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 
39  int buf_size)
40 {
41  int i, j;
42  uint32_t state;
43  ParseContext *pc = &h->parse_context;
44  int next_avc= h->is_avc ? 0 : buf_size;
45 
46 // mb_addr= pc->mb_addr - 1;
47  state = pc->state;
48  if (state > 13)
49  state = 7;
50 
51  if (h->is_avc && !h->nal_length_size)
52  av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
53 
54  for (i = 0; i < buf_size; i++) {
55  if (i >= next_avc) {
56  int nalsize = 0;
57  i = next_avc;
58  for (j = 0; j < h->nal_length_size; j++)
59  nalsize = (nalsize << 8) | buf[i++];
60  if (nalsize <= 0 || nalsize > buf_size - i) {
61  av_log(h->avctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
62  return buf_size;
63  }
64  next_avc = i + nalsize;
65  state = 5;
66  }
67 
68  if (state == 7) {
69  i += h->h264dsp.startcode_find_candidate(buf + i, next_avc - i);
70  if (i < next_avc)
71  state = 2;
72  } else if (state <= 2) {
73  if (buf[i] == 1)
74  state ^= 5; // 2->7, 1->4, 0->5
75  else if (buf[i])
76  state = 7;
77  else
78  state >>= 1; // 2->1, 1->0, 0->0
79  } else if (state <= 5) {
80  int nalu_type = buf[i] & 0x1F;
81  if (nalu_type == NAL_SEI || nalu_type == NAL_SPS ||
82  nalu_type == NAL_PPS || nalu_type == NAL_AUD) {
83  if (pc->frame_start_found) {
84  i++;
85  goto found;
86  }
87  } else if (nalu_type == NAL_SLICE || nalu_type == NAL_DPA ||
88  nalu_type == NAL_IDR_SLICE) {
89  state += 8;
90  continue;
91  }
92  state = 7;
93  } else {
94  h->parse_history[h->parse_history_count++]= buf[i];
95  if (h->parse_history_count>5) {
96  unsigned int mb, last_mb= h->parse_last_mb;
97  GetBitContext gb;
98 
100  h->parse_history_count=0;
101  mb= get_ue_golomb_long(&gb);
102  h->parse_last_mb= mb;
103  if (pc->frame_start_found) {
104  if (mb <= last_mb)
105  goto found;
106  } else
107  pc->frame_start_found = 1;
108  state = 7;
109  }
110  }
111  }
112  pc->state = state;
113  if (h->is_avc)
114  return next_avc;
115  return END_NOT_FOUND;
116 
117 found:
118  pc->state = 7;
119  pc->frame_start_found = 0;
120  if (h->is_avc)
121  return next_avc;
122  return i - (state & 5) - 5 * (state > 7);
123 }
124 
126 {
127  H264Context *h = s->priv_data;
128 
129  h->slice_type_nos = s->pict_type & 3;
130 
132  get_ue_golomb(&h->gb); // redundant_pic_count
133 
134  if (ff_set_ref_count(h) < 0)
135  return AVERROR_INVALIDDATA;
136 
137  if (h->slice_type_nos != AV_PICTURE_TYPE_I) {
138  int list;
139  for (list = 0; list < h->list_count; list++) {
140  if (get_bits1(&h->gb)) {
141  int index;
142  for (index = 0; ; index++) {
143  unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(&h->gb);
144 
145  if (reordering_of_pic_nums_idc < 3)
146  get_ue_golomb(&h->gb);
147  else if (reordering_of_pic_nums_idc > 3) {
149  "illegal reordering_of_pic_nums_idc %d\n",
150  reordering_of_pic_nums_idc);
151  return AVERROR_INVALIDDATA;
152  } else
153  break;
154 
155  if (index >= h->ref_count[list]) {
157  "reference count %d overflow\n", index);
158  return AVERROR_INVALIDDATA;
159  }
160  }
161  }
162  }
163  }
164 
168 
169  if (get_bits1(&h->gb)) { // adaptive_ref_pic_marking_mode_flag
170  int i;
171  for (i = 0; i < MAX_MMCO_COUNT; i++) {
172  MMCOOpcode opcode = get_ue_golomb_31(&h->gb);
173  if (opcode > (unsigned) MMCO_LONG) {
175  "illegal memory management control operation %d\n",
176  opcode);
177  return AVERROR_INVALIDDATA;
178  }
179  if (opcode == MMCO_END)
180  return 0;
181  else if (opcode == MMCO_RESET)
182  return 1;
183 
184  if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
185  get_ue_golomb(&h->gb);
186  if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
187  opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
188  get_ue_golomb_31(&h->gb);
189  }
190  }
191 
192  return 0;
193 }
194 
195 /**
196  * Parse NAL units of found picture and decode some basic information.
