git.sesse.net Git - ffmpeg/blob - libavcodec/h264_parser.c

   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 / MPEG-4 part10 parser.
  25  * @author Michael Niedermayer <michaelni@gmx.at>
  26  */
  27
  28 #define UNCHECKED_BITSTREAM_READER 1
  29
  30 #include <assert.h>
  31 #include <stdint.h>
  32
  33 #include "libavutil/avutil.h"
  34 #include "libavutil/error.h"
  35 #include "libavutil/log.h"
  36 #include "libavutil/mem.h"
  37 #include "libavutil/pixfmt.h"
  38
  39 #include "avcodec.h"
  40 #include "get_bits.h"
  41 #include "golomb.h"
  42 #include "h264.h"
  43 #include "h264_sei.h"
  44 #include "h264_ps.h"
  45 #include "h264data.h"
  46 #include "internal.h"
  47 #include "mpegutils.h"
  48 #include "parser.h"
  49
  50 typedef struct H264ParseContext {
  51     ParseContext pc;
  52     H264ParamSets ps;
  53     H264DSPContext h264dsp;
  54     H264POCContext poc;
  55     H264SEIContext sei;
  56     int is_avc;
  57     int nal_length_size;
  58     int got_first;
  59     int picture_structure;
  60     uint8_t parse_history[6];
  61     int parse_history_count;
  62     int parse_last_mb;
  63     int64_t reference_dts;
  64     int last_frame_num, last_picture_structure;
  65 } H264ParseContext;
  66
  67
  68 static int h264_find_frame_end(H264ParseContext *p, const uint8_t *buf,
  69                                int buf_size, void *logctx)
  70 {
  71     int i, j;
  72     uint32_t state;
  73     ParseContext *pc = &p->pc;
  74
  75     int next_avc = p->is_avc ? 0 : buf_size;
  76 //    mb_addr= pc->mb_addr - 1;
  77     state = pc->state;
  78     if (state > 13)
  79         state = 7;
  80
  81     if (p->is_avc && !p->nal_length_size)
  82         av_log(logctx, AV_LOG_ERROR, "AVC-parser: nal length size invalid\n");
  83
  84     for (i = 0; i < buf_size; i++) {
  85         if (i >= next_avc) {
  86             int nalsize = 0;
  87             i = next_avc;
  88             for (j = 0; j < p->nal_length_size; j++)
  89                 nalsize = (nalsize << 8) | buf[i++];
  90             if (nalsize <= 0 || nalsize > buf_size - i) {
  91                 av_log(logctx, AV_LOG_ERROR, "AVC-parser: nal size %d remaining %d\n", nalsize, buf_size - i);
  92                 return buf_size;
  93             }
  94             next_avc = i + nalsize;
  95             state    = 5;
  96         }
  97
  98         if (state == 7) {
  99             i += p->h264dsp.startcode_find_candidate(buf + i, next_avc - i);
 100             if (i < next_avc)
 101                 state = 2;
 102         } else if (state <= 2) {
 103             if (buf[i] == 1)
 104                 state ^= 5;            // 2->7, 1->4, 0->5
 105             else if (buf[i])
 106                 state = 7;
 107             else
 108                 state >>= 1;           // 2->1, 1->0, 0->0
 109         } else if (state <= 5) {
 110             int nalu_type = buf[i] & 0x1F;
 111             if (nalu_type == H264_NAL_SEI || nalu_type == H264_NAL_SPS ||
 112                 nalu_type == H264_NAL_PPS || nalu_type == H264_NAL_AUD) {
 113                 if (pc->frame_start_found) {
 114                     i++;
 115                     goto found;
 116                 }
 117             } else if (nalu_type == H264_NAL_SLICE || nalu_type == H264_NAL_DPA ||
 118                        nalu_type == H264_NAL_IDR_SLICE) {
 119                 state += 8;
 120                 continue;
 121             }
 122             state = 7;
 123         } else {
 124             unsigned int mb, last_mb = p->parse_last_mb;
 125             GetBitContext gb;
 126             p->parse_history[p->parse_history_count++] = buf[i];
 127
 128             init_get_bits(&gb, p->parse_history, 8*p->parse_history_count);
 129             mb= get_ue_golomb_long(&gb);
 130             if (get_bits_left(&gb) > 0 || p->parse_history_count > 5) {
 131                 p->parse_last_mb = mb;
 132                 if (pc->frame_start_found) {
 133                     if (mb <= last_mb) {
 134                         i -= p->parse_history_count - 1;
