h->is_avc ? 0 : buf_size;
239  for (;;) {
240  int src_length, dst_length, consumed, nalsize = 0;
241 
242  if (buf_index >= next_avc) {
243  nalsize = get_avc_nalsize(h, buf, buf_size, &buf_index);
244  if (nalsize < 0)
245  break;
246  next_avc = buf_index + nalsize;
247  } else {
248  buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
249  if (buf_index >= buf_size)
250  break;
251  if (buf_index >= next_avc)
252  continue;
253  }
254  src_length = next_avc - buf_index;
255 
256  state = buf[buf_index];
257  switch (state & 0x1f) {
258  case NAL_SLICE:
259  case NAL_IDR_SLICE:
260  // Do not walk the whole buffer just to decode slice header
261  if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
262  /* IDR or disposable slice
263  * No need to decode many bytes because MMCOs shall not be present. */
264  if (src_length > 60)
265  src_length = 60;
266  } else {
267  /* To decode up to MMCOs */
268  if (src_length > 1000)
269  src_length = 1000;
270  }
271  break;
272  }
273  ptr = ff_h264_decode_nal(h, buf + buf_index, &dst_length,
274  &consumed, src_length);
275  if (!ptr || dst_length < 0)
276  break;
277 
278  buf_index += consumed;
279 
280  init_get_bits(&h->gb, ptr, 8 * dst_length);
281  switch (h->nal_unit_type) {
282  case NAL_SPS:
284  break;
285  case NAL_PPS:
287  break;
288  case NAL_SEI:
290  break;
291  case NAL_IDR_SLICE:
292  s->key_frame = 1;
293 
294  h->prev_frame_num = 0;
295  h->prev_frame_num_offset = 0;
296  h->prev_poc_msb =
297  h->prev_poc_lsb = 0;
298  /* fall through */
299  case NAL_SLICE:
300  get_ue_golomb_long(&h->gb); // skip first_mb_in_slice
301  slice_type = get_ue_golomb_31(&h->gb);
302  s->pict_type = golomb_to_pict_type[slice_type % 5];
303  if (h->sei_recovery_frame_cnt >= 0) {
304  /* key frame, since recovery_frame_cnt is set */
305  s->key_frame = 1;
306  }
307  pps_id = get_ue_golomb(&h->gb);
308  if (pps_id >= MAX_PPS_COUNT) {
310  "pps_id %u out of range\n", pps_id);
311  return -1;
312  }
313  if (!h->pps_buffers[pps_id]) {
315  "non-existing PPS %u referenced\n", pps_id);
316  return -1;
317  }
318  h->pps = *h->pps_buffers[pps_id];
319  if (!h->sps_buffers[h->pps.sps_id]) {
321  "non-existing SPS %u referenced\n", h->pps.sps_id);
322  return -1;
323  }
324  h->sps = *h->sps_buffers[h->pps.sps_id];
326 
327  if(h->sps.ref_frame_count <= 1 && h->pps.ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
328  s->key_frame = 1;
329 
330  avctx->profile = ff_h264_get_profile(&h->sps);
331  avctx->level = h->sps.level_idc;
332 
333  if (h->sps.frame_mbs_only_flag) {
335  } else {
336  if (get_bits1(&h->gb)) { // field_pic_flag
337  h->picture_structure = PICT_TOP_FIELD + get_bits1(&h->gb); // bottom_field_flag
338  } else {
340  }
341  }
342 
343  if (h->nal_unit_type == NAL_IDR_SLICE)
344  get_ue_golomb(&h->gb); /* idr_pic_id */
345  if (h->sps.poc_type == 0) {
346  h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb);
347 
348  if (h->pps.pic_order_present == 1 &&
351  }
352 
353  if (h->sps.poc_type == 1 &&
355  h->delta_poc[0] = get_se_golomb(&h->gb);
356 
357  if (h->pps.pic_order_present == 1 &&
359  h->delta_poc[1] = get_se_golomb(&h->gb);
360  }
361 
362  /* Decode POC of this picture.
363  * The prev_ values needed for decoding POC of the next picture are not set here. */
364  field_poc[0] = field_poc[1] = INT_MAX;
365  ff_init_poc(h, field_poc, &s->output_picture_number);
366 
367  /* Continue parsing to check if MMCO_RESET is present.
368  * FIXME: MMCO_RESET could appear in non-first slice.