197  *
198  * @param s parser context.
199  * @param avctx codec context.
200  * @param buf buffer with field/frame data.
201  * @param buf_size size of the buffer.
202  */
204  AVCodecContext *avctx,
205  const uint8_t * const buf, int buf_size)
206 {
207  H264Context *h = s->priv_data;
208  int buf_index, next_avc;
209  unsigned int pps_id;
210  unsigned int slice_type;
211  int state = -1, got_reset = 0;
212  const uint8_t *ptr;
213  int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
214  int field_poc[2];
215 
216  /* set some sane default values */
218  s->key_frame = 0;
220 
221  h->avctx = avctx;
224 
225  if (!buf_size)
226  return 0;
227 
228  buf_index = 0;
229  next_avc = h->is_avc ? 0 : buf_size;
230  for (;;) {
231  int src_length, dst_length, consumed, nalsize = 0;
232 
233  if (buf_index >= next_avc) {
234  nalsize = get_avc_nalsize(h, buf, buf_size, &buf_index);
235  if (nalsize < 0)
236  break;
237  next_avc = buf_index + nalsize;
238  } else {
239  buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
240  if (buf_index >= buf_size)
241  break;
242  if (buf_index >= next_avc)
243  continue;
244  }
245  src_length = next_avc - buf_index;
246 
247  state = buf[buf_index];
248  switch (state & 0x1f) {
249  case NAL_SLICE:
250  case NAL_IDR_SLICE:
251  // Do not walk the whole buffer just to decode slice header
252  if ((state & 0x1f) == NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
253  /* IDR or disposable slice
254  * No need to decode many bytes because MMCOs shall not be present. */
255  if (src_length > 60)
256  src_length = 60;
257  } else {
258  /* To decode up to MMCOs */
259  if (src_length > 1000)
260  src_length = 1000;
261  }
262  break;
263  }
264  ptr = ff_h264_decode_nal(h, buf + buf_index, &dst_length,
265  &consumed, src_length);
266  if (!ptr || dst_length < 0)
267  break;
268 
269  buf_index += consumed;
270 
271  init_get_bits(&h->gb, ptr, 8 * dst_length);
272  switch (h->nal_unit_type) {
273  case NAL_SPS:
275  break;
276  case NAL_PPS:
278  break;
279  case NAL_SEI:
281  break;
282  case NAL_IDR_SLICE:
283  s->key_frame = 1;
284 
285  h->prev_frame_num = 0;
286  h->prev_frame_num_offset = 0;
287  h->prev_poc_msb =
288  h->prev_poc_lsb = 0;
289  /* fall through */
290  case NAL_SLICE:
291  get_ue_golomb_long(&h->gb); // skip first_mb_in_slice
292  slice_type = get_ue_golomb_31(&h->gb);
293  s->pict_type = golomb_to_pict_type[slice_type % 5];
294  if (h->sei_recovery_frame_cnt >= 0) {
295  /* key frame, since recovery_frame_cnt is set */
296  s->key_frame = 1;
297  }
298  pps_id = get_ue_golomb(&h->gb);
299  if (pps_id >= MAX_PPS_COUNT) {
301  "pps_id %u out of range\n", pps_id);
302  return -1;
303  }
304  if (!h->pps_buffers[pps_id]) {
306  "non-existing PPS %u referenced\n", pps_id);
307  return -1;
308  }
309  h->pps = *h->pps_buffers[pps_id];
310  if (!h->sps_buffers[h->pps.sps_id]) {
312  "non-existing SPS %u referenced\n", h->pps.sps_id);
313  return -1;
314  }
315  h->sps = *h->sps_buffers[h->pps.sps_id];
317 
318  if(h->sps.ref_frame_count <= 1 && h->pps.ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
319  s->key_frame = 1;
320 
321  avctx->profile = ff_h264_get_profile(&h->sps);
322  avctx->level = h->sps.level_idc;
323 