 135                         p->parse_history_count = 0;
 136                         goto found;
 137                     }
 138                 } else
 139                     pc->frame_start_found = 1;
 140                 p->parse_history_count = 0;
 141                 state = 7;
 142             }
 143         }
 144     }
 145     pc->state = state;
 146     if (p->is_avc)
 147         return next_avc;
 148     return END_NOT_FOUND;
 149
 150 found:
 151     pc->state             = 7;
 152     pc->frame_start_found = 0;
 153     if (p->is_avc)
 154         return next_avc;
 155     return i - (state & 5);
 156 }
 157
 158 static int scan_mmco_reset(AVCodecParserContext *s, GetBitContext *gb,
 159                            void *logctx)
 160 {
 161     H264PredWeightTable pwt;
 162     int slice_type_nos = s->pict_type & 3;
 163     H264ParseContext *p = s->priv_data;
 164     int list_count, ref_count[2];
 165
 166
 167     if (p->ps.pps->redundant_pic_cnt_present)
 168         get_ue_golomb(gb); // redundant_pic_count
 169
 170     if (slice_type_nos == AV_PICTURE_TYPE_B)
 171         get_bits1(gb); // direct_spatial_mv_pred
 172
 173     if (ff_h264_parse_ref_count(&list_count, ref_count, gb, p->ps.pps,
 174                                 slice_type_nos, p->picture_structure, logctx) < 0)
 175         return AVERROR_INVALIDDATA;
 176
 177     if (slice_type_nos != AV_PICTURE_TYPE_I) {
 178         int list;
 179         for (list = 0; list < list_count; list++) {
 180             if (get_bits1(gb)) {
 181                 int index;
 182                 for (index = 0; ; index++) {
 183                     unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(gb);
 184
 185                     if (reordering_of_pic_nums_idc < 3)
 186                         get_ue_golomb_long(gb);
 187                     else if (reordering_of_pic_nums_idc > 3) {
 188                         av_log(logctx, AV_LOG_ERROR,
 189                                "illegal reordering_of_pic_nums_idc %d\n",
 190                                reordering_of_pic_nums_idc);
 191                         return AVERROR_INVALIDDATA;
 192                     } else
 193                         break;
 194
 195                     if (index >= ref_count[list]) {
 196                         av_log(logctx, AV_LOG_ERROR,
 197                                "reference count %d overflow\n", index);
 198                         return AVERROR_INVALIDDATA;
 199                     }
 200                 }
 201             }
 202         }
 203     }
 204
 205     if ((p->ps.pps->weighted_pred && slice_type_nos == AV_PICTURE_TYPE_P) ||
 206         (p->ps.pps->weighted_bipred_idc == 1 && slice_type_nos == AV_PICTURE_TYPE_B))
 207         ff_h264_pred_weight_table(gb, p->ps.sps, ref_count, slice_type_nos,
 208                                   &pwt, p->picture_structure, logctx);
 209
 210     if (get_bits1(gb)) { // adaptive_ref_pic_marking_mode_flag
 211         int i;
 212         for (i = 0; i < MAX_MMCO_COUNT; i++) {
 213             MMCOOpcode opcode = get_ue_golomb_31(gb);
 214             if (opcode > (unsigned) MMCO_LONG) {
 215                 av_log(logctx, AV_LOG_ERROR,
 216                        "illegal memory management control operation %d\n",
 217                        opcode);
 218                 return AVERROR_INVALIDDATA;
 219             }
 220             if (opcode == MMCO_END)
 221                return 0;
 222             else if (opcode == MMCO_RESET)
 223                 return 1;
 224
 225             if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
 226                 get_ue_golomb_long(gb); // difference_of_pic_nums_minus1
 227             if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
 228                 opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
 229                 get_ue_golomb_31(gb);
 230         }
 231     }
 232
 233     return 0;
 234 }
 235
 236 /**
 237  * Parse NAL units of found picture and decode some basic information.