369  * Maybe, we should parse all undisposable non-IDR slice of this
370  * picture until encountering MMCO_RESET in a slice of it. */
371  if (h->nal_ref_idc && h->nal_unit_type != NAL_IDR_SLICE) {
372  got_reset = scan_mmco_reset(s);
373  if (got_reset < 0)
374  return got_reset;
375  }
376 
377  /* Set up the prev_ values for decoding POC of the next picture. */
378  h->prev_frame_num = got_reset ? 0 : h->frame_num;
379  h->prev_frame_num_offset = got_reset ? 0 : h->frame_num_offset;
380  if (h->nal_ref_idc != 0) {
381  if (!got_reset) {
382  h->prev_poc_msb = h->poc_msb;
383  h->prev_poc_lsb = h->poc_lsb;
384  } else {
385  h->prev_poc_msb = 0;
386  h->prev_poc_lsb =
387  h->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
388  }
389  }
390 
391  if (h->sps.pic_struct_present_flag) {
392  switch (h->sei_pic_struct) {
395  s->repeat_pict = 0;
396  break;
400  s->repeat_pict = 1;
401  break;
404  s->repeat_pict = 2;
405  break;
407  s->repeat_pict = 3;
408  break;
410  s->repeat_pict = 5;
411  break;
412  default:
413  s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
414  break;
415  }
416  } else {
417  s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
418  }
419 
420  if (h->picture_structure == PICT_FRAME) {
422  if (h->sps.pic_struct_present_flag) {
423  switch (h->sei_pic_struct) {
427  break;
431  break;
432  default:
434  break;
435  }
436  } else {
437  if (field_poc[0] < field_poc[1])
439  else if (field_poc[0] > field_poc[1])
441  else
443  }
444  } else {
447  else
450  }
451 
452  return 0; /* no need to evaluate the rest */
453  }
454  }
455  if (q264)
456  return 0;
457  /* didn't find a picture! */
458  av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
459  return -1;
460 }
461 
463  AVCodecContext *avctx,
464  const uint8_t **poutbuf, int *poutbuf_size,
465  const uint8_t *buf, int buf_size)
466 {
467  H264ParseContext *p = s->priv_data;
468  H264Context *h = &p->h;
469  ParseContext *pc = &p->pc;
470  int next;
471 
472  if (!p->got_first) {
473  p->got_first = 1;
474  if (avctx->extradata_size) {
475  h->avctx = avctx;
476  // must be done like in decoder, otherwise opening the parser,
477  // letting it create extradata and then closing and opening again
478  // will cause has_b_frames to be always set.
479  // Note that estimate_timings_from_pts does exactly this.
480  if (!avctx->has_b_frames)
481  h->low_delay = 1;
483  }
484  }
485 
487  next = buf_size;
488  } else {
489  next = h264_find_frame_end(p, buf, buf_size);
490 
491  if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
492  *poutbuf = NULL;
493  *poutbuf_size = 0;
494  return buf_size;
495  }
496 
497  if (next < 0 && next != END_NOT_FOUND) {
498  av_assert1(pc->last_index + next >= 0);
499  h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next); // update state
500  }
501  }
502 
503  parse_nal_units(s, avctx, buf, buf_size);
504 
505  if (avctx->framerate.num)
506  avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
507  if (h->sei_cpb_removal_delay >= 0) {
508  s->dts_sync_point = h->sei_buffering_period_present;
509  s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
510  s->pts_dts_delta = h->sei_dpb_output_delay;
511  } else {
512  s->dts_sync_point = INT_MIN;
513  s->dts_ref_dts_delta = INT_MIN;
514  s->pts_dts_delta = INT_MIN;
515  }
516 
517  if (s->flags & PARSER_FLAG_ONCE) {
518  s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
519  }
520 
521  *poutbuf = buf;
522  *poutbuf_size = buf_size;
523  return next;
524 }
525 
526 static int h264_split(AVCodecContext *avctx,
527  const uint8_t *buf, int buf_size)
528 {
529  int i;
530  uint32_t state = -1;
531  int has_sps = 0;
532  int has_pps = 0;
533 
534  for (i = 0; i <= buf_size; i++) {
535  if ((state & 0xFFFFFF1F) == 0x107)
536  has_sps = 1;
537  if ((state & 0xFFFFFF1F) == 0x108)
538  has_pps = 1;
539  /* if ((state&0xFFFFFF1F) == 0x101 ||
540  * (state&0xFFFFFF1F) == 0x102 ||
541  * (state&0xFFFFFF1F) == 0x105) {
542  * }
543  */
544  if ((state & 0xFFFFFF00) == 0x100 && ((state & 0xFFFFFF1F) != 0x106 || has_pps) &&
545  (state & 0xFFFFFF1F) != 0x107 && (state & 0xFFFFFF1F) != 0x108 &&
546  (state & 0xFFFFFF1F) != 0x109 && (state & 0xFFFFFF1F) != 0x10d &&
547  (state & 0xFFFFFF1F) != 0x10f) {
548  if (has_sps) {
549  while (i > 4 && buf[i - 5] == 0)
550  i--;
551  return i - 4;
552  }
553  }
554  if (i < buf_size)
555  state = (state << 8) | buf[i];
556  }
557  return 0;
558 }
559 
561 {
562  H264ParseContext *p = s->priv_data;
563  H264Context *h = &p->h;
564  ParseContext *pc = &p->pc;
565 
566  av_freep(&pc->buffer);
568 }
569 
571 {
572  H264ParseContext *p = s->priv_data;
573  H264Context *h = &p->h;
574  h->thread_context[0] = h;
575  h->slice_context_count = 1;
576  ff_h264dsp_init(&h->h264dsp, 8, 1);
577  return 0;
578 }
579 
582  .priv_data_size = sizeof(H264ParseContext),
583  .parser_init = init,
584  .parser_parse = h264_parse,
585  .parser_close = h264_close,
586  .split = h264_split,
587 };