324  if (h->sps.frame_mbs_only_flag) {
326  } else {
327  if (get_bits1(&h->gb)) { // field_pic_flag
328  h->picture_structure = PICT_TOP_FIELD + get_bits1(&h->gb); // bottom_field_flag
329  } else {
331  }
332  }
333 
334  if (h->nal_unit_type == NAL_IDR_SLICE)
335  get_ue_golomb(&h->gb); /* idr_pic_id */
336  if (h->sps.poc_type == 0) {
337  h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb);
338 
339  if (h->pps.pic_order_present == 1 &&
342  }
343 
344  if (h->sps.poc_type == 1 &&
346  h->delta_poc[0] = get_se_golomb(&h->gb);
347 
348  if (h->pps.pic_order_present == 1 &&
350  h->delta_poc[1] = get_se_golomb(&h->gb);
351  }
352 
353  /* Decode POC of this picture.
354  * The prev_ values needed for decoding POC of the next picture are not set here. */
355  field_poc[0] = field_poc[1] = INT_MAX;
356  ff_init_poc(h, field_poc, &s->output_picture_number);
357 
358  /* Continue parsing to check if MMCO_RESET is present.
359  * FIXME: MMCO_RESET could appear in non-first slice.
360  * Maybe, we should parse all undisposable non-IDR slice of this
361  * picture until encountering MMCO_RESET in a slice of it. */
362  if (h->nal_ref_idc && h->nal_unit_type != NAL_IDR_SLICE) {
363  got_reset = scan_mmco_reset(s);
364  if (got_reset < 0)
365  return got_reset;
366  }
367 
368  /* Set up the prev_ values for decoding POC of the next picture. */
369  h->prev_frame_num = got_reset ? 0 : h->frame_num;
370  h->prev_frame_num_offset = got_reset ? 0 : h->frame_num_offset;
371  if (h->nal_ref_idc != 0) {
372  if (!got_reset) {
373  h->prev_poc_msb = h->poc_msb;
374  h->prev_poc_lsb = h->poc_lsb;
375  } else {
376  h->prev_poc_msb = 0;
377  h->prev_poc_lsb =
378  h->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
379  }
380  }
381 
382  if (h->sps.pic_struct_present_flag) {
383  switch (h->sei_pic_struct) {
386  s->repeat_pict = 0;
387  break;
391  s->repeat_pict = 1;
392  break;
395  s->repeat_pict = 2;
396  break;
398  s->repeat_pict = 3;
399  break;
401  s->repeat_pict = 5;
402  break;
403  default:
404  s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
405  break;
406  }
407  } else {
408  s->repeat_pict = h->picture_structure == PICT_FRAME ? 1 : 0;
409  }
410 
411  if (h->picture_structure == PICT_FRAME) {
413  if (h->sps.pic_struct_present_flag) {
414  switch (h->sei_pic_struct) {
418  break;
422  break;
423  default:
425  break;
426  }
427  } else {
428  if (field_poc[0] < field_poc[1])
430  else if (field_poc[0] > field_poc[1])
432  else
434  }
435  } else {
438  else
441  }
442 
443  return 0; /* no need to evaluate the rest */
444  }
445  }
446  if (q264)
447  return 0;
448  /* didn't find a picture! */
449  av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
450  return -1;
451 }
452 
454  AVCodecContext *avctx,
455  const uint8_t **poutbuf, int *poutbuf_size,
456  const uint8_t *buf, int buf_size)
457 {
458  H264Context *h = s->priv_data;
459  ParseContext *pc = &h->parse_context;
460  int next;
461 
462  if (!h->got_first) {
463  h->got_first = 1;
464  if (avctx->extradata_size) {
465  h->avctx = avctx;
466  // must be done like in decoder, otherwise opening the parser,
467  // letting it create extradata and then closing and opening again
468  // will cause has_b_frames to be always set.