 238  *
 239  * @param s parser context.
 240  * @param avctx codec context.
 241  * @param buf buffer with field/frame data.
 242  * @param buf_size size of the buffer.
 243  */
 244 static inline int parse_nal_units(AVCodecParserContext *s,
 245                                   AVCodecContext *avctx,
 246                                   const uint8_t * const buf, int buf_size)
 247 {
 248     H264ParseContext *p = s->priv_data;
 249     H2645RBSP rbsp = { NULL };
 250     H2645NAL nal = { NULL };
 251     int buf_index, next_avc;
 252     unsigned int pps_id;
 253     unsigned int slice_type;
 254     int state = -1, got_reset = 0;
 255     int q264 = buf_size >=4 && !memcmp("Q264", buf, 4);
 256     int field_poc[2];
 257     int ret;
 258
 259     /* set some sane default values */
 260     s->pict_type         = AV_PICTURE_TYPE_I;
 261     s->key_frame         = 0;
 262     s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
 263
 264     ff_h264_sei_uninit(&p->sei);
 265     p->sei.frame_packing.arrangement_cancel_flag = -1;
 266
 267     if (!buf_size)
 268         return 0;
 269
 270     av_fast_padded_malloc(&rbsp.rbsp_buffer, &rbsp.rbsp_buffer_alloc_size, buf_size);
 271     if (!rbsp.rbsp_buffer)
 272         return AVERROR(ENOMEM);
 273
 274     buf_index     = 0;
 275     next_avc      = p->is_avc ? 0 : buf_size;
 276     for (;;) {
 277         const SPS *sps;
 278         int src_length, consumed, nalsize = 0;
 279
 280         if (buf_index >= next_avc) {
 281             nalsize = get_nalsize(p->nal_length_size, buf, buf_size, &buf_index, avctx);
 282             if (nalsize < 0)
 283                 break;
 284             next_avc = buf_index + nalsize;
 285         } else {
 286             buf_index = find_start_code(buf, buf_size, buf_index, next_avc);
 287             if (buf_index >= buf_size)
 288                 break;
 289             if (buf_index >= next_avc)
 290                 continue;
 291         }
 292         src_length = next_avc - buf_index;
 293
 294         state = buf[buf_index];
 295         switch (state & 0x1f) {
 296         case H264_NAL_SLICE:
 297         case H264_NAL_IDR_SLICE:
 298             // Do not walk the whole buffer just to decode slice header
 299             if ((state & 0x1f) == H264_NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
 300                 /* IDR or disposable slice
 301                  * No need to decode many bytes because MMCOs shall not be present. */
 302                 if (src_length > 60)
 303                     src_length = 60;
 304             } else {
 305                 /* To decode up to MMCOs */
 306                 if (src_length > 1000)
 307                     src_length = 1000;
 308             }
 309             break;
 310         }
 311         consumed = ff_h2645_extract_rbsp(buf + buf_index, src_length, &rbsp, &nal, 1);
 312         if (consumed < 0)
 313             break;
 314
 315         buf_index += consumed;
 316
 317         ret = init_get_bits8(&nal.gb, nal.data, nal.size);
 318         if (ret < 0)
 319             goto fail;
 320         get_bits1(&nal.gb);
 321         nal.ref_idc = get_bits(&nal.gb, 2);
 322         nal.type    = get_bits(&nal.gb, 5);
 323
 324         switch (nal.type) {
 325         case H264_NAL_SPS:
 326             ff_h264_decode_seq_parameter_set(&nal.gb, avctx, &p->ps, 0);
 327             break;
 328         case H264_NAL_PPS:
 329             ff_h264_decode_picture_parameter_set(&nal.gb, avctx, &p->ps,
 330                                                  nal.size_bits);
 331             break;
 332         case H264_NAL_SEI:
 333             ff_h264_sei_decode(&p->sei, &nal.gb, &p->ps, avctx);
 334             break;
 335         case H264_NAL_IDR_SLICE:
 336             s->key_frame = 1;
 337
 338             p->poc.prev_frame_num        = 0;
 339             p->poc.prev_frame_num_offset = 0;
 340             p->poc.prev_poc_msb          =
 341             p->poc.prev_poc_lsb          = 0;
 342         /* fall through */
 343         case H264_NAL_SLICE:
 344             get_ue_golomb_long(&nal.gb);  // skip first_mb_in_slice
 345             slice_type   = get_ue_golomb_31(&nal.gb);
 346             s->pict_type = ff_h264_golomb_to_pict_type[slice_type % 5];
 347             if (p->sei.recovery_point.recovery_frame_cnt >= 0) {
 348                 /* key frame, since recovery_frame_cnt is set */
 349                 s->key_frame = 1;
 350             }
 351             pps_id = get_ue_golomb(&nal.gb);
 352             if (pps_id >= MAX_PPS_COUNT) {
 353                 av_log(avctx, AV_LOG_ERROR,
 354                        "pps_id %u out of range\n", pps_id);
 355                 goto fail;
 356             }
 357             if (!p->ps.pps_list[pps_id]) {
 358                 av_log(avctx, AV_LOG_ERROR,
 359                        "non-existing PPS %u referenced\n", pps_id);
 360                 goto fail;
 361             }
 362
 363             av_buffer_unref(&p->ps.pps_ref);
 364             av_buffer_unref(&p->ps.sps_ref);
 365             p->ps.pps = NULL;
 366             p->ps.sps = NULL;
 367             p->ps.pps_ref = av_buffer_ref(p->ps.pps_list[pps_id]);
 368             if (!p->ps.pps_ref)
 369                 goto fail;
 370             p->ps.pps = (const PPS*)p->ps.pps_ref->data;
 371
 372             if (!p->ps.sps_list[p->ps.pps->sps_id]) {
 373                 av_log(avctx, AV_LOG_ERROR,
 374                        "non-existing SPS %u referenced\n", p->ps.pps->sps_id);
 375                 goto fail;
 376             }
 377
 378             p->ps.sps_ref = av_buffer_ref(p->ps.sps_list[p->ps.pps->sps_id]);
 379             if (!p->ps.sps_ref)
 380                 goto fail;
 381             p->ps.sps = (const SPS*)p->ps.sps_ref->data;
 382
 383             sps = p->ps.sps;
 384
 385             // heuristic to detect non marked keyframes
 386             if (p->ps.sps->ref_frame_count <= 1 && p->ps.pps->ref_count[0] <= 1 && s->pict_type == AV_PICTURE_TYPE_I)
 387                 s->key_frame = 1;
 388
 389             p->poc.frame_num = get_bits(&nal.gb, sps->log2_max_frame_num);
 390
 391             s->coded_width  = 16 * sps->mb_width;
 392             s->coded_height = 16 * sps->mb_height;
 393             s->width        = s->coded_width  - (sps->crop_right + sps->crop_left);
 394             s->height       = s->coded_height - (sps->crop_top   + sps->crop_bottom);
 395             if (s->width <= 0 || s->height <= 0) {
 396                 s->width  = s->coded_width;
 397                 s->height = s->coded_height;
 398             }
 399
 400             switch (sps->bit_depth_luma) {
 401             case 9:
 402                 if (sps->chroma_format_idc == 3)      s->format = AV_PIX_FMT_YUV444P9;
 403                 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P9;
 404                 else                                  s->format = AV_PIX_FMT_YUV420P9;
 405                 break;
 406             case 10:
 407                 if (sps->chroma_format_idc == 3)      s->format = AV_PIX_FMT_YUV444P10;
 408                 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P10;
 409                 else                                  s->format = AV_PIX_FMT_YUV420P10;
 410                 break;
 411             case 8:
 412                 if (sps->chroma_format_idc == 3)      s->format = AV_PIX_FMT_YUV444P;
 413                 else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P;
 414                 else                                  s->format = AV_PIX_FMT_YUV420P;
 415                 break;
 416             default:
 417                 s->format = AV_PIX_FMT_NONE;
 418             }
 419
 420             avctx->profile = ff_h264_get_profile(sps);
 421             avctx->level   = sps->level_idc;
 422
 423             if (sps->frame_mbs_only_flag) {
 424                 p->picture_structure = PICT_FRAME;
 425             } else {
 426                 if (get_bits1(&nal.gb)) { // field_pic_flag
 427                     p->picture_structure = PICT_TOP_FIELD + get_bits1(&nal.gb); // bottom_field_flag
 428                 } else {