469  // Note that estimate_timings_from_pts does exactly this.
470  if (!avctx->has_b_frames)
471  h->low_delay = 1;
473  }
474  }
475 
477  next = buf_size;
478  } else {
479  next = h264_find_frame_end(h, buf, buf_size);
480 
481  if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
482  *poutbuf = NULL;
483  *poutbuf_size = 0;
484  return buf_size;
485  }
486 
487  if (next < 0 && next != END_NOT_FOUND) {
488  av_assert1(pc->last_index + next >= 0);
489  h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); // update state
490  }
491  }
492 
493  parse_nal_units(s, avctx, buf, buf_size);
494 
495  if (avctx->framerate.num)
496  avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
497  if (h->sei_cpb_removal_delay >= 0) {
498  s->dts_sync_point = h->sei_buffering_period_present;
499  s->dts_ref_dts_delta = h->sei_cpb_removal_delay;
500  s->pts_dts_delta = h->sei_dpb_output_delay;
501  } else {
502  s->dts_sync_point = INT_MIN;
503  s->dts_ref_dts_delta = INT_MIN;
504  s->pts_dts_delta = INT_MIN;
505  }
506 
507  if (s->flags & PARSER_FLAG_ONCE) {
508  s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
509  }
510 
511  *poutbuf = buf;
512  *poutbuf_size = buf_size;
513  return next;
514 }
515 
516 static int h264_split(AVCodecContext *avctx,
517  const uint8_t *buf, int buf_size)
518 {
519  int i;
520  uint32_t state = -1;
521  int has_sps = 0;
522 
523  for (i = 0; i <= buf_size; i++) {
524  if ((state & 0xFFFFFF1F) == 0x107)
525  has_sps = 1;
526  /* if ((state&0xFFFFFF1F) == 0x101 ||
527  * (state&0xFFFFFF1F) == 0x102 ||
528  * (state&0xFFFFFF1F) == 0x105) {
529  * }
530  */
531  if ((state & 0xFFFFFF00) == 0x100 && (state & 0xFFFFFF1F) != 0x107 &&
532  (state & 0xFFFFFF1F) != 0x108 && (state & 0xFFFFFF1F) != 0x109) {
533  if (has_sps) {
534  while (i > 4 && buf[i - 5] == 0)
535  i--;
536  return i - 4;
537  }
538  }
539  if (i < buf_size)
540  state = (state << 8) | buf[i];
541  }
542  return 0;
543 }
544 
546 {
547  H264Context *h = s->priv_data;
548  ParseContext *pc = &h->parse_context;
549 
550  av_free(pc->buffer);
552 }
553 
555 {
556  H264Context *h = s->priv_data;
557  h->thread_context[0] = h;
558  h->slice_context_count = 1;
559  ff_h264dsp_init(&h->h264dsp, 8, 1);
560  return 0;
561 }
562 
565  .priv_data_size = sizeof(H264Context),
566  .parser_init = init,
567  .parser_parse = h264_parse,
568  .parser_close = close,
569  .split = h264_split,
570 };