 429                     p->picture_structure = PICT_FRAME;
 430                 }
 431             }
 432
 433             if (nal.type == H264_NAL_IDR_SLICE)
 434                 get_ue_golomb_long(&nal.gb); /* idr_pic_id */
 435             if (sps->poc_type == 0) {
 436                 p->poc.poc_lsb = get_bits(&nal.gb, sps->log2_max_poc_lsb);
 437
 438                 if (p->ps.pps->pic_order_present == 1 &&
 439                     p->picture_structure == PICT_FRAME)
 440                     p->poc.delta_poc_bottom = get_se_golomb(&nal.gb);
 441             }
 442
 443             if (sps->poc_type == 1 &&
 444                 !sps->delta_pic_order_always_zero_flag) {
 445                 p->poc.delta_poc[0] = get_se_golomb(&nal.gb);
 446
 447                 if (p->ps.pps->pic_order_present == 1 &&
 448                     p->picture_structure == PICT_FRAME)
 449                     p->poc.delta_poc[1] = get_se_golomb(&nal.gb);
 450             }
 451
 452             /* Decode POC of this picture.
 453              * The prev_ values needed for decoding POC of the next picture are not set here. */
 454             field_poc[0] = field_poc[1] = INT_MAX;
 455             ret = ff_h264_init_poc(field_poc, &s->output_picture_number, sps,
 456                              &p->poc, p->picture_structure, nal.ref_idc);
 457             if (ret < 0)
 458                 goto fail;
 459
 460             /* Continue parsing to check if MMCO_RESET is present.
 461              * FIXME: MMCO_RESET could appear in non-first slice.
 462              *        Maybe, we should parse all undisposable non-IDR slice of this
 463              *        picture until encountering MMCO_RESET in a slice of it. */
 464             if (nal.ref_idc && nal.type != H264_NAL_IDR_SLICE) {
 465                 got_reset = scan_mmco_reset(s, &nal.gb, avctx);
 466                 if (got_reset < 0)
 467                     goto fail;
 468             }
 469
 470             /* Set up the prev_ values for decoding POC of the next picture. */
 471             p->poc.prev_frame_num        = got_reset ? 0 : p->poc.frame_num;
 472             p->poc.prev_frame_num_offset = got_reset ? 0 : p->poc.frame_num_offset;
 473             if (nal.ref_idc != 0) {
 474                 if (!got_reset) {
 475                     p->poc.prev_poc_msb = p->poc.poc_msb;
 476                     p->poc.prev_poc_lsb = p->poc.poc_lsb;
 477                 } else {
 478                     p->poc.prev_poc_msb = 0;
 479                     p->poc.prev_poc_lsb =
 480                         p->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
 481                 }
 482             }
 483
 484             if (sps->pic_struct_present_flag && p->sei.picture_timing.present) {
 485                 switch (p->sei.picture_timing.pic_struct) {
 486                 case H264_SEI_PIC_STRUCT_TOP_FIELD:
 487                 case H264_SEI_PIC_STRUCT_BOTTOM_FIELD:
 488                     s->repeat_pict = 0;
 489                     break;
 490                 case H264_SEI_PIC_STRUCT_FRAME:
 491                 case H264_SEI_PIC_STRUCT_TOP_BOTTOM:
 492                 case H264_SEI_PIC_STRUCT_BOTTOM_TOP:
 493                     s->repeat_pict = 1;
 494                     break;
 495                 case H264_SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 496                 case H264_SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 497                     s->repeat_pict = 2;
 498                     break;
 499                 case H264_SEI_PIC_STRUCT_FRAME_DOUBLING:
 500                     s->repeat_pict = 3;
 501                     break;
 502                 case H264_SEI_PIC_STRUCT_FRAME_TRIPLING:
 503                     s->repeat_pict = 5;
 504                     break;
 505                 default:
 506                     s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
 507                     break;
 508                 }
 509             } else {
 510                 s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
 511             }
 512
 513             if (p->picture_structure == PICT_FRAME) {
 514                 s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
 515                 if (sps->pic_struct_present_flag && p->sei.picture_timing.present) {
 516                     switch (p->sei.picture_timing.pic_struct) {
 517                     case H264_SEI_PIC_STRUCT_TOP_BOTTOM:
 518                     case H264_SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
 519                         s->field_order = AV_FIELD_TT;
 520                         break;
 521                     case H264_SEI_PIC_STRUCT_BOTTOM_TOP:
 522                     case H264_SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
 523                         s->field_order = AV_FIELD_BB;
 524                         break;
 525                     default:
 526                         s->field_order = AV_FIELD_PROGRESSIVE;
 527                         break;
 528                     }
 529                 } else {
 530                     if (field_poc[0] < field_poc[1])
 531                         s->field_order = AV_FIELD_TT;
 532                     else if (field_poc[0] > field_poc[1])
 533                         s->field_order = AV_FIELD_BB;
 534                     else
 535                         s->field_order = AV_FIELD_PROGRESSIVE;
 536                 }
 537             } else {
 538                 if (p->picture_structure == PICT_TOP_FIELD)
 539                     s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
 540                 else
 541                     s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
 542                 if (p->poc.frame_num == p->last_frame_num &&
 543                     p->last_picture_structure != AV_PICTURE_STRUCTURE_UNKNOWN &&
 544                     p->last_picture_structure != AV_PICTURE_STRUCTURE_FRAME &&
 545                     p->last_picture_structure != s->picture_structure) {
 546                     if (p->last_picture_structure == AV_PICTURE_STRUCTURE_TOP_FIELD)
 547                         s->field_order = AV_FIELD_TT;
 548                     else
 549                         s->field_order = AV_FIELD_BB;
 550                 } else {
 551                     s->field_order = AV_FIELD_UNKNOWN;
 552                 }
 553                 p->last_picture_structure = s->picture_structure;
 554                 p->last_frame_num = p->poc.frame_num;
 555             }
 556
 557             av_freep(&rbsp.rbsp_buffer);
 558             return 0; /* no need to evaluate the rest */
 559         }
 560     }
 561     if (q264) {
 562         av_freep(&rbsp.rbsp_buffer);
 563         return 0;
 564     }
 565     /* didn't find a picture! */
 566     av_log(avctx, AV_LOG_ERROR, "missing picture in access unit with size %d\n", buf_size);
 567 fail:
 568     av_freep(&rbsp.rbsp_buffer);
 569     return -1;
 570 }
 571
 572 static int h264_parse(AVCodecParserContext *s,
 573                       AVCodecContext *avctx,
 574                       const uint8_t **poutbuf, int *poutbuf_size,
 575                       const uint8_t *buf, int buf_size)
 576 {
 577     H264ParseContext *p = s->priv_data;
 578     ParseContext *pc = &p->pc;
 579     int next;
 580
 581     if (!p->got_first) {
 582         p->got_first = 1;
 583         if (avctx->extradata_size) {
 584             ff_h264_decode_extradata(avctx->extradata, avctx->extradata_size,
 585                                      &p->ps, &p->is_avc, &p->nal_length_size,
 586                                      avctx->err_recognition, avctx);
 587         }
 588     }
 589
 590     if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
 591         next = buf_size;
 592     } else {
 593         next = h264_find_frame_end(p, buf, buf_size, avctx);
 594
 595         if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
 596             *poutbuf      = NULL;
 597             *poutbuf_size = 0;
 598             return buf_size;
 599         }
 600
 601         if (next < 0 && next != END_NOT_FOUND) {
 602             av_assert1(pc->last_index + next >= 0);
 603             h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next, avctx); // update state
 604         }
 605     }
 606
 607     parse_nal_units(s, avctx, buf, buf_size);
 608
 609     if (avctx->framerate.num)
 610         avctx->time_base = av_inv_q(av_mul_q(avctx->framerate, (AVRational){avctx->ticks_per_frame, 1}));
 611     if (p->sei.picture_timing.cpb_removal_delay >= 0) {
 612         s->dts_sync_point    = p->sei.buffering_period.present;
 613         s->dts_ref_dts_delta = p->sei.picture_timing.cpb_removal_delay;
 614         s->pts_dts_delta     = p->sei.picture_timing.dpb_output_delay;
 615     } else {
 616         s->dts_sync_point    = INT_MIN;
 617         s->dts_ref_dts_delta = INT_MIN;
 618         s->pts_dts_delta     = INT_MIN;
 619     }
 620
 621     if (s->flags & PARSER_FLAG_ONCE) {
 622         s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
 623     }
 624
 625     if (s->dts_sync_point >= 0) {
 626         int64_t den = avctx->time_base.den * (int64_t)avctx->pkt_timebase.num;
 627         if (den > 0) {
 628             int64_t num = avctx->time_base.num * (int64_t)avctx->pkt_timebase.den;
 629             if (s->dts != AV_NOPTS_VALUE) {
 630                 // got DTS from the stream, update reference timestamp
 631                 p->reference_dts = s->dts - av_rescale(s->dts_ref_dts_delta, num, den);
 632             } else if (p->reference_dts != AV_NOPTS_VALUE) {
 633                 // compute DTS based on reference timestamp
 634                 s->dts = p->reference_dts + av_rescale(s->dts_ref_dts_delta, num, den);
 635             }
 636
 637             if (p->reference_dts != AV_NOPTS_VALUE && s->pts == AV_NOPTS_VALUE)
 638                 s->pts = s->dts + av_rescale(s->pts_dts_delta, num, den);
 639
 640             if (s->dts_sync_point > 0)
 641                 p->reference_dts = s->dts; // new reference
 642         }
 643     }
 644
 645     *poutbuf      = buf;
 646     *poutbuf_size = buf_size;
 647     return next;
 648 }
 649
 650 static int h264_split(AVCodecContext *avctx,
 651                       const uint8_t *buf, int buf_size)
 652 {
 653     uint32_t state = -1;
 654     int has_sps    = 0;
 655     int has_pps    = 0;
 656     const uint8_t *ptr = buf, *end = buf + buf_size;
 657     int nalu_type;
 658
 659     while (ptr < end) {
 660         ptr = avpriv_find_start_code(ptr, end, &state);
 661         if ((state & 0xFFFFFF00) != 0x100)
 662             break;
 663         nalu_type = state & 0x1F;
 664         if (nalu_type == H264_NAL_SPS) {
 665             has_sps = 1;
 666         } else if (nalu_type == H264_NAL_PPS)
 667             has_pps = 1;
 668         /* else if (nalu_type == 0x01 ||
 669          *     nalu_type == 0x02 ||
 670          *     nalu_type == 0x05) {
 671          *  }
 672          */
 673         else if ((nalu_type != H264_NAL_SEI || has_pps) &&
 674                   nalu_type != H264_NAL_AUD && nalu_type != H264_NAL_SPS_EXT &&
 675                   nalu_type != 0x0f) {
 676             if (has_sps) {
 677                 while (ptr - 4 > buf && ptr[-5] == 0)
 678                     ptr--;
 679                 return ptr - 4 - buf;
 680             }
 681         }
 682     }
 683
 684     return 0;
 685 }
 686
 687 static void h264_close(AVCodecParserContext *s)
 688 {
 689     H264ParseContext *p = s->priv_data;
 690     ParseContext *pc = &p->pc;
 691
 692     av_freep(&pc->buffer);
 693
 694     ff_h264_sei_uninit(&p->sei);
 695     ff_h264_ps_uninit(&p->ps);
 696 }
 697
 698 static av_cold int init(AVCodecParserContext *s)
 699 {
 700     H264ParseContext *p = s->priv_data;
 701
 702     p->reference_dts = AV_NOPTS_VALUE;
 703     p->last_frame_num = INT_MAX;
 704     ff_h264dsp_init(&p->h264dsp, 8, 1);
 705     return 0;
 706 }
 707
 708 AVCodecParser ff_h264_parser = {
 709     .codec_ids      = { AV_CODEC_ID_H264 },
 710     .priv_data_size = sizeof(H264ParseContext),
 711     .parser_init    = init,
 712     .parser_parse   = h264_parse,
 713     .parser_close   = h264_close,
 714     .split          = h264_split,
 715 };